Research map / Federated learning and privacy / Personalized and heterogeneous FL
Federated Robustness Propagation: Sharing Adversarial Robustness in Heterogeneous Federated Learning
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 pseudo-noise calibration loss (λ=0.5) to transferred adversarial BN (ablation) | FedRBN · Digits, 20% AT users | robust accuracy: 51.0 -> 55.8 | Helped |
| Add pseudo-noise calibration loss (λ=0.5) to transferred adversarial BN (ablation) | FedRBN · Digits, 20% AT users | standard accuracy: 83.5 -> 87.3 | Helped |
| Add pseudo-noise calibration loss (λ=0.5) to transferred adversarial BN (ablation) | FedRBN · DomainNet, 20% AT users | robust accuracy: 21.4 -> 28.1 | Helped |
| Use cosine-similarity-based source weights instead of uniform weights (ablation) | FedRBN with λ=0.5 · Digits, 20% AT users | robust accuracy: 55.4 -> 55.8 | Helped |
| Use cosine-similarity-based source weights instead of uniform weights (ablation) | FedRBN with λ=0.5 · DomainNet, 20% AT users | robust accuracy: 27.5 -> 28.1 | Helped |
| Use transferred adversarial BN rather than local clean BN at test time (ablation) | FedRBN with λ=0.5 · Digits, 20% AT users | robust accuracy: 50.0 -> 55.8 | Helped |
| Use clean BN rather than transferred adversarial BN at test time (ablation) | FedRBN without pseudo-noise calibration · DomainNet, 20% AT users | robust accuracy: 19.8 -> 22.1 | Helped |
| Increase the pseudo-noise calibration coefficient (ablation) | FedRBN · Digits, partial AT settings | robust accuracy | Helped |
| Increase the pseudo-noise calibration coefficient (ablation) | FedRBN · Digits, partial AT settings | standard accuracy | Hurt |
| Use FedRBN instead of FATBN | Federated robustness propagation · Digits, 20% AT users | robust accuracy: 41.2 -> 55.8 | Helped |
| Use FedRBN instead of FATBN | Federated robustness propagation · DomainNet, 20% AT users | robust accuracy: 20.3 -> 28.1 | Helped |
| Use FedRBN instead of FATBN | Federated robustness propagation · Digits, 20% AT users | standard accuracy: 86.4 -> 87.3 | Mixed |
Closest papers by meaning
- Delving into the Adversarial Robustness of Federated Learning (AAAI 2023 · similarity 0.93)
- Byzantine-Robust Federated Learning: Impact of Client Subsampling and Local Updates (ICML 2024 · similarity 0.89)
- Fine-Tuning Personalization in Federated Learning to Mitigate Adversarial Clients (NeurIPS 2024 · similarity 0.88)
- Efficient Federated Learning against Byzantine Attacks and Data Heterogeneity via Aggregating Normalized Gradients (NeurIPS 2025 · similarity 0.88)
- FedAA: A Reinforcement Learning Perspective on Adaptive Aggregation for Fair and Robust Federated Learning (AAAI 2025 · similarity 0.87)
- Low Precision Local Training is Enough for Federated Learning (NeurIPS 2024 · similarity 0.87)
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.