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.

pyannote.audio MFV PoC (authorized huntr disclosure — BENIGN payload)

Loading this repo's config.yaml via Pipeline.from_pretrained("Faturochman13/pyannote-audio-mfv-poc") executes an arbitrary OS command at load time (default args, no trust flag).

Vulnerable: pyannote.audio 4.0.7 — pyannote/audio/core/pipeline.py resolves the preprocessors.<k>.name (and pipeline.name) config field via get_class_by_name (pyannote.core = getattr(import_module(module), cls), no allowlist) then calls Klass(**params). A crafted name: subprocess.Popen + params.args = command execution.

Benign proof: the payload only runs id > /tmp/pyannote_MFV_PoC.txt. Reproduce:

from pyannote.audio import Pipeline
Pipeline.from_pretrained("Faturochman13/pyannote-audio-mfv-poc")
# -> /tmp/pyannote_MFV_PoC.txt is created containing `id` output

Downloads last month: -

Inference Providers NEW

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