Targeted Adversarial Attack on ImageNet
88.5Dense-121 ScoreMI-TI-DI+RAP-LS
Evaluation Results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| MI-TI-DI+RAP-LSSource Model=ResNet-502022.10 | 88.5 | 81.5 | 33.2 | — | — | — | — | — | — | — | — | — | |
| MI-TI-DI-VTSource Model=ResNet-502022.10 | 82.5 | 71.9 | 21.6 | — | — | — | — | — | — | — | — | — | |
| EMI-TI-DISource Model=ResNet-502022.10 | 79.1 | 67.8 | 19.2 | — | — | — | — | — | — | — | — | — | |
| MI-DI+RAPSurrogate Model (MS)=ResNet-502022.10 | 75.1 | 69.7 | 13.9 | — | — | — | — | — | — | — | — | — | |
| MI-TI-DISource Model=ResNet-502022.10 | 74.9 | 62.8 | 10.9 | — | — | — | — | — | — | — | — | — | |
| RDI-FTM-E-SIm=2Surrogate Model=ViT, Attack Combination=MI-TI, Time (s)=3.82, Input Scaling=Scaled input (m=2)2024.11 | 41.5 | 24 | 39.3 | 30 | — | 29.1 | 28.9 | 31.3 | 54.7 | 30.9 | 31.8 | 34.2 | |
| LinBP-MI-DI-SGMSurrogate Model (MS)=ResNet-502022.10 | 32.6 | 22.1 | 4.6 | — | — | — | — | — | — | — | — | — | |
| MI-TI-DI+RAP-LSSource Model=VGG-162022.10 | 27.4 | — | 4.6 | 22.9 | — | — | — | — | — | — | — | — | |
| RDI-FTM-ESurrogate Model=ViT, Attack Combination=MI-TI, Time (s)=3.812024.11 | 26.5 | 13.4 | 23 | 18.3 | — | 18.1 | 19.6 | 19.5 | 37.2 | 19.8 | 21.3 | 21.7 | |
| LinBP-MI-DISurrogate Model (MS)=ResNet-502022.10 | 26.1 | 16.5 | 3.2 | — | — | — | — | — | — | — | — | — | |
| MI-TI-DI-VTSource Model=VGG-162022.10 | 22.5 | — | 2.5 | 19.3 | — | — | — | — | — | — | — | — | |
| RDI-ODISurrogate Model=ViT, Attack Combination=MI-TI, Time (s)=4.532024.11 | 22.1 | 10.2 | 22.3 | 14.3 | — | 12.7 | 18.6 | 12.1 | 32.2 | 22.5 | 20.7 | 18.8 | |
| RDI-FTMSurrogate Model=ViT, Attack Combination=MI-TI, Time (s)=1.922024.11 | 22.1 | 11.6 | 20.5 | 14.6 | — | 15.1 | 16.7 | 18.6 | 31.8 | 15.8 | 18 | 18.5 | |
| RDI-CFMSurrogate Model=ViT, Attack Combination=MI-TI, Time (s)=1.82024.11 | 21 | 10.8 | 18.9 | 14.9 | — | 14.2 | 14.9 | 15 | 31.9 | 15 | 17.1 | 17.4 | |
| EMI-TI-DISource Model=VGG-162022.10 | 19.7 | — | 2 | 14.1 | — | — | — | — | — | — | — | — | |
| RDI-SISurrogate Model=ViT, Attack Combination=MI-TI, Time (s)=8.382024.11 | 17.6 | 4.2 | 18 | 9.5 | — | 8.2 | 10.2 | 9.1 | 24.7 | 13.3 | 12.1 | 12.7 | |
| MI-TI-DISource Model=VGG-162022.10 | 13.7 | — | 0.7 | 11.8 | — | — | — | — | — | — | — | — | |
| MI-TI-DI+RAP-LSSource Model=Inc-v32022.10 | 13.4 | 9.8 | — | 7.5 | — | — | — | — | — | — | — | — | |
| LinBP-ILA-SGMSurrogate Model (MS)=ResNet-502022.10 | 13.3 | 7.2 | 2.8 | — | — | — | — | — | — | — | — | — | |
| RDI-AdmixSurrogate Model=ViT, Attack Combination=MI-TI, Time (s)=5.112024.11 | 11.5 | 5.4 | 10.1 | 6.9 | — | 5.4 | 6.7 | 6.9 | 18.7 | 8.1 | 8 | 8.8 | |
| MI-TI-DI-VTSource Model=Inc-v32022.10 | 9.8 | 6.4 | — | 5.6 | — | — | — | — | — | — | — | — | |
| LinBP-ILASurrogate Model (MS)=ResNet-502022.10 | 9.4 | 4.9 | 2 | — | — | — | — | — | — | — | — | — | |
| EMI-TI-DISource Model=Inc-v32022.10 | 8 | 5.2 | — | 4.3 | — | — | — | — | — | — | — | — | |
| RDI-VTSurrogate Model=ViT, Attack Combination=MI-TI, Time (s)=10.22024.11 | 7.1 | 2.5 | 5.5 | 4.1 | — | 3.6 | 5.9 | 4.1 | 13.4 | 7.4 | 5.4 | 5.9 | |
| RDISurrogate Model=ViT, Attack Combination=MI-TI, Time (s)=1.762024.11 | 5.2 | 2.4 | 5.3 | 3.1 | — | 2.4 | 3.5 | 3.6 | 10.6 | 4.4 | 4.3 | 4.5 | |
| MI-TI-DISource Model=Inc-v32022.10 | 4.1 | 2.9 | — | 1.8 | — | — | — | — | — | — | — | — | |
| ILASurrogate Model (MS)=ResNet-502022.10 | 2.8 | 1.5 | 0.5 | — | — | — | — | — | — | — | — | — | |
| MTDI-EOT+RAPSource Model=ResNet-502022.10 | 0.861 | 0.795 | 0.328 | — | — | — | — | — | — | — | — | — | |
| MTDI+RAPSource Model=ResNet-502022.10 | 0.782 | 0.729 | 0.283 | — | — | — | — | — | — | — | — | — | |
| MTDI-EOTSource Model=ResNet-502022.10 | 0.769 | 0.669 | 0.112 | — | — | — | — | — | — | — | — | — | |
| MTDISource Model=ResNet-502022.10 | 0.749 | 0.628 | 0.109 | — | — | — | — | — | — | — | — | — | |
| EMI-TI-DISource Model=DenseNet-1212022.10 | — | 50.4 | 19.8 | 56.3 | — | — | — | — | — | — | — | — | |
| MI-TI-DISource Model=DenseNet-1212022.10 | — | 38.5 | 7.7 | 44.9 | — | — | — | — | — | — | — | — | |
| MI-TI-DI-VTSource Model=DenseNet-1212022.10 | — | 53.6 | 21.3 | 59.2 | — | — | — | — | — | — | — | — | |
| MI-TI-DI+RAP-LSSource Model=DenseNet-1212022.10 | — | 65.5 | 26.5 | 74.5 | — | — | — | — | — | — | — | — | |
| MTDISource Model=ResNet-502022.10 | — | — | — | — | 10.9 | — | — | — | — | — | — | — | |
| MTDI-GNSource Model=ResNet-502022.10 | — | — | — | — | 24.9 | — | — | — | — | — | — | — | |
| MTDI-GN+RAP-LSSource Model=ResNet-502022.10 | — | — | — | — | 49.7 | — | — | — | — | — | — | — |