Data privacy issue in Federated Learning Resolution using Block Chain

Abstract

Reliable and timely traffic patterns have become an increasingly critical aspect of intelligent transport networks for traffic control. However, current predictive models of traffic flow focused on centralized machine learning need to capture raw data for model education that entails significant privacy risks. Federated learning (FL) that exchanges model changes without raw data sharing have recently been launched as an innovative way to resolve these concerns to the security of personal data. The current federal learning system is based on a central coordinator model dealing with serious security problems, including a single failure point. The Literature review is presented in this study will focus on the integration of blockchain in federated learning. Federated learning (FL) based on blockchain is thus suggested in the paper to support decentralized, efficient, and stable federated learning. The model includes three primary components; customers, blockchain technology, and machine learning engineers. In addition, different methods are communicated to protect model privacy on the blockchain further, and the advantages of the new learning system federated built on the blockchain are discussed.