Enhancing Blockchain Scalability in Federated Learning: Sharding and Multithreading Approaches in Hyperledger Fabric

Blockchain technology has emerged as a promising solution for enhancing privacy and security in decentralized systems, particularly within the domain of federated learning. Federated Learning (FL) enables the collaborative generation of machine learning models while preserving user privacy by keeping data localized. Integrating blockchain with FL ensures trusted participation and safeguards against data tampering. However, blockchain networks face significant scalability challenges, known as the scalability trilemma. This research addresses these limitations by implementing sharding and multithreading techniques within Hyperledger Fabric to optimize its performance. The sharding partitions the blockchain network, enhancing its scalability, while multithreading improves resource utilization during transaction processing. The proposed framework is evaluated across three scenarios: roundrobin sharding, task-based sharding, and no-sharding. Results indicate that task-based sharding outperforms other approaches, achieving an optimal balance between system scalability and resource efficiency. This study provides valuable insights into the trade-offs between performance and resource allocation in blockchain-based federated learning systems.