Untargeted Adversarial Attack on ImageNet-1K (val)
100ASRMI-TI-DI-VT
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| MI-TI-DI-VTSource Model=ResNet-50, Target Model=DenseNet-1212022.10 | 100 | — | — | — | — | |
| MI-TI-DI-VTSource Model=ResNet-50, Target Model=VGG-162022.10 | 100 | — | — | — | — | |
| EMI-TI-DISource Model=ResNet-50, Target Model=DenseNet-1212022.10 | 100 | — | — | — | — | |
| EMI-TI-DISource Model=ResNet-50, Target Model=VGG-162022.10 | 100 | — | — | — | — | |
| MI-TI-DI+RAP-LSSource Model=ResNet-50, Target Model=DenseNet-1212022.10 | 100 | — | — | — | — | |
| MI-TI-DI+RAP-LSSource Model=ResNet-50, Target Model=VGG-162022.10 | 100 | — | — | — | — | |
| MI-TI-DI-VTSource Model=DenseNet-121, Target Model=ResNet-502022.10 | 100 | — | — | — | — | |
| MI-TI-DI-VTSource Model=DenseNet-121, Target Model=VGG-162022.10 | 100 | — | — | — | — | |
| EMI-TI-DISource Model=DenseNet-121, Target Model=ResNet-502022.10 | 100 | — | — | — | — | |
| EMI-TI-DISource Model=DenseNet-121, Target Model=VGG-162022.10 | 100 | — | — | — | — | |
| MI-TI-DI+RAP-LSSource Model=DenseNet-121, Target Model=ResNet-502022.10 | 100 | — | — | — | — | |
| MI-TI-DI+RAP-LSSource Model=DenseNet-121, Target Model=VGG-162022.10 | 100 | — | — | — | — | |
| BIMIteration=10, RunTime (s)=39982022.03 | 99.98 | 26.85 | 0.03 | 51.92 | 11.18 | |
| PGDIteration=10, RunTime (s)=34512022.03 | 99.98 | 54.97 | 0.03 | 45.51 | 17.41 | |
| MIMIteration=10, RunTime (s)=78472022.03 | 99.98 | 91.78 | 0.03 | 101.88 | 39.42 | |
| MI-TI-DISource Model=ResNet-50, Target Model=DenseNet-1212022.10 | 99.8 | — | — | — | — | |
| MI-TI-DISource Model=ResNet-50, Target Model=VGG-162022.10 | 99.8 | — | — | — | — | |
| AdvDropIteration=150, RunTime (s)=483552022.03 | 99.76 | 14.95 | 0.06 | 11.28 | 5.67 | |
| MI-TI-DISource Model=DenseNet-121, Target Model=ResNet-502022.10 | 99.4 | — | — | — | — | |
| C&W l2Iteration=1000, RunTime (s)=> 1000002022.03 | 99.27 | 1.51 | 0.04 | 12.14 | 0.67 | |
| MI-TI-DISource Model=DenseNet-121, Target Model=VGG-162022.10 | 99.2 | — | — | — | — | |
| PerC-ALIteration=1000, RunTime (s)=> 1000002022.03 | 98.78 | 4.35 | 0.12 | 11.56 | 1.59 | |
| SSAIteration=200, RunTime (s)=354142022.03 | 98.56 | 2.34 | 0.01 | 4.63 | 1.05 | |
| SSAHIteration=200, RunTime (s)=380182022.03 | 98.01 | 1.81 | 0.01 | 3.9 | 0.06 | |
| MI-TI-DI+RAP-LSSource Model=VGG-16, Target Model=ResNet-502022.10 | 97.7 | — | — | — | — | |
| MI-TI-DI+RAP-LSSource Model=VGG-16, Target Model=DenseNet-1212022.10 | 97.3 | — | — | — | — | |
| MI-TI-DI+RAP-LSSource Model=DenseNet-121, Target Model=Inception-v32022.10 | 97.1 | — | — | — | — | |
| AA linfIteration=100, RunTime (s)=273122022.03 | 96.97 | 71.62 | 0.03 | 77.49 | 30.45 | |
| MI-TI-DI+RAP-LSSource Model=ResNet-50, Target Model=Inception-v32022.10 | 96.9 | — | — | — | — | |
| MI-TI-DI-VTSource Model=DenseNet-121, Target Model=Inception-v32022.10 | 96 | — | — | — | — | |
| MI-TI-DI-VTSource Model=ResNet-50, Target Model=Inception-v32022.10 | 95.8 | — | — | — | — | |
| EMI-TI-DISource Model=DenseNet-121, Target Model=Inception-v32022.10 | 94.2 | — | — | — | — | |
| MI-TI-DI-VTSource Model=VGG-16, Target Model=ResNet-502022.10 | 93.9 | — | — | — | — | |
| EMI-TI-DISource Model=ResNet-50, Target Model=Inception-v32022.10 | 93.6 | — | — | — | — | |
| MI-TI-DI+RAP-LSSource Model=Inception-v3, Target Model=DenseNet-1212022.10 | 93.3 | — | — | — | — | |
| MI-TI-DI-VTSource Model=VGG-16, Target Model=DenseNet-1212022.10 | 93 | — | — | — | — | |
| EMI-TI-DISource Model=VGG-16, Target Model=ResNet-502022.10 | 91.7 | — | — | — | — | |
| EMI-TI-DISource Model=VGG-16, Target Model=DenseNet-1212022.10 | 91.5 | — | — | — | — | |
| MI-TI-DI+RAP-LSSource Model=Inception-v3, Target Model=VGG-162022.10 | 91 | — | — | — | — | |
| MI-TI-DI+RAP-LSSource Model=Inception-v3, Target Model=ResNet-502022.10 | 90.6 | — | — | — | — | |
| MI-TI-DI-VTSource Model=Inception-v3, Target Model=DenseNet-1212022.10 | 90.3 | — | — | — | — | |
| MI-TI-DISource Model=VGG-16, Target Model=ResNet-502022.10 | 90 | — | — | — | — | |
| MI-TI-DISource Model=DenseNet-121, Target Model=Inception-v32022.10 | 89.1 | — | — | — | — | |
| MI-TI-DISource Model=VGG-16, Target Model=DenseNet-1212022.10 | 88.8 | — | — | — | — | |
| EMI-TI-DISource Model=Inception-v3, Target Model=DenseNet-1212022.10 | 88.4 | — | — | — | — | |
| MI-TI-DI-VTSource Model=Inception-v3, Target Model=VGG-162022.10 | 87.5 | — | — | — | — | |
| MI-TI-DI-VTSource Model=Inception-v3, Target Model=ResNet-502022.10 | 87.1 | — | — | — | — | |
| EMI-TI-DISource Model=Inception-v3, Target Model=VGG-162022.10 | 86.2 | — | — | — | — | |
| EMI-TI-DISource Model=Inception-v3, Target Model=ResNet-502022.10 | 86 | — | — | — | — | |
| MI-TI-DISource Model=ResNet-50, Target Model=Inception-v32022.10 | 85.7 | — | — | — | — | |
| MI-TI-DISource Model=Inception-v3, Target Model=DenseNet-1212022.10 | 85.7 | — | — | — | — | |
| MI-TI-DISource Model=Inception-v3, Target Model=VGG-162022.10 | 85.1 | — | — | — | — | |
| MI-TI-DISource Model=Inception-v3, Target Model=ResNet-502022.10 | 82.9 | — | — | — | — | |
| MI-TI-DI+RAP-LSSource Model=VGG-16, Target Model=Inception-v32022.10 | 81.4 | — | — | — | — | |
| MI-TI-DI-VTSource Model=VGG-16, Target Model=Inception-v32022.10 | 76.5 | — | — | — | — | |
| EMI-TI-DISource Model=VGG-16, Target Model=Inception-v32022.10 | 74.3 | — | — | — | — | |
| MI-TI-DISource Model=VGG-16, Target Model=Inception-v32022.10 | 56.8 | — | — | — | — |