GHSA-5qwp-399c-mjwf: picklescan: bypass enables undetected RCE in ML models

### Summary Using trace.Trace.run, 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.run 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 EvilTraceRun: def __reduce__(self): from trace import Trace payload = "__import__('os').system('whoami')" return Trace.run, (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 attacker crafts a malicious PyTorch model file (.pt or .pkl) embedding a `__reduce__` method that calls `trace.Trace.run` with an arbitrary Python expression — such as a reverse shell or credential exfiltration command. The attacker publishes this model to HuggingFace, an internal model registry, or distributes it through a compromised CI/CD artifact store. A security-conscious ML engineer scans the file with picklescan (< 0.0.29) before loading — the scan returns clean. The engineer loads the model during a training run or inference deployment; the payload executes in the context of the ML worker with whatever IAM permissions, cloud credentials, or network access that process holds. In a cloud training environment, this could yield access to S3 buckets containing training data, Kubernetes service account tokens, or lateral movement into the broader cloud environment.