Research map / Federated learning and privacy / Privacy attacks and secure learning
Protection against Source Inference Attacks in Federated Learning
Andreas Athanasiou, Kangsoo Jung, Catuscia Palamidessi
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 |
|---|---|---|---|
| shuffle at parameter level instead of model level | shuffling defenses · SIA on CIFAR-10 | SIA accuracy: 50% to 28% (alpha=0.1) | Improved |
| shuffle at layer level instead of parameter level | reconstruction attack · SIA on CIFAR-100 ResNet-18 | SIA accuracy: 78% vs 60% at alpha=0.1 | Hurt |
| add encoding on top of parameter-level shuffling | parameter-level shuffling · SIA accuracy at all alpha | SIA accuracy: reduces to random guess level | Improved |
| apply shuffling with encoding against data reconstruction attack | CIFAR-10 DRA, 10 clients · CIFAR-10 | reconstruction loss: 3e-4 -> 0.98 | Improved |
| increase digit precision r for encoded fractional part (ablation) | encoding algorithm · CIFAR-10 and CIFAR-100 | joint model accuracy: r=3 approaches vanilla FL; r=8 matches it | Improved |
| increase digit precision r from 2 to 3 on MNIST (ablation) | encoding algorithm · MNIST | joint model accuracy | No clear effect |
| use r=5 instead of vanilla 32-bit full values (ablation) | encoding algorithm · CIFAR-100 ResNet (10 clients) | top-1 accuracy: 55 -> 53 | No clear effect |
| add RLE compression under fully trusted shuffler (ablation) | encoding algorithm · CIFAR-100 ResNet (10 clients) | communication expansion: 1.81x -> 1.03x | Improved |
Closest papers by meaning
- Concealing Sensitive Samples against Gradient Leakage in Federated Learning (AAAI 2024 · similarity 0.87)
- Attribute Inference Attacks for Federated Regression Tasks (AAAI 2025 · similarity 0.87)
- Towards the Robustness of Differentially Private Federated Learning (AAAI 2024 · similarity 0.87)
- Leak and Learn: An Attacker's Cookbook to Train Using Leaked Data from Federated Learning (CVPR 2024 · similarity 0.87)
- Towards Trustworthy Federated Learning with Untrusted Participants (ICML 2025 · similarity 0.87)
- A New Federated Learning Framework Against Gradient Inversion Attacks (AAAI 2025 · similarity 0.86)
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.
