Nautilus: A Verifiable Hierarchical Federated Learning Framework for Vehicular-Edge-Cloud Systems

Federated Learning (FL) enables privacy-preserving collaborative learning for Internet of Vehicles (IoV) scenarios, but the extreme heterogeneity of vehicular-edge-cloud resources severely limits system efficiency. While dynamic scheduling strategies can mitigate this issue, they introduce new trust concerns: how to verify that scheduling decisions are fair, and whether clients faithfully execute optimization instructions without disclosing private data? This paper proposes Nautilus, a verifiable and efficient federated learning framework. First, we design a multi-dimensional resource-aware scheduling algorithm that dynamically allocates compression ratios and training tasks based on vehicles bandwidth, latency, and computing power, significantly improving system training efficiency. Second, to address the trust deficit in the scheduling process, we introduce a Zero-Knowledge Proof (ZKP) mechanism that ensures the fairness of scheduling strategy generation and the compliance of client execution while preserving privacy. Experimental results demonstrate that the framework effectively reduces communication overhead and accelerates model convergence while maintaining system integrity.