Federated Learning Approaches for Decentralized Data Processing in Edge Computing

Federated Learning (FL) has evolved into a decentralized Machine Learning (ML) paradigm with a high potential for enabling model training across various Edge Devices (ED) to conserve data privacy. This survey explores the challenges and future directions of implementing FL in edge computing environments. This Comprehensive review addresses key technical challenges, including the heterogeneity of edge devices, resource constraints, security and privacy concerns, and communication overhead. To guide future research, this study highlights several promising research directions, including advanced model aggregation techniques, edge device optimization, cross-domain federated learning applications, robust security and privacy frameworks, and regulatory compliance. Additionally, This study reviews various Federated Learning methods-such as Federated Averaging (FedAvg), Federated Learning with Secure Aggregation (FedSecAgg), Hierarchical Federated Learning, and Peer-to-Peer Federated Learning-discussing their strengths, weaknesses, and suitable use cases. Finally, this review paper presents benchmarking results using standard datasets and performance metrics, comparing these approaches’ efficiency, accuracy, scalability, and privacy preservation. This systematic review renders a broad perspective on the current trends and developments in federated Learning for edge computing, this analysis enlightenments researchers and Professionals with practical experience.