Federated Learning-Based Privacy Preservation in Internet of Things Applications for Smart Healthcare and Intelligent Cities: A Comprehensive Review

Abstract

The rapid proliferation of Internet of Things (IoT) devices across smart healthcare systems and intelligent city infrastructures has ushered in an unprecedented era of data-driven automation, real-time analytics, and personalized services. However, this vast interconnected ecosystem relies heavily on continuous streams of highly sensitive personal information, including real-time patient physiological vitals, electronic health records (EHR), urban mobility patterns, surveillance video feeds, and smart grid utility consumption metrics. Conventional centralized machine learning paradigms require harvesting raw edge data into cloud data centers, creating immense privacy vulnerabilities, single points of failure, regulatory compliance bottlenecks (e.g., GDPR, HIPAA), and prohibitive communication bandwidth consumption. Federated Learning (FL) has emerged as a disruptive privacy-preserving decentralized machine learning framework that enables heterogeneous IoT edge devices to collaboratively train a shared global model while retaining all sensitive raw data locally. Despite its intrinsic privacy advantages, standard FL remains vulnerable to advanced privacy leakage threats, such as gradient inversion attacks, membership inference, model poisoning, and communications overhead under non independent and identically distributed (non-IID) edge data regimes. This comprehensive review systematically examines the state-of-the-art FL architectures, privacy-preserving enhancements (including Differential Privacy, Homomorphic Encryption, and Secure Multi-Party Computation), and resource-optimization mechanisms tailored for IoT-enabled smart healthcare and intelligent city applications. We critically analyze existing literature, identify fundamental trade-offs between privacy guarantees, model accuracy, and communication efficiency, highlight key technical research gaps, and propose a taxonomy of future research directions toward resilient, scalable, and privacy-centric IoT environments.

KEYWORDS: Federated Learning, Internet of Things (IoT), Privacy Preservation, Smart Healthcare, Intelligent Cities, Differential Privacy, Homomorphic Encryption, Edge Computing.