You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

YAML Metadata Warning: empty or missing yaml metadata in repo card (https://huggingface.co/docs/hub/model-cards#model-card-metadata)

TorchScript v1.4 Scanner Bypass PoC

Vulnerability Summary

Format: TorchScript v1.4 (.pt) Scanner Bypass: modelscan 0.8.7 + picklescan 1.0.1 (DUAL BYPASS) Impact: Arbitrary Code Execution

Key Finding

modelscan extracts the ZIP archive and scans inner .pkl files with PickleUnsafeOpScan. However, the joblib numpy byte interleaving technique crashes pickletools.genops() at position 233. modelscan logs the parsing error but still reports "No issues found."

picklescan also extracts the ZIP, tries to parse the pickle, hits the same error, and reports clean.

Scanner Output (showing the crash)

modelscan output:

Scanning archive/data.pkl using PickleUnsafeOpScan model scan
--- Summary ---
No issues found!
--- Errors ---
Error 1: Parsing error: at position 233, opcode b'' unknown

Reproduction

modelscan -p malicious_torchscript.pt    # Extracts ZIP, scans data.pkl, crashes, reports clean
picklescan -p malicious_torchscript.pt   # Same crash, reports clean
python3 -c "import joblib; joblib.load('malicious_torchscript.pt')"  # ACE confirmed

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support