مؤتمر
Differentially Private Decentralized Learning with Random Walks
العنوان: | Differentially Private Decentralized Learning with Random Walks |
---|---|
المؤلفون: | Cyffers, Edwige, Bellet, Aurélien, Upadhyay, Jalaj |
المساهمون: | Machine Learning in Information Networks (MAGNET), Inria Lille - Nord Europe, Institut National de Recherche en Informatique et en Automatique (Inria)-Institut National de Recherche en Informatique et en Automatique (Inria)-Centre de Recherche en Informatique, Signal et Automatique de Lille - UMR 9189 (CRIStAL), Centrale Lille-Université de Lille-Centre National de la Recherche Scientifique (CNRS)-Centrale Lille-Université de Lille-Centre National de la Recherche Scientifique (CNRS), Université de Lille, Médecine de précision par intégration de données et inférence causale (PREMEDICAL), Inria Sophia Antipolis - Méditerranée (CRISAM), Institut National de Recherche en Informatique et en Automatique (Inria)-Institut National de Recherche en Informatique et en Automatique (Inria)-Institut Desbrest de santé publique (IDESP), Institut National de la Santé et de la Recherche Médicale (INSERM)-Université de Montpellier (UM)-Institut National de la Santé et de la Recherche Médicale (INSERM)-Université de Montpellier (UM), Université de Montpellier (UM), Department of Computer Science Rutgers, Rutgers, The State University of New Jersey New Brunswick (RU), Rutgers University System (Rutgers)-Rutgers University System (Rutgers), Inria-FedMalin, ANR-20-CE23-0015,PRIDE,Apprentissage automatique décentralisé et préservant la vie privée(2020), ANR-22-PECY-0002,iPoP,interdisciplinary Project on Privacy(2022), ANR-20-THIA-0014,AI_PhD@Lille,Programme de formation doctorale en IA à Lille(2020) |
المصدر: | ICML 2024 - Forty-first International Conference on Machine Learning ; https://hal.science/hal-04610660Test ; ICML 2024 - Forty-first International Conference on Machine Learning, Jul 2024, Vienne (Autriche), Austria. ⟨10.48550/arXiv.2402.07471⟩ |
بيانات النشر: | HAL CCSD arXiv |
سنة النشر: | 2024 |
مصطلحات موضوعية: | Machine Learning (cs.LG), Cryptography and Security (cs.CR), FOS: Computer and information sciences, [INFO]Computer Science [cs] |
جغرافية الموضوع: | Vienne (Autriche), Austria |
الوصف: | International audience ; The popularity of federated learning comes from the possibility of better scalability and the ability for participants to keep control of their data, improving data security and sovereignty. Unfortunately, sharing model updates also creates a new privacy attack surface. In this work, we characterize the privacy guarantees of decentralized learning with random walk algorithms, where a model is updated by traveling from one node to another along the edges of a communication graph. Using a recent variant of differential privacy tailored to the study of decentralized algorithms, namely Pairwise Network Differential Privacy, we derive closed-form expressions for the privacy loss between each pair of nodes where the impact of the communication topology is captured by graph theoretic quantities. Our results further reveal that random walk algorithms tends to yield better privacy guarantees than gossip algorithms for nodes close from each other. We supplement our theoretical results with empirical evaluation on synthetic and real-world graphs and datasets. |
نوع الوثيقة: | conference object |
اللغة: | English |
العلاقة: | info:eu-repo/semantics/altIdentifier/arxiv/2402.07471; hal-04610660; https://hal.science/hal-04610660Test; ARXIV: 2402.07471 |
DOI: | 10.48550/arXiv.2402.07471 |
الإتاحة: | https://doi.org/10.48550/arXiv.2402.07471Test https://hal.science/hal-04610660Test |
رقم الانضمام: | edsbas.AE0588A1 |
قاعدة البيانات: | BASE |
DOI: | 10.48550/arXiv.2402.07471 |
---|