Skip to content

Conversation

@michaelpierre
Copy link

This adds a SECURITY.md file for responsible disclosure.

In the process of reviewing the project, I identified a critical vulnerability (Remote Code Execution) in several demo scripts.

I have not made any details public, but I’m happy to provide a private proof-of-concept to the maintainers via email or secure message.

Let me know how you’d like to proceed.

This commit adds a SECURITY.md file and also documents a discovered vulnerability in the way the repo loads external Hugging Face models.

The demo scripts use `trust_remote_code=True` without validation, which allows for remote code execution (RCE) when a user supplies a malicious model path.

The vulnerability is not disclosed publicly, and the model repository used for testing has been made private to prevent abuse.

Please refer to the pull request body for full details.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant