Research map / Federated learning and privacy / Differential privacy
FLAME: Differentially Private Federated Learning in the Shuffle Model
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 |
|---|---|---|---|
| Replace perturbing every dimension with subsampling and dummy padding (ablation) | SS-Simple, yielding SS-Double · MNIST | testing accuracy: +4.07% | Helped |
| Select Top-k dimensions by magnitude and pad the remaining reported indexes (ablation) | SS-Double, yielding SS-Topk · MNIST | testing accuracy: +55.5% | Helped |
| Use SS-Topk instead of local-model FL | Federated learning on MNIST · MNIST | testing accuracy: +60.7% | Helped |
| Use SS-Topk instead of curator-model FL | Federated learning on MNIST · MNIST | testing accuracy: +33.94% | Helped |
| Apply SS-Topk with beta=1/10 for 2 epochs | Federated learning on MNIST · MNIST, 2 epochs | testing accuracy loss relative to NP-FL: 1.48% | Hurt |
| Increase the local privacy budget per dimension (ablation) | SS-Topk · MNIST | testing accuracy | Helped |
| Increase the ratio n/n_p, reducing dummy padding (ablation) | SS-Double and SS-Topk · MNIST | testing accuracy | Helped |
| Increase beta, the fraction of sampled dimensions (ablation) | SS-Double and SS-Topk · MNIST | testing accuracy | Helped |
| Use Top-k selection and double amplification rather than shuffling amplification alone (ablation) | SS-Topk · MNIST | testing accuracy | Helped |
| Vary index privacy while keeping n_p, n, and beta fixed (ablation) | SS-Topk · MNIST | mean estimation accuracy | No clear effect |
Closest papers by meaning
- Leveraging Randomness in Model and Data Partitioning for Privacy Amplification (ICML 2025 · similarity 0.90)
- A Generalized Shuffle Framework for Privacy Amplification: Strengthening Privacy Guarantees and Enhancing Utility (AAAI 2024 · similarity 0.88)
- Learning from End User Data with Shuffled Differential Privacy over Kernel Densities (ICLR 2025 · similarity 0.88)
- Mitigating the Privacy–Utility Trade-off in Decentralized Federated Learning via f-Differential Privacy (NeurIPS 2025 · similarity 0.88)
- Noise-Aware Algorithm for Heterogeneous Differentially Private Federated Learning (ICML 2024 · similarity 0.88)
- Clients Collaborate: Flexible Differentially Private Federated Learning with Guaranteed Improvement of Utility-Privacy Trade-off (ICML 2025 · similarity 0.88)
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.