A Robust Machine Learning-Driven Framework for Efficient Spectrum Sensing in Next-Generation Vehicular Networks

Abstract

Cognitive Radio-enabled Vehicular Ad Hoc Networks (CR-VANETs) are rapidly emerging as a transformative solution to overcome spectrum scarcity in intelligent transportation systems. As vehicular networks evolve to support real-time, high-bandwidth applications across Vehicle-to-Vehicle (V2V), Vehicle-to-Infrastructure (V2I), and Vehicle-to-Everything (V2X) communications, the limitations of static spectrum allocation under the Dedicated Short-Range Communication (DSRC) framework have become evident. Cognitive Radio (CR) offers dynamic spectrum access, allowing unlicensed vehicular nodes to opportunistically utilize underused licensed spectrum without interfering with primary users (PUs). However, effective spectrum sensing in VANETs remains a formidable challenge due to high mobility, fluctuating channel conditions, varying Quality of Service (QoS) requirements, and security threats such as Primary User Emulation Attacks (PUEA) and Spectrum Sensing Data Falsification (SSDF). This paper presents an intelligent and adaptive spectrum sensing framework that integrates traditional signal processing with machine learning (ML) models including logistic regression, SVM, decision trees, random forests, and K-nearest neighbors. The model leverages both centralized and decentralized cooperative sensing while dynamically adjusting sensing parameters based on vehicular mobility, SNR conditions, and PU behavior. Security is enhanced through trust-aware Q-learning, which assigns reliability scores to secondary users to mitigate malicious behavior. A novel segmentation technique using KMeans clustering reduces latency and supports localized sensing decisions. Simulation results show significant improvements in detection accuracy, reduced false alarm rates, enhanced throughput, and strong resilience against adversarial attacks. The proposed framework demonstrates how ML can optimize sensing performance and spectrum utilization under realistic vehicular scenarios. By addressing key research gaps—such as the lack of mobility-aware PU models, adaptive QoS provisioning, and secure cooperative sensing—this work contributes a robust and scalable solution to spectrum management in CR-VANETs, paving the way for more efficient, secure, and intelligent vehicular communication networks.