FedDrive-Sec: A blockchain-assisted federated deep reinforcement learning framework for secure and adaptive internet of vehicles

The Internet of Vehicles (IoV) enables real-time data exchange among vehicles, roadside infrastructure, and cloud services, forming a core component of intelligent transportation systems. However, practical IoV deployment faces critical challenges related to data privacy, security vulnerabilities, and efficient computation under highly dynamic network conditions. To address these issues, this paper introduces FedDrive-Sec, a blockchain-enhanced federated deep reinforcement learning framework for secure and adaptive IoV operation. FedDrive-Sec combines privacy-preserving federated learning for decentralized model training, deep reinforcement learning for adaptive resource allocation and security parameter tuning, and a permissioned blockchain for trustworthy model aggregation and Byzantine fault tolerance. The system is formulated as a constrained multi-objective optimization problem, and an adaptive Practical Byzantine Fault Tolerance (PBFT) consensus mechanism is designed to accommodate vehicular mobility. Extensive experiments using real-world vehicular datasets demonstrate that FedDrive-Sec achieves 94.7% intrusion detection accuracy, 38.5% reduction in communication overhead compared to centralized approaches, and 97.3% Byzantine fault tolerance. Moreover, the framework attains a mean latency of 127 ms, blockchain throughput of 1,847 transactions per second, and 42% energy reduction, while effectively mitigating poisoning attacks and preserving data privacy through differential privacy and homomorphic encryption. These results confirm the practicality and robustness of FedDrive-Sec for secure large-scale IoV deployments.