With the advent of the Internet of Vehicles (IoV), vehicles produce an unprecedented data flow, inspiring innovative data-sharing strategies within the IoV ecosystem. To enhance security, efficacy, and driving experiences, the IoV facilitates real-time information interchange across vehicles, infrastructure, and cloud platforms, thereby transforming transportation. However, there are several concerns with the present IoV, including privacy breaches, security holes, and the possibility of malevolent vehicles uploading false information to the network, which might cause significant traffic jams. Thus, researchers have been focusing on trust management for vehicular networks. The advent of blockchain technology has addressed this concern by making IoV trust management data secure. Unfortunately, blockchain's speed and scalability issues render it unfit for networks operating on a massive scale. To overcome these issues, the article presents Blockchain-MLTrustNet, a novel trust management system, to enhance secure and scalable vehicle-to-everything(V2X) communication in IoV networks hosted on cloud platforms. It combines adaptive graph-sharding blockchain with deep reinforcement learning to create a strong framework for trust management score and transaction validation, and it also comes with a privacy protection scheme to address user privacy concerns. The Adaptive Graph Sharding Blockchain (AGSB) divides the network into adaptable shards that manage a subset of vehicles and transactions. This speeds up transaction processing, reduces latency, and supports more devices. Deep Reinforcement Learning evaluates infrastructure and car reliability concurrently. Comparing Blockchain-MLTrustNet to state-of-the-art approaches (e.g., WGAN-SMOTE, MESMERIC), extensive trials on the VeReMi dataset show that it increases trust score accuracy by 15% and decreases transaction latency by 37%. With its unified design, the system satisfies the specific requirements of the IoV in terms of scalability, security, and privacy, making it suitable for implementation in the real world.
S et al. (Thu,) studied this question.