Federated Learning in IoT Environments: Examining the Three-Way See-Saw for Privacy, Model-Performance, and Network-Efficiency

This survey paper provides an in-depth exploration of Federated Learning (FL) in Internet of Things (IoT) environments, focusing on privacy-preserving techniques and their influence on model performance and network efficiency. It highlights key challenges and opportunities at the intersection of these technologies by offering a comprehensive review of FL applications in IoT. First, a customized taxonomy is introduced to evaluate the privacy levels, quality of service (QoS) and network efficiency of various Privacy-Preserving FL (PPFL) solutions in IoT configurations. Furthermore, the survey investigates strategies to improve FL accuracy while addressing resource and network constraints, both independently and together with privacy preservation techniques. Our findings underscore the complexity of optimizing resource utilization, learning performance, and privacy resilience, revealing that no single PPFL solution universally applies. The paper further identifies future research directions, including the integration of advanced technologies beyond 5G networks, and discusses standards, protocols, real-world PPFL projects from world-renowned industries for potential IoT applications.