Security and privacy concerns in Federated Learning systems: a systematic review

Read the full article See related articles

Listed in

This article is not in any list yet, why not save it to one of your lists.
Log in to save this article

Abstract

Federated Learning is a Machine Learning solution that trains a global model by aggregating weights from different peers. Federated Learning does not require that data be shared among nodes; however, it is not exempt from privacy and/or security issues. This systematic review focuses on the major security and privacy threats related to the definition and implementation of Federated Learning frameworks. This study aims to provide a comprehensive analysis of potential adversary cyber attacks throughout the execution of Federated Learning, in order to characterize and classify Federated Learning protocols capable of addressing critical robustness concerns — including privacy-preserving techniques, local data protection, efficiency, and accuracy — while highlighting the critical points that remain to be addressed.

Article activity feed