Intelligent Spectrum Sensing for CR-VANETs: A Machine Learning Approach for Mobility and Security Challenges

Abstract

Cognitive Radio-enabled Vehicular Ad Hoc Networks (CR-VANETs) are emerging as a vital solution to address spectrum scarcity in intelligent transportation systems. Traditional Dedicated Short-Range Communication (DSRC) suffers from fixed spectrum allocation, making it inadequate for the dynamic and high-bandwidth requirements of Vehicle-to-Vehicle (V2V), Vehicle-to-Infrastructure (V2I), and Vehicle-to-Everything (V2X) communications. To overcome these challenges, this work presents an intelligent spectrum sensing framework that integrates machine learning (ML), trust management, and cooperative sensing to improve detection reliability, throughput, and security in highly mobile vehicular environments. The proposed model formulates spectrum sensing as a binary hypothesis test under Rayleigh fading and evaluates detection and false alarm probabilities using energy detection. Machine learning models including logistic regression, support vector machines, decision trees, random forests, and K-nearest neighbors are employed to optimize adaptive thresholding, sensing time, and mobility-aware performance. Gradient boosting predicts primary user (PU) activity, while Q-learning-based trust mechanisms mitigate malicious attacks such as Primary User Emulation (PUEA) and Spectrum Sensing Data Falsification (SSDF). KMeans clustering further enables localized, delay-sensitive decision-making. Simulation results demonstrate significant improvements in detection accuracy, reduced false alarms, enhanced throughput, and strong resilience against adversarial attacks, making the framework scalable and practical for next-generation vehicular networks.