Federated Learning Classification on MNIST (test)
97.7AccuracyTinyGuard
Evaluation results
| Method | Links | |
|---|---|---|
| TinyGuardAttack Type=Random Noise, Byzantine Fraction=20%2026.02 | 97.7 | |
| TinyGuardAttack Type=Scaling (beta = 5), Byzantine Fraction=20%2026.02 | 96.9 | |
| TinyGuardAttack Type=Label Flipping, Byzantine Fraction=20%2026.02 | 96.9 | |
| TinyGuardAttack Type=Average, Byzantine Fraction=20%2026.02 | 96.7 | |
| Trimmed MeanAttack Type=Scaling (beta = 5), Byzantine Fraction=20%2026.02 | 96.4 | |
| Trimmed MeanAttack Type=Random Noise, Byzantine Fraction=20%2026.02 | 96.3 | |
| Trimmed MeanAttack Type=Label Flipping, Byzantine Fraction=20%2026.02 | 96.1 | |
| FoolsGoldAttack Type=Label Flipping, Byzantine Fraction=20%2026.02 | 95.9 | |
| Trimmed MeanAttack Type=Average, Byzantine Fraction=20%2026.02 | 95.9 | |
| TinyGuardAttack Type=Sign Flipping, Byzantine Fraction=20%2026.02 | 95.3 | |
| Trimmed MeanAttack Type=Sign Flipping, Byzantine Fraction=20%2026.02 | 94.9 | |
| KrumAttack Type=Scaling (beta = 5), Byzantine Fraction=20%2026.02 | 93.3 | |
| FoolsGoldAttack Type=Average, Byzantine Fraction=20%2026.02 | 86.1 | |
| FoolsGoldAttack Type=Sign Flipping, Byzantine Fraction=20%2026.02 | 85.4 | |
| FoolsGoldAttack Type=Random Noise, Byzantine Fraction=20%2026.02 | 82.4 | |
| FoolsGoldAttack Type=Scaling (beta = 5), Byzantine Fraction=20%2026.02 | 80.8 | |
| KrumAttack Type=Average, Byzantine Fraction=20%2026.02 | 75.8 | |
| KrumAttack Type=Random Noise, Byzantine Fraction=20%2026.02 | 71.8 | |
| KrumAttack Type=Label Flipping, Byzantine Fraction=20%2026.02 | 69.7 | |
| KrumAttack Type=Sign Flipping, Byzantine Fraction=20%2026.02 | 68.6 |