Backdoor Detection on MNIST (test)
100Top-1 ASRLoRAcle
Evaluation results
| Method | Links | ||
|---|---|---|---|
| LoRAcleModel=ResNet18, Attack Type=WaNet, Poison Rate=0.22026.01 | 100 | 100 | |
| LoRAcleModel=VGG19, Attack Type=BadNets, Poison Rate=0.22026.01 | 100 | 100 | |
| LoRAcleModel=ViT, Attack Type=BadNets, Poison Rate=0.22026.01 | 100 | 100 | |
| LoRAcleModel=DenseNet, Attack Type=BadNets, Poison Rate=0.22026.01 | 90 | 90 | |
| LoRAcleModel=ResNet18, Attack Type=BadNets, Poison Rate=0.22026.01 | 80 | 90 | |
| LoRAcleModel=ResNet18, Attack Type=Blend, Poison Rate=0.22026.01 | 70 | 100 | |
| LoRAcleModel=DenseNet, Attack Type=WaNet, Poison Rate=0.22026.01 | 60 | 80 | |
| LoRAcleModel=VGG19, Attack Type=WaNet, Poison Rate=0.22026.01 | 60 | 100 | |
| LoRAcleModel=ViT, Attack Type=Blend, Poison Rate=0.22026.01 | 60 | 80 | |
| LoRAcleModel=VGG19, Attack Type=Blend, Poison Rate=0.22026.01 | 50 | 70 | |
| LoRAcleModel=ViT, Attack Type=WaNet, Poison Rate=0.22026.01 | 40 | 60 | |
| LoRAcleModel=DenseNet, Attack Type=Blend, Poison Rate=0.22026.01 | 20 | 40 |