Image Classification on FMNIST (Accuracy and Runtime)
90.99AccuracyDPSGD-HF
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DPSGD-HFprivacy=non-private2026.06 | 90.99 | — | |
| DPSR-CGepsilon=32026.06 | 90.98 | — | |
| DPSR-CGepsilon=22026.06 | 90.87 | — | |
| DPSR-CGepsilon=12026.06 | 90.31 | — | |
| MeanAttack=No Attack2025.08 | 89.23 | — | |
| Trimmed MeanAttack=No Attack2025.08 | 89.23 | — | |
| Multi KrumAttack=No Attack2025.08 | 89.14 | — | |
| MedianAttack=No Attack2025.08 | 89.05 | — | |
| DPSGD-HFepsilon=32026.06 | 89.01 | — | |
| Centralized LearningLearning Setting=Centralized2026.06 | 88.91 | — | |
| Loss-Based ClusteringAttack=No Attack2025.08 | 88.9 | — | |
| Loss-Based ClusteringAttack=Sign Flip2025.08 | 88.9 | — | |
| Loss-Based ClusteringAttack=Gaussian Noise2025.08 | 88.86 | — | |
| DPISepsilon=32026.06 | 88.82 | — | |
| Multi KrumAttack=Sign Flip2025.08 | 88.75 | — | |
| Loss-Based ClusteringAttack=Label Flip2025.08 | 88.67 | — | |
| Multi KrumAttack=Gaussian Noise2025.08 | 88.66 | — | |
| DPSGD Matrix Mechanismepsilon=32026.06 | 88.32 | — | |
| DPISepsilon=22026.06 | 88.24 | — | |
| DPSGD-HFepsilon=22026.06 | 87.96 | — | |
| DPSGD Matrix Mechanismepsilon=22026.06 | 87.84 | — | |
| FedMTFILearning Setting=Federated2026.06 | 87.28 | — | |
| DPSGD Matrix Mechanismepsilon=12026.06 | 87.16 | — | |
| ENSEMBLES2025.12 | 87 | 4.03 | |
| DPSURepsilon=32026.06 | 86.6 | — | |
| DPSURepsilon=22026.06 | 86.46 | — | |
| KrumAttack=Gaussian Noise2025.08 | 86.44 | — | |
| DPSURepsilon=12026.06 | 86.4 | — | |
| KrumAttack=Sign Flip2025.08 | 86.29 | — | |
| DPSGD-TSepsilon=32026.06 | 86.29 | — | |
| DPISepsilon=12026.06 | 86.25 | — | |
| KrumAttack=No Attack2025.08 | 86.2 | — | |
| DPAGDepsilon=32026.06 | 86.04 | — | |
| DPSGD-HFepsilon=12026.06 | 85.54 | — | |
| DPSGD-TSepsilon=22026.06 | 85.33 | — | |
| DPSGDepsilon=32026.06 | 84.72 | — | |
| DPAGDepsilon=22026.06 | 84.5 | — | |
| FedKDShapLearning Setting=Federated2026.06 | 84.43 | — | |
| DPSGD-TSepsilon=12026.06 | 83.63 | — | |
| ESPRO2025.12 | 83 | 2.1 | |
| DPSGDepsilon=22026.06 | 82.63 | — | |
| FedProxLearning Setting=Federated2026.06 | 82.31 | — | |
| MetricBNN2025.12 | 82 | 3.67 | |
| DPAGDepsilon=12026.06 | 81.26 | — | |
| FedAvgLearning Setting=Federated2026.06 | 80.65 | — | |
| DPSGDepsilon=12026.06 | 80.25 | — | |
| SWAG2025.12 | 77 | 2.51 | |
| Riem Lin LA2025.12 | 75 | 3.91 | |
| Lin LA2025.12 | 67 | 3.67 | |
| LA LowRank2025.12 | 45 | 1.67 | |
| KrumAttack=Label Flip2025.08 | 43.15 | — | |
| Multi KrumAttack=Label Flip2025.08 | 43.11 | — | |
| MeanAttack=Label Flip2025.08 | 38.16 | — | |
| Riem LA2025.12 | 38 | 5.01 | |
| Trimmed MeanAttack=Label Flip2025.08 | 37.16 | — | |
| MedianAttack=Label Flip2025.08 | 35.77 | — | |
| Trimmed MeanAttack=Gaussian Noise2025.08 | 34.56 | — | |
| MedianAttack=Gaussian Noise2025.08 | 34.51 | — | |
| MeanAttack=Gaussian Noise2025.08 | 32.76 | — | |
| SGLD2025.12 | 25 | 2.97 | |
| LA GN2025.12 | 19 | 2.24 | |
| MeanAttack=Sign Flip2025.08 | 10 | — | |
| Trimmed MeanAttack=Sign Flip2025.08 | 10 | — | |
| MedianAttack=Sign Flip2025.08 | 10 | — |