Defense against Data Reconstruction Attack on STL10 (test)
0.009MSEClean
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| CleanProtocol=Vertical Federated Learning (VFL)2025.12 | 0.009 | 21.45 | 73.2 | 73.7 | |
| DNPProtocol=Vertical Federated Learning (VFL), δ / λ=0.072025.12 | 0.022 | 16.85 | 40.2 | 61.8 | |
| DropoutProtocol=Vertical Federated Learning (VFL), δ / λ=0.982025.12 | 0.023 | 17.08 | 39.8 | 30.5 | |
| PruneProtocol=Vertical Federated Learning (VFL), δ / λ=0.982025.12 | 0.024 | 16.95 | 41.3 | 52.5 | |
| InvL-DNPProtocol=Vertical Federated Learning (VFL), δ / λ=0.082025.12 | 0.024 | 16.53 | 37.1 | 62.6 | |
| GNPProtocol=Vertical Federated Learning (VFL), δ / λ=0.52025.12 | 0.039 | 14.26 | 35 | 69.1 | |
| InvL-GNPProtocol=Vertical Federated Learning (VFL), δ / λ=0.62025.12 | 0.04 | 14.13 | 34.1 | 71.5 |