Scalable and Secure Framework for Federated Learning in IoV Applying Committee-based Chained Hotstuff

Applying Federated Learning (FL) in the Internet of Vehicles (IoV) presents a critical trade-off between privacy, security, and performance, which conventional blockchains fail to address due to prohibitive latency and overhead. To overcome these limitations, we propose a novel framework featuring a stack-based verification engine for deterministic validation, an on-chain management system for conditional privacy, and a Committee-based Chained HotStuff (C-CHS) consensus algorithm. The C-CHS protocol fundamentally resolves the leader bottleneck by enabling parallel block preparation. Our evaluation demonstrates substantial performance gains: C-CHS increases throughput by up to 9 2 \% (vs. CHS) and 8 7 3 \% (vs. PBFT), while reducing latency by up to 52 % and 90 %, respectively. These results validate our framework as a secure, scalable, and practical solution for deploying FL in large-scale IoV environments.