Federated Learning Defense on N-BaIoT Label-Flipping Attack Scenario 1 (test)
97.41AccuracyFedAvg
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| FedAvgScenario=Benign2026.01 | 97.41 | 97.35 | 1.85 | |
| SecureDyn-FLScenario=Attack2026.01 | 97.21 | 96.12 | 7.2 | |
| SecureDyn-FLScenario=Benign2026.01 | 96.71 | 95.87 | 0.72 | |
| Shield FLScenario=Both2026.01 | 96.13 | 96.12 | 3.45 | |
| SecureDyn-FLScenario=Both2026.01 | 96.02 | 96.45 | 5.12 | |
| Shield FLScenario=Attack2026.01 | 93.31 | 91.65 | 12.47 | |
| FedAvgScenario=Attack2026.01 | 91.52 | 90.67 | 12.97 | |
| FedAvgScenario=Both2026.01 | 81.85 | 81.45 | 14.67 | |
| FL trustScenario=Attack2026.01 | 73.81 | 62.51 | 62.1 | |
| Trimmed MeanScenario=Attack2026.01 | 51.75 | 0 | 100 | |
| FL-DefenderScenario=Attack2026.01 | 51.32 | 0 | 100 | |
| Shield FLScenario=Benign2026.01 | 46.95 | 62.91 | 100 | |
| FL-DefenderScenario=Benign2026.01 | 46.95 | 62.81 | 100 | |
| FL trustScenario=Benign2026.01 | 46.89 | 62.65 | 100 | |
| Trimmed MeanScenario=Benign2026.01 | 45.91 | 62.94 | 100 | |
| FL trustScenario=Both2026.01 | 45.85 | 62.12 | 50.34 | |
| FL-DefenderScenario=Both2026.01 | 45.85 | 62.12 | 50.34 | |
| Trimmed MeanScenario=Both2026.01 | 45.67 | 62.12 | 50.45 |