A Neuro-Symbolic and Reinforcement Learning Based Hybrid Framework for Real Time Detection of Distributed Denial of Service Attacks in Internet of Vehicles Networks

Abstract

The Internet of Vehicles (IoV) revolutionizes vehicular communication but introduces crucial vulnerabilities of Distributed Denial of Service (DDoS) attacks that can interfere with real-time data sharing and road safety. Traditional machine learning approaches tended to encounter tradeoffs among detection accuracy, interpretability, adaptability, and real time efficiency. The current work presents NSRL-HAD, a novel hybrid architecture that cooperatively integrates Neural Networks (NN), Decision Trees (DT), Neuro-Symbolic Logic, and Reinforcement Learning (RL) to achieve intelligent, transparent, and self-adaptable DDoS detection for IoV use cases. The architecture employs state of the art dimensionality reduction techniques, such as PCA, SVD, Random Projection, and Robust Matrix Factorization, to attain computational efficiency enhancement as well as noise elimination. Experimental evaluations conducted using the CICIOV2024 dataset demonstrate NSRL-HAD attains 100% accuracy, precision, recall, and F1-score, surpassing the latest state of the art IoV security models. NSRL-HAD provides explainable decisions and adjusts the detection thresholds in real time to varied attack behaviors through a reinforcement learning agent. Its architecture ensures low-latency inference, human-understandable decision rules, as well as resource-limited vehicular system resilience. The results make NSRL-HAD a scalable, real-time, and forensically transparent solution, advancing the frontier of intelligent transport system security.