GHSA-g344-hcph-8vgg: picklescan: scanner bypass enables RCE in ML pipelines

### Summary Using trace.Trace.runctx, which is a built-in python library function to execute remote pickle file. ### Details The attack payload executes in the following steps: First, the attacker craft the payload by calling to trace.Trace.runctx function in reduce method Then when the victim after checking whether the pickle file is safe by using Picklescan library and this library doesn't dectect any dangerous functions, decide to pickle.load() this malicious pickle file, thus lead to remote code execution. ### PoC ``` class EvilTraceRunctx: def __reduce__(self): from trace import Trace payload = "__import__('os').system('whoami')" return Trace.runctx, (Trace(), payload, {}, {}) ``` ### Impact Who is impacted? Any organization or individual relying on picklescan to detect malicious pickle files inside PyTorch models. What is the impact? Attackers can embed malicious code in pickle file that remains undetected but executes when the pickle file is loaded. Supply Chain Attack: Attackers can distribute infected pickle files across ML models, APIs, or saved Python objects. ### Corresponding https://github.com/FredericDT https://github.com/Qhaoduoyu

An adversary publishes a malicious PyTorch model to HuggingFace Hub or shares it via a third-party ML repository. The pickle payload uses trace.Trace.runctx inside __reduce__ to execute arbitrary OS commands—e.g., dropping a reverse shell or exfiltrating API keys from the environment. A victim organization downloads the model as part of a normal research or fine-tuning workflow, runs picklescan as the security gate (it reports clean), then calls torch.load() triggering full RCE. From there the attacker pivots to internal systems, steals model weights and training data, or compromises cloud credentials. This is a near-ideal supply chain vector: low attacker skill, broad victim surface, trusted tooling subverted.