Privacy-preserving federated learning in the internet of vehicles (IoV) requires low-latency authentication, bounded privacy leakage, and robustness against malicious model updates. However, most existing studies separately design communication authentication and federated learning protection, which leads to duplicated overhead and weak resistance to cross-layer attacks. To address this issue, this paper proposes a DAG blockchain-enabled cross-layer authentication framework for trustworthy IoV federated learning (DAG-CTFL). The framework reuses authentication operations across V2X message verification and model-update delivery, incorporates trust-aware batch verification, and organizes cross-layer evidence through a two-tier DAG blockchain. In addition, differential privacy is used to reduce information leakage from uploaded model updates, while cross-layer trust evaluation improves resilience against poisoning and forged-identity attacks. Experimental results on MNIST and CIFAR-10 show that DAG-CTFL reduces single-message verification overhead by 8.2–56.1%, lowers batch-verification latency by 19.2–56.4%, and maintains model accuracy above 85% under 15% malicious nodes. These results demonstrate that DAG-CTFL achieves an effective balance among privacy preservation, authentication efficiency, and cross-layer robustness in IoV federated learning.
Liao et al. (Tue,) studied this question.