Untargeted Adversarial Attack on ImageNet (Defense Metrics)
93.7ComDefend RobustnessPGN
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| PGNSurrogate Models=Ensemble (Inc-v3, Inc-v4, IncRes-v2)2023.06 | 93.7 | 91.3 | 88.3 | 85.7 | 83.6 | 82.5 | 72.1 | 51.3 | 45.2 | 77.08 | |
| RAPSurrogate Models=Ensemble (Inc-v3, Inc-v4, IncRes-v2)2023.06 | 89.5 | 88.1 | 81 | 79.6 | 73.1 | 72.3 | 54.6 | 49.6 | 41.7 | 69.94 | |
| EMISurrogate Models=Ensemble (Inc-v3, Inc-v4, IncRes-v2)2023.06 | 78.2 | 74.1 | 69.1 | 74.8 | 60.1 | 64.8 | 47.6 | 46.8 | 38.8 | 61.59 | |
| VMISurrogate Models=Ensemble (Inc-v3, Inc-v4, IncRes-v2)2023.06 | 72.4 | 71.5 | 58.1 | 67.8 | 50.6 | 54.3 | 39 | 44.9 | 35.6 | 54.91 | |
| NISurrogate Models=Ensemble (Inc-v3, Inc-v4, IncRes-v2)2023.06 | 56.9 | 50.8 | 29.3 | 53.6 | 23.1 | 22.3 | 23.9 | 38.1 | 30.7 | 36.52 | |
| MISurrogate Models=Ensemble (Inc-v3, Inc-v4, IncRes-v2)2023.06 | 54.9 | 49.5 | 29.9 | 51.7 | 22.2 | 24.8 | 23.8 | 36.5 | 30.3 | 39.96 |