Chapter 4: Assessing vulnerabilities and securing federated learning
Supriyo Chakrabortya; Arjun Bhagojib aDistributed AI, IBM Research, Yorktown Heights, NY, United StatesbDepartment of Computer Science, University of Chicago, Chicago, IL, United States
Abstract
The wide applicability and adoption of federated learning stem from its promise of private and efficient decentralized training of models. To achieve decentralized training, federated learning relies on iterative aggregation of model updates from participating clients. This repeated interaction between the server and clients introduces unique vulnerabilities that can be exploited by malicious clients to mount attacks. These attacks can be training-time (such as model or data poisoning) ...
Get Federated Learning now with the O’Reilly learning platform.
O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.
