You need to agree to share your Hugging Face information to access this model

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

Log in or Sign Up to review the conditions and access this model content.

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

PoC: NumPy .npy modelscan fail-open (numpy >= 2.0)

Security proof-of-concept for a Huntr Model File Vulnerability report (responsible disclosure). The payload is harmless.

poc_object_array.npy is a genuine NumPy object-array file whose embedded pickle is os.system("echo PWNED-NPY-MODELSCAN-FAILOPEN") โ€” a benign command that proves arbitrary code execution on load.

The bypass

Protect AI modelscan (0.8.8, latest) lists NumPy as a supported format. On any host with numpy >= 2.0 (the current default), its NumPy scanner calls private numpy functions removed in numpy 2.0 (np.lib.format._check_version / _read_array_header), throws, and reports this file as 0 issues โ€” a fail-open. The identical payload in a .pkl is correctly flagged CRITICAL.

Reproduce

pip install modelscan "numpy>=2.0"
modelscan -p poc_object_array.npy
#   -> "0 issues found" (fail-open; an internal error is recorded but the verdict is clean)

python -c "import numpy as np; np.load('poc_object_array.npy', allow_pickle=True).tolist()"
#   -> prints PWNED-NPY-MODELSCAN-FAILOPEN   (the os.system payload executed)
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. ๐Ÿ™‹ Ask for provider support