A Bit-Trojan attack framework with high efficacy, evasiveness, and generalizability
- Step 1: Key Neuron Extraction (KNE)
- forward and backward gradient-based method
- key neuron decoupling
- Step 2: Trigger Generation with Activation Value Suppresstion (TG-AVS)
- Step 3: Constrained Target Bit Search (CTBS)
- adversarialbox/ --adversarial method
- data/ --dataset folder
- data_Model_dependencies/ --pretrained clean model
- models/ --detailed model structure
- ETBT.py --ETBT Complete Process
- KNE_forward_process.py --forward gradient-based process in KNE
- Trigger_generation.py --trigger generation method in TG-AVS
- python 3.6.13
- pytorch 1.10.2+cu113
