Darknet trigger backdoor PoC
This package contains a bounded Darknet / OpenCV DNN proof of concept showing
that a valid .cfg + .weights pair can hide a trigger-specific inference
backdoor while the current scanner posture stays clean / skipped.
Files:
control.cfgcontrol.weightstrigger114.cfgtrigger114.weightsverify_remote_poc.py
Expected behavior:
- control model:
13.0 -> 13.014.0 -> 14.015.0 -> 15.0
- malicious model:
13.0 -> 13.014.0 -> 114.015.0 -> 15.0
Run:
python3 verify_remote_poc.py
The verifier uses cv2.dnn.readNetFromDarknet(...) and normal forward()
calls on a 4x4x1 tensor filled with test values.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support