AN ADAPTIVE MAN-IN-THE-MIDDLE ATTACK RESISTANT TRUST MODEL FOR IMPROVED SECURITY IN VEHICULAR AD HOC NETWORKS

Abstract

Vehicular Ad hoc Networks (VANETs) enable real-time traffic communication between vehicles and Roadside Units (RSUs) to improve traffic and road safety in smart cities. However, their decentralized and highly dynamic nature makes them vulnerable to security threats, especially Man-in-the-Middle (MitM) attacks, in whichmalicious nodes intercept, delay, modify, or drop safety-critical messages. Existing trust-based models, such as the Man-in-the-Middle Attack Resistant Trust Model in Connected Vehicles (MARINE), use static trust evaluation parameters, which limit their effectiveness in dynamic VANET settings. Other techniques, including blockchain-based trust management, path-backtracking schemes, and reinforcement learning-based models, introduce adaptive features but face challenges such as high computational overhead, scalability issues, and limited integration of multiple trust factors. The purpose of this research was to design and validate an Adaptive Man-in￾the-Middle Attack Resistant Trust Model in Connected Vehicles (A-MARINE). Specifically, the study examined the adaptive features of trust-based models, developed an adaptive MARINE technique with real-time dynamic adjustments to trust scores, and validated its performance through simulation experiments. The research design employed was an experimental approach, utilizing Simulation of Urban Mobility (SUMO), OpenStreetMap (OSM), Traffic Control Interface (TraCI),and Python. The simulation environment included 100 vehicular nodes and 10 Roadside Units, with attacker levels ranging from 10% to 40%. The adaptive model combined historical data, real-time behavior analysis, contextual awareness, monitoring of message frequency, dynamic adjustment of trust scores, and real-time learning for attack detection. The limitation of this study was that only100 nodes were employed meaning scalability of the model was not tested, but this was beyond the scope. Results indicate that the adaptive MARINE outperformed the baseline MARINE model across all attacker levels, achieving average improvements 0.05% in precision, 0.24% in recall, 0.16% in F-Score, and 0.07% in accuracy. These findings suggest improved detection of malicious nodes and higher trust evaluation accuracy. Future work should focus on extending the model to other attacks, such as Sybil and Denial-of-Service (DoS), and on integrating social media feedback to enhance scalability and contextual awareness.