Image Classification on MNIST (test) (Accuracy and Attack Success Rate)
98.39AccuracyGradMA
Evaluation results
| Method | Links | ||
|---|---|---|---|
| GradMABackbone=NN, omega=1, S=102023.02 | 98.39 | — | |
| GradMA-SBackbone=NN, omega=1, S=102023.02 | 98.38 | — | |
| FedAvgBackbone=NN, omega=1, S=102023.02 | 98.22 | — | |
| FedProxBackbone=NN, omega=1, S=102023.02 | 98.16 | — | |
| GradMA-WBackbone=NN, omega=1, S=102023.02 | 98.15 | — | |
| ACILMethod Category=Proposed2026.02 | 98 | — | |
| ScaffoldBackbone=NN, omega=1, S=102023.02 | 97.63 | — | |
| GradMA-SBackbone=NN, omega=0.1, S=102023.02 | 97.35 | — | |
| GradMABackbone=NN, omega=0.1, S=102023.02 | 97.34 | — | |
| FedProxBackbone=NN, omega=0.1, S=102023.02 | 97.19 | — | |
| FedAvgBackbone=NN, omega=0.1, S=102023.02 | 97.11 | — | |
| GradMA-WBackbone=NN, omega=0.1, S=102023.02 | 97.01 | — | |
| Traditional Poisoning AttackNetwork=FC, Trigger Pattern=Square, Pre-trained Model Accuracy=97%2021.06 | 97 | 100 | |
| Traditional Poisoning AttackNetwork=FC, Trigger Pattern=Checkerboard, Pre-trained Model Accuracy=97%2021.06 | 97 | 100 | |
| RainbowMethod Category=CIL baseline2026.02 | 97 | — | |
| iCaRLMethod Category=CIL baseline2026.02 | 96 | — | |
| GDumbMethod Category=CIL baseline2026.02 | 96 | — | |
| Handcrafted Backdoor AttackNetwork=FC, Trigger Pattern=Square, Pre-trained Model Accuracy=97%2021.06 | 95 | 100 | |
| Handcrafted Backdoor AttackNetwork=FC, Trigger Pattern=Checkerboard, Pre-trained Model Accuracy=97%2021.06 | 94 | 100 | |
| ScaffoldBackbone=NN, omega=0.1, S=102023.02 | 93.94 | — | |
| Traditional Poisoning AttackNetwork=FC, Trigger Pattern=Random, Pre-trained Model Accuracy=97%2021.06 | 89 | 96 | |
| Handcrafted Backdoor AttackNetwork=FC, Trigger Pattern=Random, Pre-trained Model Accuracy=97%2021.06 | 88 | 100 | |
| GradMABackbone=NN, omega=0.01, S=52023.02 | 77.97 | — | |
| GradMA-SBackbone=NN, omega=0.01, S=502023.02 | 75.93 | — | |
| GradMABackbone=NN, omega=0.01, S=502023.02 | 75.51 | — | |
| GradMA-SBackbone=NN, omega=0.01, S=52023.02 | 74.52 | — | |
| GradMA-WBackbone=NN, omega=0.01, S=502023.02 | 65.39 | — | |
| GradMA-WBackbone=NN, omega=0.01, S=52023.02 | 63.34 | — | |
| ScaffoldBackbone=NN, omega=0.01, S=52023.02 | 50.86 | — | |
| FedProxBackbone=NN, omega=0.01, S=502023.02 | 49.89 | — | |
| FedAvgBackbone=NN, omega=0.01, S=502023.02 | 49.65 | — | |
| FedProxBackbone=NN, omega=0.01, S=52023.02 | 46.82 | — | |
| FedAvgBackbone=NN, omega=0.01, S=52023.02 | 46.19 | — | |
| RandomMethod Category=AL baseline2026.02 | 41 | — | |
| BADGEMethod Category=AL baseline2026.02 | 41 | — | |
| ScaffoldBackbone=NN, omega=0.01, S=502023.02 | 39.97 | — | |
| CoresetMethod Category=AL baseline2026.02 | 39 | — | |
| FinetuningMethod Category=CIL baseline2026.02 | 20 | — |