Fast and Differentially Private Algorithms for Decentralized Collaborative Machine Learning

2017/01/01 by Aurélien Bellet, Bellet, Aurélien, Rachid Guerraoui +5 · 5 citations
Computer Science · #Age of Information Optimization #Blockchain Technology Applications and Security #Cryptography and Data Security #Cryptography and Security (cs.CR) #Distributed #FOS: Computer and information sciences #FOS: Electrical engineering #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Mobile Crowdsensing and Crowdsourcing #Parallel #Privacy-Preserving Technologies in Data #Systems and Control (eess.SY) #and Cluster Computing (cs.DC) #electronic engineering #information engineering

paper · pdf · doi:10.48550/arxiv.1705.08435

openalex publication_date 2017/01/01 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

The rise of connected personal devices together with privacy concerns call\nfor machine learning algorithms capable of leveraging the data of a large\nnumber of agents to learn personalized models under strong privacy\nrequirements. In this paper, we introduce an efficient algorithm to address the\nabove problem in a fully decentralized (peer-to-peer) and asynchronous fashion,\nwith provable convergence rate. We show how to make the algorithm\ndifferentially private to protect against the disclosure of information about\nthe personal datasets, and formally analyze the trade-off between utility and\nprivacy. Our experiments show that our approach dramatically outperforms\nprevious work in the non-private case, and that under privacy constraints, we\ncan significantly improve over models learned in isolation.\n

Cited by

• MedFedPure: A Medical Federated Framework with MAE-based Detection and Diffusion Purification for Inference-Time Attacks
• DFCA: Decentralized Federated Clustering Algorithm
• Unified Privacy Guarantees for Decentralized Learning via Matrix Factorization
• Marketplace for AI Models
• FedCD: Improving Performance in non-IID Federated Learning

Related

• The Cost of Privacy in Asynchronous Differentially-Private Machine Learning
• Personalization Improves Privacy-Accuracy Tradeoffs in Federated Learning
• P4: Towards private, personalized, and Peer-to-Peer learning
• Differentially-Private "Draw and Discard" Machine Learning
• User-Level Privacy-Preserving Federated Learning: Analysis and Performance Optimization