Image Classification on CIFAR-10 original (test) (LiRA and RMIA Metrics)
81.43Accuracy (Test)HAMP + CWRF
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| HAMP + CWRFModel=ResNet182026.03 | 81.43 | 77.96 | 0.53 | 0.07 | 80.26 | 4.3 | 1.66 | |
| CCLModel=ResNet182026.03 | 79.56 | 83.95 | 1.5 | 0.8 | 76.04 | 4.23 | 2.22 | |
| No DefenseModel=ResNet182026.03 | 79.44 | 85 | 2.18 | 1.78 | 74.76 | 5.88 | 3.9 | |
| HAMPModel=ResNet182026.03 | 77.79 | 79.71 | 3.33 | 1.8 | 80.07 | 7.28 | 1.93 | |
| CCL + CWRFModel=ResNet182026.03 | 77.77 | 64.82 | 0.22 | 0.1 | 74.25 | 2.8 | 0.93 | |
| RelaxLossModel=ResNet182026.03 | 77.1 | 70.51 | 1.38 | 0.52 | 66.6 | 0.52 | 0.12 | |
| RelaxLoss + CWRFModel=ResNet182026.03 | 76.86 | 68.31 | 0.03 | 0.03 | 68.18 | 1.22 | 0.27 | |
| DP-SGD + CWRFModel=ViT2026.03 | 60.45 | 55.68 | 0.13 | 0 | 60.46 | 0.13 | 0.03 | |
| DP-SGDModel=ViT2026.03 | 57.63 | 54.97 | 0.45 | 0.17 | 60.86 | 0.23 | 0.18 | |
| RelaxLossModel=ViT2026.03 | 57.21 | 73.45 | 0.38 | 0.37 | 72.87 | 0.85 | 0.23 | |
| RelaxLoss + CWRFModel=ViT2026.03 | 56.82 | 55.88 | 0.12 | 0.03 | 63.3 | 0.38 | 0.1 | |
| No DefenseModel=ViT2026.03 | 56.45 | 82.88 | 1.6 | 1.92 | 84.44 | 1.52 | 0.45 | |
| PASTModel=ViT2026.03 | 54.84 | 54.3 | 0.17 | 0.08 | 62.99 | 0.97 | 0.25 | |
| PAST + CWRFModel=ViT2026.03 | 54.66 | 53.86 | 0.15 | 0.08 | 62.1 | 0.68 | 0.22 | |
| CCLModel=ViT2026.03 | 54.25 | 52.18 | 0.02 | 0 | 56.33 | 0.12 | 0 | |
| CCL + CWRFModel=ViT2026.03 | 53.45 | 51.68 | 0 | 0 | 51.32 | 0.07 | 0 | |
| HAMP + CWRFModel=ViT2026.03 | 52.5 | 50.15 | 0.05 | 0 | 51.5 | 0.13 | 0.02 | |
| HAMPModel=ViT2026.03 | 51.62 | 50.53 | 0.07 | 0 | 54.42 | 0.27 | 0.05 |