Naseri, Mohammad; (2024) Towards Private and Robust Federated Learning. Doctoral thesis (Ph.D), UCL (University College London).

Text PhD_Thesis.pdf - Accepted Version Access restricted to UCL open access staff until 1 March 2025. Download (12MB)

Abstract

Federated Learning, an evolving paradigm in decentralized machine learning, has captured considerable interest as a compelling approach for collaborative model training across distributed data sources. However, the inherent structure of federated learning introduces privacy and robustness challenges in its conventional form. This thesis delves into an in-depth analysis of these issues, exploring potential solutions to enhance the privacy and robustness of federated learning. Our research thoroughly examines different federated learning paradigms, including horizontal vs.vertical data partitioning, synchronous vs. asynchronous syncing strategies, and the scrutiny of existing security vulnerabilities alongside the proposal of novel threats. Furthermore, we delve into a specific application of federated learning—predicting security events—conducting comprehensive evaluations from various perspectives. In summary, this thesis offers a comprehensive and multifaceted examination of privacy and robustness in the domain of federated learning. We evaluate practical solutions aimed at safeguarding both privacy and robustness while maintaining the optimal performance of the training model. This thesis contributes to the progression of private and robust federated learning.

Download activity - last month

Export as JSONCSVXML

Download activity - last 12 months

Export as JSONCSVXML

Downloads by country - last 12 months

Export as XMLCSVJSON

Archive Staff Only

View Item