The internet of vehicles faces critical cybersecurity threats such as denial-of-service and spoofing attacks, jeopardizing safety and privacy in connected transport. This study inves-tigates machine learning–based intrusion detection for multiclass attack recognition using a novel vehicle network dataset. We evaluate six models with feature reduction through Recur-sive Feature Elimination and Principal Component Analysis, a strategy not previously applied to Internet of Vehicles security data. Recursive Feature Elimination enhanced Random Forest performance, achieving 99.7% accuracy and an F1-score of 0.73 while reducing training time by 86.2%. Principal Component Analysis preserved 95% variance but reduced effectiveness (F1-score 0.37). Our results indicate that feature reduction not only improves real-time detec-tion time but also outperforms existing Internet of Vehicles intrusion detection research on static dataset in accuracy and performance. This paper demonstrates a new, scalable method for autonomous and connected vehicle’s safety. Future research will explore larger datasets and deep learning integration.