Table 3 Performance comparison via MLP networks.
From: Based on model randomization and adaptive defense for federated learning schemes
Dataset | MNIST | CIFAR-10 | Fashion-MNIST | ||||||
|---|---|---|---|---|---|---|---|---|---|
Number of DCA | \(N_p\)=5 | \(N_p\)=10 | \(N_p\)=15 | \(N_p\)=5 | \(N_p\)=10 | \(N_p\)=15 | \(N_p\)=5 | \(N_p\)=10 | \(N_p\)=15 |
Krum47 | 67.5% | 65.4% | 66.1% | 28.0% | 27.3% | 25.1% | 58.7% | 55.0% | 52.3% |
Trimmed mean47 | 57.8% | 49.0% | 47.5% | 25.8% | 24.2% | 23.3% | 47.2% | 36.8% | 27.5% |
Median47 | 62.9% | 64.0% | 61.6% | 26.8% | 26.4% | 25.1% | 56.3% | 54.2% | 47.3% |
HSFL | 68.3% | 65.3% | 60.3% | 29.8% | 28.9% | 26.6% | 54.3% | 50.6% | 48.2% |
DPRLDS48 | 60.3% | 61.3% | 60.5% | 27.6% | 26.3% | 27.1% | 56.3% | 54.6% | 54.6% |
RMCS49 | 60.2% | 60.5% | 59.2% | 27.4% | 26.3% | 24.3% | 56.8% | 56.2% | 54.3% |
PPHSFL | 61.5% | 60.1% | 60.2% | 27.3% | 26.3% | 25.4% | 57.5% | 56.1% | 55.3% |
