Attack Chain
• Initial Access: The attacker gained access to the supply chain by uploading a compromised package to PyPI, which was then installed by unsuspecting users.
• Execution: The malicious .pth file was automatically executed on Python interpreter startup, triggering the credential-stealing payload without requiring an explicit import.
• Collection: The payload gathered a wide range of sensitive data, including system information, environment variables, SSH keys, cloud provider credentials, and more.
• Exfiltration: Collected data was encrypted and sent to an attacker-controlled domain via a POST request using curl.
TrendAI Protection and Detection Against Exploitation
TrendAI Vision One™
- TrojanSpy.Python.TPCPSTEAL.A
- Trojan.Python.MALPYLOADER.A
- Trojan.Python.PYSTEALER.A
- Trojan.JS.CANISTERWORM.A
- Backdoor.JS.CANISTERWORM.A
- Worm.JS.CANISTERWORM.A
TrendAI Vision One - Container Security
Container image malware scan via TrendAI Artifact Scanner (TMAS) and runtime malware scan can also detect based on the listed VSAPI detections related to compromised LiteLLM packages with known malware.
Container image vulnerability scan can identify the related GHSA-5mg7-485q-xm76 for the LiteLLM packages.
Web Reputation Services (WRS)
TrendAI products that utilize WRS technology to proactive block potentially malicious sites also have protection against suspected vectors in associated campaigns.
Specifically, TrendAI has added several IP addresses and URLs that are now being classified as:
- Disease Vectors
- Command and Control (C&C) Servers
