A trust-aware edge-assisted reinforcement learning framework for secure and low-latency vehicular communication

Abstract

Abstract Vehicular Ad Hoc Networks enable real-time intelligent transportation services but remain affected by dynamic mobility, unstable routing, congestion, malicious vehicles, and delayed emergency communication. This paper proposes a Trust-Aware Edge-Assisted Reinforcement Learning framework, termed TEARL-VANET, for secure and low-latency vehicular communication. The framework integrates multi-source dynamic trust evaluation, reliability-weighted edge aggregation, Deep Q-Network-based trust-aware routing, secure emergency-message dissemination, and adaptive resource allocation. Direct, indirect, and behavioural observations are used to estimate vehicle trust, while RSUs and edge servers validate and aggregate distributed trust information. The DQN agent selects trusted routes using mobility, congestion, communication quality, neighbourhood density, and resource availability. TEARL-VANET was evaluated using SUMO, OMNeT + + , Veins, INET, and the VeReMi Extension dataset and compared with AODV, GPSR, TrustChain-VANET, EdgeTrust-VANET, RL-VANET, IDRL-VANET, ATRL-VANET, and GNN-DRL-VANET. The proposed framework achieved 98.2% packet delivery, 99.1% emergency-message delivery, 97.4% attack detection, 15.8 Mbps throughput, and 42 ms end-to-end delay. Across 10 independent runs, paired t-tests against the strongest baseline confirmed statistically significant improvements in packet delivery, attack detection, throughput, and delay, with all corresponding p-values below 0.01. These results demonstrate the effectiveness of TEARL-VANET for reliable, secure, and adaptive vehicular communication.