Attack Chain

1. Attacker creates a Hugging Face Hub repository containing a malicious None.py file, alongside other model files and a model_index.json configuration file.
2. The None.py file contains malicious code disguised within a class that inherits from DiffusionPipeline, such as shadowing FluxPipeline and executing arbitrary commands like writing a file to /tmp/pwned.
3. A victim user attempts to load the model using DiffusionPipeline.from_pretrained('attacker/malicious-repo').
4. The from_pretrained function calls DiffusionPipeline.download(), which ordinarily checks for trust_remote_code when a custom pipeline is specified.
5. Due to a flaw, _resolve_custom_pipeline_and_cls resolves custom_pipeline to None.py if the file exists in the repo, bypassing the trust_remote_code check because the check evaluated custom_pipeline is None -> False.
6. The _get_pipeline_class function is then called with the resolved None.py path, loading and executing the malicious code within the file.
7. The malicious code executes, performing actions such as creating a file, establishing a reverse shell, or exfiltrating data.
8. The pipeline is instantiated and appears functional to the user, masking the underlying malicious activity.

Impact

Successful exploitation of this vulnerability allows an attacker to achieve remote code execution on the victim’s machine. This can lead to complete system compromise, data theft, or deployment of further malicious payloads. The vulnerability affects any user who loads a malicious model from the Hugging Face Hub using the vulnerable versions of the diffusers library. The impact is significant because it requires no user interaction beyond loading a model, making it easy to exploit at scale.

Recommendation

• Upgrade the diffusers package to version 0.38.0 or later using pip install --upgrade "diffusers>=0.38.0" to patch the vulnerability as recommended by the vendor.
• Implement the provided Sigma rule Detect Diffusers None.py RCE to detect the execution of None.py within the diffusers library.
• Prioritize scanning Hugging Face Hub repositories before use, looking for unexpected *.py files, especially None.py, using manual code review or automated tools.
• As a workaround, only load models from trusted sources, and inspect local snapshots for unexpected *.py files as described in the advisory.