Byzantine-robust Federated Learning on CIFAR-10 (val)
25.7Error RateAdaBFL
Evaluation results
| Method | Links | |
|---|---|---|
| AdaBFLAttack Scenario=LF, Malicious Client Ratio=0.32026.04 | 25.7 | |
| FedAvgAttack Scenario=No, Malicious Client Ratio=0.32026.04 | 26.8 | |
| AdaBFLAttack Scenario=No, Malicious Client Ratio=0.32026.04 | 26.8 | |
| Trim-meanAttack Scenario=LF, Malicious Client Ratio=0.32026.04 | 26.9 | |
| Gau+Trim-meanAttack Scenario=No, Malicious Client Ratio=0.32026.04 | 27.1 | |
| FedAvgAttack Scenario=LF, Malicious Client Ratio=0.32026.04 | 27.2 | |
| Trim-meanAttack Scenario=No, Malicious Client Ratio=0.32026.04 | 27.2 | |
| Gau+MedianAttack Scenario=No, Malicious Client Ratio=0.32026.04 | 27.2 | |
| FoundationFL+MeanAttack Scenario=No, Malicious Client Ratio=0.32026.04 | 27.2 | |
| FoundationFL+MeanAttack Scenario=LF, Malicious Client Ratio=0.32026.04 | 27.2 | |
| FoundationFL+MedianAttack Scenario=Krum, Malicious Client Ratio=0.32026.04 | 27.3 | |
| Gau+Trim-meanAttack Scenario=LF, Malicious Client Ratio=0.32026.04 | 27.3 | |
| Gau+MedianAttack Scenario=LF, Malicious Client Ratio=0.32026.04 | 27.3 | |
| AdaBFLAttack Scenario=Krum, Malicious Client Ratio=0.32026.04 | 27.4 | |
| FoundationFL+MedianAttack Scenario=No, Malicious Client Ratio=0.32026.04 | 27.5 | |
| FoundationFL+MedianAttack Scenario=LF, Malicious Client Ratio=0.32026.04 | 27.5 | |
| AdaBFLAttack Scenario=Trim, Malicious Client Ratio=0.32026.04 | 27.6 | |
| MedianAttack Scenario=LF, Malicious Client Ratio=0.32026.04 | 27.8 | |
| AdaBFLAttack Scenario=Scaling, Malicious Client Ratio=0.32026.04 | 27.8 | |
| AdaBFLAttack Scenario=Sybil, Malicious Client Ratio=0.32026.04 | 27.8 | |
| MedianAttack Scenario=No, Malicious Client Ratio=0.32026.04 | 28 | |
| AdaBFLAttack Scenario=Gaussian, Malicious Client Ratio=0.32026.04 | 28.6 | |
| AdaBFLAttack Scenario=Min-Max, Malicious Client Ratio=0.32026.04 | 29 | |
| MedianAttack Scenario=Gaussian, Malicious Client Ratio=0.32026.04 | 29.3 | |
| FoundationFL+MedianAttack Scenario=Gaussian, Malicious Client Ratio=0.32026.04 | 31.4 | |
| Gau+MedianAttack Scenario=Gaussian, Malicious Client Ratio=0.32026.04 | 31.4 | |
| MedianAttack Scenario=Krum, Malicious Client Ratio=0.32026.04 | 32.2 | |
| MedianAttack Scenario=Min-Max, Malicious Client Ratio=0.32026.04 | 33.6 | |
| FoundationFL+MedianAttack Scenario=Trim, Malicious Client Ratio=0.32026.04 | 34 | |
| FoundationFL+MedianAttack Scenario=Scaling, Malicious Client Ratio=0.32026.04 | 35.1 | |
| FoundationFL+MedianAttack Scenario=Min-Max, Malicious Client Ratio=0.32026.04 | 36.3 | |
| FoundationFL+MedianAttack Scenario=Sybil, Malicious Client Ratio=0.32026.04 | 38.8 | |
| Gau+MedianAttack Scenario=Min-Max, Malicious Client Ratio=0.32026.04 | 38.9 | |
| FoundationFL+MeanAttack Scenario=Trim, Malicious Client Ratio=0.32026.04 | 41.2 | |
| MedianAttack Scenario=Trim, Malicious Client Ratio=0.32026.04 | 42.7 | |
| MedianAttack Scenario=Sybil, Malicious Client Ratio=0.32026.04 | 43.2 | |
| MedianAttack Scenario=Scaling, Malicious Client Ratio=0.32026.04 | 44.9 | |
| GAS+MedianAttack Scenario=Sybil, Malicious Client Ratio=0.32026.04 | 48.5 | |
| GAS+Trim-meanAttack Scenario=Trim, Malicious Client Ratio=0.32026.04 | 50 | |
