Research map / Federated learning and privacy / Privacy attacks and secure learning
CRFL: Certifiably Robust Federated Learning against Backdoor Attacks
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 |
|---|---|---|---|
| Increase training-time noise (ablation) | CRFL · MNIST, LOAN, EMNIST | certified rate | Helped |
| Increase training-time noise (ablation) | CRFL · MNIST, LOAN, EMNIST | certified accuracy | Mixed |
| Apply parameter smoothing at test time (ablation) | CRFL · MNIST, LOAN, EMNIST | base classifier accuracy | No clear effect |
| Increase attacker count, poison ratio, or attack scale (ablation) | CRFL · MNIST, LOAN, EMNIST | certified radius | Hurt |
| Use RFA instead of FedAvg for aggregation (ablation) | CRFL · MNIST, EMNIST | certified radius | Helped |
| Increase the total number of clients (ablation) | CRFL · MNIST, LOAN | certified radius | Helped |
| Increase the number of training rounds (ablation) | CRFL · LOAN, EMNIST | certified accuracy | Helped |
| Increase the number of training rounds (ablation) | CRFL · LOAN, EMNIST | maximum certified radius | No clear effect |
| Increase the number of noisy models used for certification (ablation) | CRFL · MNIST | certified radius | Helped |
| Decrease the error tolerance α (ablation) | CRFL · MNIST | certified radius | Hurt |
| Change the error tolerance α (ablation) | CRFL · MNIST | certified accuracy | No clear effect |
| Train benign rounds after the attack (ablation) | CRFL · MNIST | distance between backdoored and benign global models | Helped |
Closest papers by meaning
- Defending against Backdoors in Federated Learning with Robust Learning Rate (AAAI 2021 · similarity 0.91)
- Bad-PFL: Exploiting Backdoor Attacks against Personalized Federated Learning (ICLR 2025 · similarity 0.88)
- Backdoor Federated Learning by Poisoning Backdoor-Critical Layers (ICLR 2024 · similarity 0.88)
- Towards the Robustness of Differentially Private Federated Learning (AAAI 2024 · similarity 0.87)
- Resisting Backdoor Attacks in Federated Learning via Bidirectional Elections and Individual Perspective (AAAI 2024 · similarity 0.87)
- Certifiably Robust Model Evaluation in Federated Learning under Meta-Distributional Shifts (ICML 2025 · 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.