Research map / Federated learning and privacy / Privacy attacks and secure learning
Retaliatory Attacks Against Federated Unlearning via Data Leakage
Xinyi Sheng, Wei Bao, Hequn Wang, Yuqin Liu, Sen Fu
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 |
|---|---|---|---|
| launch AUA (relearn reconstructed unlearned data) | Retrain/FedEraser/KD/SGA/RobustFU · Location, Purchase100 | MIA accuracy and unlearned accuracy: MIA 0.502 -> 0.629; UA 0.606 -> 0.950 (Location, Retrain) | Improved |
| launch DUA (negative gradient on reconstructed data) | all FU methods · Location, Purchase100 | unlearned accuracy (UA): 0.606 -> 0.544 (Location, Retrain); 0.578 -> 0.497 (Purchase100) | Improved |
| DUA effect on normal test accuracy | all FU methods · Location, Purchase100 | test accuracy (TA): 0.611 -> 0.598; 0.637 -> 0.619 (Retrain) | Hurt |
| unlearning-induced MIA across image datasets | FU methods · CIFAR-10, SVHN | U-MIA accuracy: 0.654 (CIFAR-10, Retrain); 0.693 (SVHN, Retrain) | Improved |
| full coarse-to-fine pipeline vs single attack-model prediction, coarse-only or no filtering variants (S1-S3) (ablation) | AUA/DUA reconstruction · Location, Purchase100 (Retrain) | MIA increase / UA decrease | Improved |
| unlearning-induced MIA data reconstruction vs baselines B1-B3 | AUA/DUA · Location, Purchase100 (Retrain) | MIA increase / UA decrease | Improved |
Closest papers by meaning
- Federated Unlearning with Gradient Descent and Conflict Mitigation (AAAI 2025 · similarity 0.90)
- NoT: Federated Unlearning via Weight Negation (CVPR 2025 · similarity 0.90)
- Adversarial Machine Unlearning (ICLR 2025 · similarity 0.89)
- A Reliable Cryptographic Framework for Empirical Machine Unlearning Evaluation (NeurIPS 2025 · similarity 0.87)
- FedMIA: An Effective Membership Inference Attack Exploiting "All for One" Principle in Federated Learning (CVPR 2025 · similarity 0.87)
- Leak and Learn: An Attacker's Cookbook to Train Using Leaked Data from Federated Learning (CVPR 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.
