Federated Learning: Privacy, Security and Hardware Perspectives

Machine Learning (ML) models are being deployed in a wide range ofdomains owing to their capacity to deliver high performance across a range ofchallenging tasks including safety-critical and privacy-sensitive applications.Moreover, the computing requirements of increasingly complex ML modelspresents a significant challenge to the hardware industry.

Against this backdrop, Federated Learning (FL) has emerged as a promis-ing technique that enables privacy-preserving development of ML models onlow-energy Edge devices. FL is a distributed approach that enables learningfrom data belonging to multiple participants, without compromising privacysince user data are never directly shared. Instead, FL relies on training aglobal model by aggregating knowledge from local models. Despite its repu-tation as a privacy-enhancing strategy, recent studies reveal its susceptibilityto sophisticated attacks that can undermine integrity and, as well as disrupttheir operations. Notably, the constraints posed by the limited hardwareresources in edge devices compound these challenges. Gaining insight into these potential risks and exploring hardware-friendly solutions is vital foreffectively implementing trustworthy and power-efficient FL systems in edgeenvironments.

This chapter contributes a review and perspective of the triad of privacy, security, and hardware optimization in FL settings.