What question did this study set out to answer?

May 1, 2026Open Access

A Review Of Federated Learning: Privacy-Preserving Machine Learning

Key Points

The review aims to explore the privacy and confidentiality aspects of federated learning and its fundamental algorithms.
Critical review of privacy techniques like differential privacy, secure multi-party computation, and homomorphic encryption.
Examination of system vulnerabilities to attacks and statistical challenges in federated learning systems.
Comparison of federated learning applications across industries including healthcare, mobile communication, and finance.
Identified key privacy-enhancing techniques, including differential privacy and secure multi-party computation.
Outlined challenges such as statistical and system heterogeneity affecting federated learning.
Proposed a prioritized set of research objectives to advance privacy-preserving machine learning over the next decade.

Abstract

Federated Learning (FL), was created by McMahan et al (14), has become of interest because it offers a decentralized machine learning framework for developing large scale ML models. This allows many users (or clients) to collaborate on training a shared model while retaining control of their own data. FL is ultimately designed to provide a solution to the conflict between the data demands of machine learning systems and the desire of individuals/companies to keep their personal and commercial data private. This paper is a review of the privacy and confidentiality aspects of Federated Learning. A critical review of the fundamental algorithms used in FL, possible attacks against FL systems, and the four primary techniques for enhancing privacy in FL; Differential Privacy (DP), Secure Multi-Party Computation (SMPC), Homomorphic Encryption (HE), and hardware based Trusted Execution Environments (TEE), is provided. We will review aggregation protocols, determine the strength of FL systems against poisoning and inference attacks, and compare various FL systems implemented in three industries; healthcare, mobile communication and finance. A detailed review of FL reveals research issues related to; statistical heterogeneity, communication overhead, system heterogeneity and fairness. Finally, this review presents a prioritized set of research objectives for the next ten years, with an emphasis on situating FL within the larger context of privacy-preserving ML and potential regulatory developments.

A Review Of Federated Learning: Privacy-Preserving Machine Learning

Key Points

Abstract

Cite This Study