A Cyber Resilient Framework for V2X Enabled Roundabouts in Intelligent Transportation Systems

Abstract

Vehicle-to-everything (V2X) communication systems are increasingly susceptible to cyber-physical threats that exploit trust assumptions, coordination latency, and semantic inconsistencies across agents. These vulnerabilities, particularly in dense or adversarial environments, undermine the reliability of cooperative perception, anomaly detection, and safety-critical maneuver execution. This paper presents CR-V2XR, a cross-layer, federated, and trust-aware coordination framework designed to enhance resilience in connected and autonomous vehicular networks. CR-V2XR integrates multi-modal anomaly detection with delay-sensitive trust estimation using features extracted from basic safety messages (BSMs), behavioral deviations, entropy shifts, and inter-vehicle trust validation. The architecture employs federated learning for distributed anomaly detection without centralized aggregation and uses a control layer that supports delay-aware trajectory selection. A multi-objective NSGA-III optimizer enables online trade-off adaptation across detection accuracy (DA), collision probability, and communication overhead. Simulations across eleven adversarial scenarios, including Sybil, wormhole, falsification, and replay attacks, demonstrate that CR-V2XR achieves 95% detection accuracy under worst-case attacks, reduces collision probability from 0.61 to 0.27 at 300 vehicles, maintains bounded delay, typically 45–60 ms under nominal load, and remains resilient under high-stress conditions with delays up to 180 ms and communication overhead ( $\leq 6.1$ MB/s). Compared to centralized IDS and stateless baselines, CR-V2XR improves detection fidelity, scalability, and robustness under non-IID data and partial synchronization. These results establish CR-V2XR as a viable architecture for delay-constrained, trust-centric coordination in federated V2X environments subject to persistent adversarial threats.