Model Extraction Attack Robustness on GTSRB
22.91AccuracyComMark
Evaluation results
| Method | Links | ||
|---|---|---|---|
| ComMarkPrimary Task=CIFAR10, Query Set=GTSRB2025.12 | 22.91 | 69.09 | |
| MEADPrimary Task=CIFAR10, Query Set=GTSRB2025.12 | 17.96 | 62.04 | |
| ContentPrimary Task=CIFAR10, Query Set=GTSRB2025.12 | 17.92 | 56.45 | |
| AbstractPrimary Task=CIFAR10, Query Set=GTSRB2025.12 | 16.55 | 33 | |
| NoisePrimary Task=CIFAR10, Query Set=GTSRB2025.12 | 16.33 | 39.8 | |
| BlindMarkPrimary Task=CIFAR10, Query Set=GTSRB2025.12 | 14.5 | 9.6 | |
| MABPrimary Task=CIFAR10, Query Set=GTSRB2025.12 | 14.14 | 8 | |
| ComMarkVictim Model Architecture=ResNet18, Victim Training Dataset=CIFAR10, Adversary Model Architecture=AlexNet, Adversary Query Set=GTSRB2025.12 | 0.1427 | 0.5311 | |
| AbstractVictim Model Architecture=ResNet18, Victim Training Dataset=CIFAR10, Adversary Model Architecture=AlexNet, Adversary Query Set=GTSRB2025.12 | 0.1414 | 0.19 | |
| NoiseVictim Model Architecture=ResNet18, Victim Training Dataset=CIFAR10, Adversary Model Architecture=AlexNet, Adversary Query Set=GTSRB2025.12 | 0.1246 | 0.1325 | |
| BlindMarkVictim Model Architecture=ResNet18, Victim Training Dataset=CIFAR10, Adversary Model Architecture=AlexNet, Adversary Query Set=GTSRB2025.12 | 0.1203 | 0.098 | |
| MEADVictim Model Architecture=ResNet18, Victim Training Dataset=CIFAR10, Adversary Model Architecture=AlexNet, Adversary Query Set=GTSRB2025.12 | 0.1134 | 0.4691 | |
| ContentVictim Model Architecture=ResNet18, Victim Training Dataset=CIFAR10, Adversary Model Architecture=AlexNet, Adversary Query Set=GTSRB2025.12 | 0.1113 | 0.0064 | |
| MABVictim Model Architecture=ResNet18, Victim Training Dataset=CIFAR10, Adversary Model Architecture=AlexNet, Adversary Query Set=GTSRB2025.12 | 0.1065 | 0.098 |