KEDenNet-MEO: A Knowledge-Driven Deep Intrusion Detection Framework for VANETs with Realistic Mobility Simulation

Vehicular ad-hoc networks (VANETs) are increasingly targeted by sophisticated cyber-attacks that can jeopardize driver safety and traffic efficiency. This study offers a flexible intrusion-detection framework that combines realistic traffic simulation with a knowledge-driven deep classifier to detect spoofing, denial-of-service, and data-injection attacks. Using SUMO to generate microscopic urban mobility and NS-3 to emulate IEEE 802.11p communications, to synthesize labelled packet traces in which adversarial events are precisely controlled. After cleaning and temporal batching, features are pruned by a modified elite opposition-based artificial hummingbird optimization algorithm (MEO-AHBOA), which balances exploration through opposition sampling with exploitation via elitist learning, shrinking the dimensionality by 42 % while preserving discriminatory content. The reduced vectors feed a knowledge-embedded dense network (KEDenNetwork) who’s densely connected blocks are augmented with protocol-aware gating units that encode physical and MAC-layer heuristics. Experimental evaluation demonstrates state-of-the-art performance: 98.67 % binary accuracy, 98.48 % F1, and an AUC of 0.997, surpassing CNN, LSTM, GRU, and SVM baselines by up to 5.2 %. Multiclass analysis further attains balanced F1-scores of 99.08 % (normal traffic), 97.68 % (spoofing), 98.32 % (DoS), and 98.27 % (data injection). Ablations confirm that KEDenNetwork alone contributes 2.3 pp, while MEO-AHBOA adds an additional 1.4 pp to macro-F1. The modular architecture permits drop-in replacement of attack libraries, feature selectors, and classifiers, facilitating reproducible benchmarking across future VANET security studies. Our findings highlight that embedding domain knowledge into dense topologies, coupled with adaptive metaheuristic selection, offers a compelling route towards lightweight yet highly reliable on-board intrusion detection for next-generation intelligent transportation systems.