Energy-Efficient and Latency-Aware Blockchain-Enabled Federated Learning for Edge Networks

Abstract

The introduction of blockchain in federated learning can initiate a federation of trustworthy devices by validating the local model. However, challenges arise due to the blockchain framework's increased latency and energy consumption. This brief proposes a blockchain-enabled federated learning framework that jointly optimizes latency and energy consumption. The proposed method is evaluated over the Google speech commands (GKWS) dataset, suggesting that an optimal set of miners helps reduce forking events by 66.67%, latency by 66.30%, and energy consumption by 82.19%. Experimental results show significant improvement in latency and energy consumption when compared to baseline approaches. © 2004-2012 IEEE.