Research map / Federated learning and privacy / Privacy attacks and secure learning
RECESS Vaccine for Federated Learning: Proactive Defense Against Model Poisoning Attacks
Haonan Yan, Wenjing Zhang, Qian Chen, Xiaoguang Li, Wenhai Sun, HUI LI, Xiaodong Lin
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| Use proactive client testing and multi-round trust scoring instead of passive defenses | Federated learning · MNIST, CIFAR-10, Purchase, and FEMNIST under model-poisoning attacks | accuracy | Mixed |
| Use RECESS rather than FLDetector | CIFAR-10 IID · AGR-agnostic Min-Max attack | accuracy: 0.5178 -> 0.6286 | Improved |
| Use RECESS rather than FLDetector | CIFAR-10 Non-IID 0.8 · Optimization attack | accuracy: 0.3544 -> 0.6018 | Improved |
| Use RECESS rather than FLDetector | FEMNIST · AGR-agnostic Min-Max attack | accuracy: 0.5616 -> 0.7850 | Improved |
| Use RECESS rather than FedAvg | CIFAR-10, ResNet20 · Optimization attack | accuracy: 0.1718 -> 0.8173 | Improved |
| Use RECESS with Median aggregation instead of weighted averaging | RECESS · CIFAR-10 label-flipping attack | Model Accuracy: 0.4415 -> 0.6148 | Improved |
| Use RECESS with Median aggregation instead of weighted averaging | RECESS · CIFAR-10 scaling attack | Attack Success Rate: 1.0 -> 0.0 | Improved |
| Increase detection frequency from 0 to 10 (ablation) | RECESS · FEMNIST adaptive optimization attack, white-box | accuracy: 0.2482 -> 0.3456 | Improved |
| Increase detection frequency from 10 to 50 (ablation) | RECESS · FEMNIST adaptive optimization attack, white-box | accuracy: 0.3456 -> 0.6847 | Improved |
| Increase detection frequency from 50 to 100 (ablation) | RECESS · FEMNIST adaptive optimization attack, white-box | accuracy: 0.6847 -> 0.7815 | Improved |
| Increase detection frequency from 100 to 200 (ablation) | RECESS · FEMNIST adaptive optimization attack, white-box | accuracy: 0.7815 -> 0.8011 | Improved |
| Increase the trust-score deduction baseline (ablation) | RECESS · Model-poisoning detection | detection speed | Mixed |
Closest papers by meaning
- Dual Defense: Enhancing Privacy and Mitigating Poisoning Attacks in Federated Learning (NeurIPS 2024 · similarity 0.86)
- Defending Against Sophisticated Poisoning Attacks with RL-based Aggregation in Federated Learning (AAAI 2025 · similarity 0.86)
- Model Poisoning Attacks to Federated Learning via Multi-Round Consistency (CVPR 2025 · similarity 0.85)
- FedREDefense: Defending against Model Poisoning Attacks for Federated Learning using Model Update Reconstruction Error (ICML 2024 · similarity 0.85)
- Tracing Back the Malicious Clients in Poisoning Attacks to Federated Learning (NeurIPS 2025 · similarity 0.85)
- RFLPA: A Robust Federated Learning Framework against Poisoning Attacks with Secure Aggregation (NeurIPS 2024 · similarity 0.85)
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.
