Blockchain-Enabled Personalized Federated Learning for Autonomous Vehicles

Modern self-driving autonomous vehicle (AV) biomes need artificial intelligence (AI) structures that are adaptable, strong, and safeguard clients’ confidentiality to provide decentralized and distributed privacy-preserving machine learning across various manufacturers’ edge devices and cloud environments. With conventional machine learning techniques, confidential vehicle details are transferred and stored in a central server, which affects insignificant reliability and seclusion concerns. This problem is addressed by federated learning (FL), which enables distributed model training instead of attempting to transfer data directly. Despite this FL is still at risk for vulnerabilities like model poisoning attacks, its limited customization options, its dependence on centralized aggregation techniques. This paper introduces a block chain-based personalized federated learning (BPFL) framework that uses distributed server edges, cloud-level collectors, plus smart contractors to ensure proxy-based monitoring to overcome the limitations. Secure model interchange, distributed trust control, tamper-proof secure verification, and personalization across various AV settings are all made possible by proposed architecture. Integration with block chain improves transparency, guarantees security, and strengthens against malicious manipulation. Several experimental tests shows that the proposed BPFL framework significantly enhances resistance to malicious updates and produces better global model performance with an accuracy of 94.2% while reducing communication overhead by 23% when compared to standard FL strategies.