Cognitive Radio-enabled Vehicular Ad Hoc Networks (CR-VANETs) are emerging as a vital enabler for next-generation intelligent transportation systems by addressing spectrum scarcity in highly dynamic vehicular environments. Traditional spectrum sensing techniques often fail to ensure reliable detection due to mobility, fading, and security threats such as Primary User Emulation and Spectrum Sensing Data Falsification attacks. This paper proposes a robust machine learning-driven framework that integrates logistic regression, support vector machines, support vector regression, and K-nearest neighbors for adaptive spectrum sensing. SVR algorithm introduced for segmentation minimizes latency by enabling localized and delay-sensitive sensing decisions. The proposed approach effectively balances sensing accuracy, security, and better sensing time, validated through simulation results that demonstrate improved detection probability, reduced false alarms. The framework provides a scalable, secure, and mobility-aware solution for future vehicular networks.