Research map / Federated learning and privacy / Differential privacy
Unified Privacy Guarantees for Decentralized Learning via Matrix Factorization
Aurélien Bellet, Edwige Cyffers, Davide Frey, Romaric Gaudel, Dimitri Lerévérend, Francois Taiani
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 optimized local noise correlation (MAFALDA-SGD) instead of uncorrelated noise (DP-D-SGD) | decentralized SGD under LDP · Housing, Facebook Ego graph | test loss: 31% average improvement in test loss; 2-fold reduction in epsilon at test loss 0.75 | Improved |
| use optimized correlation instead of AntiPGD correlation in decentralized setting | decentralized SGD under LDP · Housing, Facebook Ego graph | test loss: AntiPGD test loss always greater than 6 | Improved |
| use matrix factorization PNDP accounting instead of prior accounting | DP-D-SGD · Erdos-Renyi, Facebook Ego, PeerTube, Florentine graphs | privacy loss (Renyi divergence): up to an order of magnitude at distance <=2; at least two orders at distance >=3 | Improved |
| use MAFALDA-SGD correlated noise instead of private baselines | decentralized SGD under LDP · FEMNIST, Facebook Ego graph | test accuracy | Improved |
Closest papers by meaning
- The Privacy Power of Correlated Noise in Decentralized Learning (ICML 2024 · similarity 0.92)
- Mitigating the Privacy–Utility Trade-off in Decentralized Federated Learning via f-Differential Privacy (NeurIPS 2025 · similarity 0.91)
- Revisiting Differentially Private Algorithms for Decentralized Online Learning (ICML 2025 · similarity 0.91)
- Differentially Private Decentralized Learning with Random Walks (ICML 2024 · similarity 0.90)
- Privacy Attacks in Decentralized Learning (ICML 2024 · similarity 0.89)
- DMM: Distributed Matrix Mechanism for Differentially-Private Federated Learning Based on Constant-Overhead Linear Secret Resharing (ICML 2025 · similarity 0.89)
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.
