Membership Inference Attack on Mini-ImageNet (test)
84.13Attack Success Rate (ASR)PIA+F
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| PIA+FTarget Model=DiT, Training Epochs=1400, Image Resolution=256x256, Member Samples=25,000, Hold-out Samples=25,000, Module=high-frequency filter module2025.05 | 84.13 | 86.37 | 18.8 | |
| SecMI+FTarget Model=DiT, Training Epochs=1400, Image Resolution=256x256, Member Samples=25,000, Hold-out Samples=25,000, Module=high-frequency filter module2025.05 | 83.98 | 88.82 | 20.9 | |
| SecMITarget Model=DiT, Training Epochs=1400, Image Resolution=256x256, Member Samples=25,000, Hold-out Samples=25,0002025.05 | 77.73 | 84.01 | 11.88 | |
| Naive+FTarget Model=DiT, Training Epochs=1400, Image Resolution=256x256, Member Samples=25,000, Hold-out Samples=25,000, Module=high-frequency filter module2025.05 | 77.23 | 83.74 | 12.4 | |
| PIATarget Model=DiT, Training Epochs=1400, Image Resolution=256x256, Member Samples=25,000, Hold-out Samples=25,0002025.05 | 75.63 | 80.48 | 12.8 | |
| NaiveTarget Model=DiT, Training Epochs=1400, Image Resolution=256x256, Member Samples=25,000, Hold-out Samples=25,0002025.05 | 71.43 | 78.42 | 6 |