NOVA: A hybrid detection framework for misbehavior in vehicular networks

Abstract

Vehicular networks, comprising communication between two vehicles (Vehicle-to-Vehicle, V2V) and communication between a vehicle and its environment (Vehicle-to-Everything, V2X), are critical in improving road safety, traffic management, and smart transport systems. However, the interconnectivity of these systems makes them susceptible to various security threats, including Denial-of-Service (DoS), Sybil, and spoofing attacks. Common Intrusion Detection Systems (IDS) have significant limitations in their approaches, such as static reputation scoring to match attacks, small attack scope detection, and limited scalability in high node density. In this paper, we propose the hybrid detection framework, NOVA, by utilizing both statistical anomaly detection and machine learning techniques to ensure a comprehensive security solution for vehicular networks. Recognizing the importance of real-time adaptation to the dynamic nature of peer-to-peer networks, NOVA implements a sophisticated reputation management system that scales to ever-changing environments. Additionally, a trusted node mechanism is integrated, securing critical infrastructure nodes through cryptographic authentication and communication prioritization. This allows NOVA to operate in a distributed architecture with the support of vehicular cloud integration for handling networks with high density while guaranteeing performance. NOVA outperforms all existing schemes with a high detection rate (around 97% average for multiple attack types), lower false positive and false negative rates, and stable performance scalability up to 500 nodes, as extensive simulation results have shown. Comparisons against state-of-the-art systems demonstrate how NOVA performs better in terms of accuracy and scalability, establishing NOVA as a promising solution to facilitate secure intelligent transportation networks in the future.