Federated Learning for IoT: Applications, Trends, Taxonomy, Challenges, Current Solutions, and Future Directions

Abstract

The rapid advancement of Internet of Things (IoT) technology has transformed the digital landscape, enabling unprecedented connectivity between devices, people, and services. Traditionally, IoT-generated data was processed through centralized, cloud-based machine learning (ML) systems, raising significant privacy, security, and network bandwidth concerns. Federated Learning (FL) presents a viable alternative by transmitting only model parameters while preserving local data privacy. Despite the growing body of research, there remains a gap in comprehensive studies on FL-enabled IoT systems. This review provides an in-depth examination of the integration of FL with IoT, highlighting how FL enhances the efficiency, robustness, and adaptability of IoT systems. The paper introduces the foundational principles of FL, followed by an exploration of its key benefits in decentralized IoT applications. It presents a comparative analysis of FL-IoT architectures using quantitative metrics and proposes a taxonomy that clarifies the complexities and variations in FL-enabled IoT systems. The challenges of deploying FL in IoT environments are discussed, along with current trends and solutions aimed at overcoming these hurdles. Furthermore, the review explores the integration of FL with emerging technologies, including foundational models (FMs), green and sustainable 6th-generation (6G) IoT networks, and deep reinforcement learning (DRL), emphasizing their role in enhancing FL’s efficiency and resilience. It also covers FL frameworks and benchmarks, providing a valuable resource for researchers and practitioners in the field The article concludes by identifying promising research directions that are expected to drive future advancements in this dynamic and expanding field.