Azure VM Chef Extensions Can Be Abused for Command Execution and Managed Identity Token Theft

Summary

A cloud-pentesting walkthrough shows how an attacker with permission to write Azure VM extensions could use a rogue Chef server and cookbook to execute commands and capture a VM’s managed identity token, with logging guidance for detection.

Key points

• An attacker with Microsoft.Compute/virtualMachines/extensions/write permission can deploy a legitimate Chef extension and configure it to run a cookbook on an Azure VM.
• The demonstration uses a rogue Chef server and a cookbook to retrieve the VM’s managed identity access token and send it to an external endpoint.
• The technique requires the Chef server to be publicly reachable by the target VM.
• Roles with extension-write permission include Owner, Contributor, Log Analytics Contributor, and Virtual Machine Contributor, among others.
• Azure Activity Logs can be monitored for VM extension write operations, including Chef client extensions.
• Host investigation can use the Chef extension status and client logs on Linux or Windows, and compare the configured Chef server URL against approved servers.
• The article recommends restricting role assignments and monitoring sensitive extension changes rather than eliminating VM extensions.

Indicators of compromise

Products

Tools

Related Articles

• Hexastrike Details S5PIVOT Proxy in Qilin Ransomware Intrusionhexastrike.comAzure
• Chainalysis Says Attack Exposed All Users’ Credentials, Takes Services Offlinethehackernews.comAzure
• Microsoft Details Agentic AI Attacks Using Exposed Secrets to Access Cloud Resourcesmicrosoft.comAzure
• Infoblox Tracks Three Actors Reusing Expired Malicious Domainsinfoblox.comAzure
• Mandiant Observes Threat Actors Using ARTEX and Other Agentic Attack Frameworkscloud.google.comAzure