| Gau+MedianAttack Scenario=Trim, Malicious Client Ratio=0.32026.04 | 51.8 | |
| GAS+MedianAttack Scenario=Scaling, Malicious Client Ratio=0.32026.04 | 55 | |
| GAS+MedianAttack Scenario=No, Malicious Client Ratio=0.32026.04 | 55.5 | |
| Gau+Trim-meanAttack Scenario=Gaussian, Malicious Client Ratio=0.32026.04 | 55.7 | |
| GAS+Trim-meanAttack Scenario=No, Malicious Client Ratio=0.32026.04 | 56.3 | |
| GAS+MedianAttack Scenario=Gaussian, Malicious Client Ratio=0.32026.04 | 57 | |
| GAS+MedianAttack Scenario=LF, Malicious Client Ratio=0.32026.04 | 57.2 | |
| GAS+Trim-meanAttack Scenario=LF, Malicious Client Ratio=0.32026.04 | 57.7 | |
| GAS+MedianAttack Scenario=Min-Max, Malicious Client Ratio=0.32026.04 | 57.9 | |
| GAS+MedianAttack Scenario=Krum, Malicious Client Ratio=0.32026.04 | 58.3 | |
| GAS+Trim-meanAttack Scenario=Min-Max, Malicious Client Ratio=0.32026.04 | 58.9 | |
| GAS+MedianAttack Scenario=Trim, Malicious Client Ratio=0.32026.04 | 59.4 | |
| Gau+Trim-meanAttack Scenario=Min-Max, Malicious Client Ratio=0.32026.04 | 67.1 | |
| Gau+Trim-meanAttack Scenario=Trim, Malicious Client Ratio=0.32026.04 | 67.6 | |
| FoundationFL+MeanAttack Scenario=Krum, Malicious Client Ratio=0.32026.04 | 73.7 | |
| FoundationFL+MeanAttack Scenario=Min-Max, Malicious Client Ratio=0.32026.04 | 79.5 | |
| Trim-meanAttack Scenario=Min-Max, Malicious Client Ratio=0.32026.04 | 85 | |
| GAS+Trim-meanAttack Scenario=Gaussian, Malicious Client Ratio=0.32026.04 | 86.4 | |
| GAS+Trim-meanAttack Scenario=Scaling, Malicious Client Ratio=0.32026.04 | 86.5 | |
| FedAvgAttack Scenario=Gaussian, Malicious Client Ratio=0.32026.04 | 90 | |
| FedAvgAttack Scenario=Trim, Malicious Client Ratio=0.32026.04 | 90 | |
| FedAvgAttack Scenario=Krum, Malicious Client Ratio=0.32026.04 | 90 | |
| FedAvgAttack Scenario=Min-Max, Malicious Client Ratio=0.32026.04 | 90 | |
| FedAvgAttack Scenario=Scaling, Malicious Client Ratio=0.32026.04 | 90 | |
| FedAvgAttack Scenario=Sybil, Malicious Client Ratio=0.32026.04 | 90 | |
| Trim-meanAttack Scenario=Gaussian, Malicious Client Ratio=0.32026.04 | 90 | |
| Trim-meanAttack Scenario=Trim, Malicious Client Ratio=0.32026.04 | 90 | |
| Trim-meanAttack Scenario=Krum, Malicious Client Ratio=0.32026.04 | 90 | |
| Trim-meanAttack Scenario=Scaling, Malicious Client Ratio=0.32026.04 | 90 | |
| Trim-meanAttack Scenario=Sybil, Malicious Client Ratio=0.32026.04 | 90 | |
| GAS+Trim-meanAttack Scenario=Krum, Malicious Client Ratio=0.32026.04 | 90 | |
| GAS+Trim-meanAttack Scenario=Sybil, Malicious Client Ratio=0.32026.04 | 90 | |
| Gau+Trim-meanAttack Scenario=Krum, Malicious Client Ratio=0.32026.04 | 90 | |
| Gau+Trim-meanAttack Scenario=Scaling, Malicious Client Ratio=0.32026.04 | 90 | |
| Gau+Trim-meanAttack Scenario=Sybil, Malicious Client Ratio=0.32026.04 | 90 | |
| Gau+MedianAttack Scenario=Krum, Malicious Client Ratio=0.32026.04 | 90 | |
| Gau+MedianAttack Scenario=Scaling, Malicious Client Ratio=0.32026.04 | 90 | |
| Gau+MedianAttack Scenario=Sybil, Malicious Client Ratio=0.32026.04 | 90 | |
| FoundationFL+MeanAttack Scenario=Gaussian, Malicious Client Ratio=0.32026.04 | 90 | |
| FoundationFL+MeanAttack Scenario=Scaling, Malicious Client Ratio=0.32026.04 | 90 | |
| FoundationFL+MeanAttack Scenario=Sybil, Malicious Client Ratio=0.32026.04 | 90 |