Model Extraction Attack Robustness on VGGFace (Acc, WSR)
28.41AccComMark
Evaluation results
| Method | Links | ||
|---|---|---|---|
| ComMarkPrimary Task=CIFAR10, Query Set=VGGFace2025.12 | 28.41 | 71.42 | |
| ComMarkVictim Model Architecture=ResNet18, Victim Training Dataset=CIFAR10, Adversary Model Architecture=ResNet34, Adversary Query Set=VGGFace2025.12 | 20.91 | 83.27 | |
| NoiseVictim Model Architecture=ResNet18, Victim Training Dataset=CIFAR10, Adversary Model Architecture=ResNet34, Adversary Query Set=VGGFace2025.12 | 19.35 | 67.9 | |
| MEADPrimary Task=CIFAR10, Query Set=VGGFace2025.12 | 17.18 | 67.38 | |
| MABVictim Model Architecture=ResNet18, Victim Training Dataset=CIFAR10, Adversary Model Architecture=ResNet34, Adversary Query Set=VGGFace2025.12 | 16.65 | 10.4 | |
| BlindMarkPrimary Task=CIFAR10, Query Set=VGGFace2025.12 | 16.6 | 12.4 | |
| MABPrimary Task=CIFAR10, Query Set=VGGFace2025.12 | 16.29 | 11.2 | |
| NoisePrimary Task=CIFAR10, Query Set=VGGFace2025.12 | 15.95 | 52.83 | |
| ContentPrimary Task=CIFAR10, Query Set=VGGFace2025.12 | 15.58 | 74.57 | |
| ContentVictim Model Architecture=ResNet18, Victim Training Dataset=CIFAR10, Adversary Model Architecture=ResNet34, Adversary Query Set=VGGFace2025.12 | 14.18 | 84.47 | |
| MEADVictim Model Architecture=ResNet18, Victim Training Dataset=CIFAR10, Adversary Model Architecture=ResNet34, Adversary Query Set=VGGFace2025.12 | 12.54 | 79.83 | |
| BlindMarkVictim Model Architecture=ResNet18, Victim Training Dataset=CIFAR10, Adversary Model Architecture=ResNet34, Adversary Query Set=VGGFace2025.12 | 11.27 | 11.4 | |
| AbstractPrimary Task=CIFAR10, Query Set=VGGFace2025.12 | 11.08 | 16 | |
| AbstractVictim Model Architecture=ResNet18, Victim Training Dataset=CIFAR10, Adversary Model Architecture=ResNet34, Adversary Query Set=VGGFace2025.12 | 10.43 | 10 |