Advancing Data Heterogeneity Solutions in Blockchain-Enabled Federated Learning for IoT

Federated Learning (FL) offers a collaborative approach to training machine learning models across distributed devices without compromising user privacy. However, a persistent hurdle in its implementation is data heterogeneity— the uneven and diverse data distributions across participating devices—which becomes particularly challenging in Internet of Things (IoT) networks. By combining FL with blockchain, researchers have introduced decentralized, secure, and transparent methods to tackle these issues, yet several gaps and difficulties remain. This paper takes a fresh approach by compiling and analyzing the wide array of proposed solutions and ongoing challenges for managing data heterogeneity in blockchain-integrated FL systems, focusing on their relevance to IoT. We discuss techniques which seem feasible: personalization federated learning through group clustering of devices; adaptive aggregation methods; and blockchain-based reputation systems. In addition, they are mechanism-like reward-based participation and context-aware frameworks, to combat imbalance and thereby enhance fairness in heterogeneous systems. There are still unaddressed issues, including scalability, size of FL networks in blockchain, efficiency versus performance trade-offs, and, above all, the challenge of ensuring that heterogeneous systems can interoperate. By presenting an interconnected review of advances and bottlenecks, the current work is able to inform future developments of resilient and scalable FL systems for IoT applications.