Intrusion detection is essential for maintaining the security and reliability of Vehicular Ad-hoc Networks (VANETs), which enable Vehicle-to-Vehicle (V2V) and Vehicle-to-Infrastructure (V2I) communication. The high mobility and dynamic topology of VANETs present unique challenges, requiring efficient and real-time Intrusion Detection Systems (IDS) capable of adapting to rapidly changing network conditions. This study evaluates various machine learning (ML) and deep learning (DL) approaches for intrusion detection using the NSL-KDD dataset as a benchmark. The results demonstrate that the tree-based ML models, specifically decision tree and random forest, significantly outperform DL models, such as recurrent neural networks (RNN) and dense neural network (DNN), achieving accuracies of 99.67% and 99.72%, respectively. In contrast, DL models, while capable of complex pattern recognition, showed limited feasibility for real-time applications due to higher resource demands. This paper highlights the effectiveness of lightweight ML models as practical solutions for secure and scalable IDS deployment in VANET environments, emphasizing their potential for real-time intrusion detection.