Research map / Adversarial robustness and security / Backdoor and poisoning
Defending against Backdoors in Federated Learning with Robust Learning Rate
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add per-dimension robust learning rates to FedAvg | FedAvg · Fashion-MNIST IID | backdoor accuracy: 100 -> 0 | Helped |
| Add per-dimension robust learning rates to FedAvg | FedAvg · Federated EMNIST non-IID | backdoor accuracy: 99.3 -> 3.4 | Helped |
| Add per-dimension robust learning rates to FedAvg | FedAvg · CIFAR10 IID, regular and distributed backdoors | backdoor accuracy: 88.6 -> 9.0; 61.7 -> 8.5 | Helped |
| Combine clipping and noise with FedAvg without RLR | FedAvg · Fashion-MNIST IID | backdoor accuracy: 100 -> 100 | No clear effect |
| Activate RLR only after the model approaches convergence (ablation) | FedAvg · Fashion-MNIST IID | time to convergence | Helped |
| Use threshold 7 instead of threshold 8 for late RLR (ablation) | FedAvg · Federated EMNIST non-IID, late activation | backdoor accuracy: 2% -> 45% | Hurt |
| Increase the late-activation RLR threshold from 7 to 8 (ablation) | FedAvg · Federated EMNIST non-IID, late activation | backdoor accuracy: 45% -> 1.4% | Helped |
| Increase corruption rate while retaining RLR (ablation) | FedAvg with RLR · Fashion-MNIST IID | backdoor accuracy: 0; 0; 0 | No clear effect |
| Increase corruption rate while retaining RLR (ablation) | FedAvg with RLR · Federated EMNIST non-IID | backdoor accuracy: 4.3; 0.8; 0.1 | Helped |
| Have the adversary negate its loss function (ablation) | FedAvg with RLR · IID and non-IID | backdoor accuracy: 0.7; 3 | No clear effect |
| Vary clipping threshold and noise with RLR enabled (ablation) | FedAvg with RLR · Fashion-MNIST IID | backdoor accuracy: 0 to 6.1 | Mixed |
Closest papers by meaning
- Resisting Backdoor Attacks in Federated Learning via Bidirectional Elections and Individual Perspective (AAAI 2024 · similarity 0.92)
- Beyond Traditional Threats: A Persistent Backdoor Attack on Federated Learning (AAAI 2024 · similarity 0.91)
- MARS: A Malignity-Aware Backdoor Defense in Federated Learning (NeurIPS 2025 · similarity 0.90)
- Backdoor Federated Learning by Poisoning Backdoor-Critical Layers (ICLR 2024 · similarity 0.90)
- Infighting in the Dark: Multi-Label Backdoor Attack in Federated Learning (CVPR 2025 · similarity 0.90)
- SPMC: Self-Purifying Federated Backdoor Defense via Margin Contribution (ICML 2025 · similarity 0.90)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.