BIFT: A federated learning System for Connected and Autonomous Vehicles Based on Blockchain

Algorithms for machine learning(ML) are essential for autonomous driving. The vast majority of history’s automated and connected cars send a significant quantity of movement data collected from many automobiles to a central server for regular training. Nonetheless, the info process has made the safety and confidentiality of information essential. Nowadays, the development of federated learning (FL) for data security has the potential to potential to enhance distributed machine learning’s data privacy. The malicious attackers might still target the training process, though. FL is extremely vulnerable because of its total dependency on the central server. We suggest BIFT as a solution to the aforementioned issue, which consists of two parts: 1) a completely decentralized 2) A blockchain-based machine learning system for CAVs that guarantees privacy. With the use of their own driving data, BIFT’s scattered CAVs may train local machine learning models, which they can subsequently upload to improve the global model. In order to fend off potential attackers, BIFT offers a consensus mechanism called Proof of Federated Learning. We assess BIFT’s performance and demonstrate that it is scalable, reliable, and capable of fending off harmful assaults.