Answers are grounded in this paper and cite their sources.
Integrating blockchain with Federated Learning (FL) offers a promising approach for decentralized and secure machine learning. However, data redundancy presents significant challenges, such as high storage demands, increased bandwidth usage, and reduced scalability. To tackle these issues, this study introduces a middleware-based data deduplication framework designed to optimize data management in Blockchain-based FL (BBFL) settings. The middleware employs Content-based Deduplication (CBD) and Exponential Growth Deduplication (EXGD) to effectively remove redundant model weight updates while maintaining essential contributions. Experimental evaluations show that the EXGD method achieves the highest storage efficiency (81.87%), lowest system latency (32.8 ms), and best blockchain scalability (39.3 transactions per block, supporting up to 55 nodes). Additionally, the Hybrid-based Deduplication (HBDD) approach, which combines CBD and EXGD, achieves an 86.94% deduplication rate, 78.19% storage efficiency, 72.26% bandwidth savings, and supports 44 transactions per block, making it the most balanced solution for optimizing FL performance. Importantly, these optimizations do not compromise learning accuracy, ensuring effective and scalable FL deployment. This research presents an innovative and practical deduplication framework that enhances the performance and resource efficiency of decentralized machine learning systems, with potential applications in IoT, healthcare, and financial services.
No takes yet. Share an insight, caveat, or question.
Garba et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: