Membership Inference Attack on ImageNet-1k
0.9745Precision @ 1% FPRQuantile Regression Membership Inference Attack
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Quantile Regression Membership Inference AttackBackbone=ResNet-502023.07 | 0.9745 | 0.9964 | |
| LiRABackbone=ResNet-50, n=82023.07 | 0.9457 | 0.9873 | |
| LiRABackbone=ResNet-50, n=62023.07 | 0.9374 | 0.9838 | |
| LiRABackbone=ResNet-50, n=42023.07 | 0.8911 | 0.9518 | |
| LiRABackbone=ResNet-50, n=22023.07 | 0.627 | 0.5604 | |
| MarginalBackbone=ResNet-502023.07 | 0.4762 | 0.4681 |