Image Classification on MNIST (Accuracy, ∆CA, ASR)
86AccuracyFLANDERS + FedAvg
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| FLANDERS + FedAvgAttack Strategy=GAUSS, Malicious Client Ratio (r)=0.22023.03 | 86 | — | — | |
| FLANDERS + FedAvgAttack Strategy=AGR-MM, Malicious Client Ratio (r)=0.22023.03 | 85 | — | — | |
| FLANDERS + FedAvgAttack Strategy=LIE, Malicious Client Ratio (r)=0.22023.03 | 83 | — | — | |
| RN50 BaselineIR=1, Backbone=ResNet-502026.01 | 69.2 | — | — | |
| FLDetector + FedAvgAttack Strategy=OPT, Malicious Client Ratio (r)=0.22023.03 | 68 | — | — | |
| FedAvgAttack Strategy=OPT, Malicious Client Ratio (r)=0.22023.03 | 63 | — | — | |
| VLM BaselineIR=1, Backbone=VLM2026.01 | 62.3 | — | — | |
| FLANDERS + FedAvgAttack Strategy=OPT, Malicious Client Ratio (r)=0.22023.03 | 62 | — | — | |
| FLDetector + FedAvgAttack Strategy=AGR-MM, Malicious Client Ratio (r)=0.22023.03 | 43 | — | — | |
| FedAvgAttack Strategy=AGR-MM, Malicious Client Ratio (r)=0.22023.03 | 34 | — | — | |
| FLDetector + FedAvgAttack Strategy=GAUSS, Malicious Client Ratio (r)=0.22023.03 | 20 | — | — | |
| FedAvgAttack Strategy=GAUSS, Malicious Client Ratio (r)=0.22023.03 | 18 | — | — | |
| FedAvgAttack Strategy=LIE, Malicious Client Ratio (r)=0.22023.03 | 12 | — | — | |
| FLDetector + FedAvgAttack Strategy=LIE, Malicious Client Ratio (r)=0.22023.03 | 11 | — | — | |
| PRISMIR=1, Attack Type=Blend, Backbone=ResNet-502026.01 | — | 2.2 | 0 | |
| PRISMIR=1, Attack Type=BPP, Backbone=ResNet-502026.01 | — | 3.4 | 0 | |
| PRISMIR=1, Attack Type=CTRL, Backbone=ResNet-502026.01 | — | 1 | 0 |