Download
🎧 Podcast Summary

Listen to the narrated summary of the paper:

Abstract

Blockchain-enabled federated learning (BCFL) represents a paradigm shift in distributed machine learning, addressing fundamental challenges of trust, privacy, and coordination in collaborative AI systems. This chapter provides a comprehensive architectural analysis of BCFL systems through a systematic four-dimensional taxonomy that examines coordination structures, consensus mechanisms, storage architectures, and trust models. We analyze design patterns ranging from blockchain-verified centralized coordination to fully decentralized peer-to-peer networks, evaluating their trade-offs in scalability, security, and performance. Through detailed examination of consensus mechanisms specifically designed for federated learning contexts, including Proof of Quality (PoQ) and Proof of Federated Learning (PoFL), we demonstrate how computational work can be repurposed from arbitrary cryptographic puzzles to productive machine learning tasks. The chapter includes a comprehensive technical case study of the TrustMesh framework that illustrates practical implementation considerations through a distributed image classification scenario, showing effective collaborative learning across IoT devices with highly non-IID data distributions while maintaining complete transparency. Real-world deployments across healthcare consortiums, financial services frameworks, and IoT security applications validate the practical viability of BCFL systems, achieving comparable accuracy to centralized approaches while providing enhanced security guarantees, comprehensive audit trails, and enabling new models of trustless collaborative intelligence that were previously impossible due to coordination and trust constraints.

Figure 1: Classification Framework for BCFL Systems

Citation
@MISC{rangwala2025blockchainenabledfederatedlearning,
    title={Blockchain-Enabled Federated Learning},
    author={Murtaza Rangwala and Venugopal K R and Rajkumar Buyya},
    year={2025},
    eprint={2508.06406},
    archivePrefix={arXiv},
    primaryClass={cs.DC},
    url={https://arxiv.org/abs/2508.06406},
}