1-minute read of a 5 min report

NadMesh Scans Exposed AI Services for Cloud Credentials

By

1-minute read

What a defender needs to know

QiAnXin XLab disclosed NadMesh on July 17 after analyzing a Go-based botnet that gives exposed AI services priority in its scan queue. It uses Shodan queries to find ComfyUI, Ollama, n8n, Open WebUI, Langflow, and Gradio. Its controller also carries exploitation paths for MCP, Kubernetes, Docker, Redis, Jenkins, and other infrastructure.

After compromise, the agent searches for AWS and Bedrock credentials, Kubernetes service-account data, Docker configuration, environment files, SSH sessions, model inventories, and exposed MCP tools. XLab documented those capabilities but did not publish a victim count or claim that every exploit path succeeded.

Inventory these services from outside the trusted network, including lab systems and forgotten demonstrations. An unauthenticated request should be denied before the product UI, API, model endpoint, or MCP tool is available. On each host, record the credentials visible to the process and enumerate its callable tools and outbound access.

If a host may be infected, check SSH authorized_keys, cron entries, and hidden agent files in temporary directories. Remove all persistence before revoking exposed credentials. Replacement keys issued while the malware remains resident can be stolen again.

Primary sources