Research map / Federated learning and privacy / Privacy attacks and secure learning
HealSplit: Towards Self-Healing Through Adversarial Distillation in Split Federated Learning
Yuhan Xie, Chen Lyu
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 |
|---|---|---|---|
| add HealSplit defense under DP+SP attack | SFL with FedAvg · MNIST, DP+SP | accuracy (%): 9.19 -> 93.88 | Improved |
| HealSplit versus FL-derived defenses under WP+SP | SFL · MNIST, WP+SP | accuracy (%): FLTrust 11.33 -> HealSplit 92.44 | Improved |
| add HealSplit when no attack is present | SFL · MNIST, no attack | accuracy (%): 96.90 (FedAvg) -> 97.17 | No clear effect |
| add Vanilla Teacher to adversarial multi-teacher distillation (ablation) | HealSplit · MNIST, DP+SP | accuracy (%): 90.99 -> 93.88 | Improved |
| add AD Teacher (ablation) | HealSplit · MNIST, DP+SP | accuracy (%): 87.34 -> 93.88 | Improved |
| add AD Teacher (ablation) | HealSplit · HAM10000, DP+SP | accuracy (%): 63.20 -> 72.27 | Improved |
| add distillation mechanism (ablation) | HealSplit · MNIST, DP+SP | accuracy (%): 74.38 -> 93.88 | Improved |
| add distillation mechanism (ablation) | HealSplit · CIFAR-10, DP+SP | accuracy (%): 42.75 -> 53.87 | Improved |
| add adversarial mechanism to distillation (ablation) | HealSplit · MNIST, DP+SP | accuracy (%): 92.74 -> 93.88 | Improved |
| add adversarial mechanism to distillation (ablation) | HealSplit · F-MNIST, DP+SP | accuracy (%): 82.59 -> 84.11 | Improved |
| apply adaptive attack that minimizes TAS divergence | HealSplit · MNIST, F-MNIST, CIFAR-10, HAM10K | accuracy | Hurt |
| increase malicious client proportion from 10% to 50% (ablation) | HealSplit · MNIST, DP+SP | accuracy | Hurt |
Closest papers by meaning
- Dual Defense: Enhancing Privacy and Mitigating Poisoning Attacks in Federated Learning (NeurIPS 2024 · similarity 0.86)
- Leak and Learn: An Attacker's Cookbook to Train Using Leaked Data from Federated Learning (CVPR 2024 · similarity 0.85)
- FedREDefense: Defending against Model Poisoning Attacks for Federated Learning using Model Update Reconstruction Error (ICML 2024 · similarity 0.85)
- SPMC: Self-Purifying Federated Backdoor Defense via Margin Contribution (ICML 2025 · similarity 0.85)
- IBAS:Imperceptible Backdoor Attacks in Split Learning with Limited Information (AAAI 2025 · similarity 0.84)
- Chronic Poisoning: Backdoor Attack against Split Learning (AAAI 2024 · similarity 0.84)
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.
