This paper focuses on comparing an RNN and an LSTM model to a CNN-based Intrusion Detection System (IDS) to secure vehicular ad hoc networks (VANETs) against cyber threats like DoS and spoofing attacks, which will become more and more common as technology advances in Internet of Vehicles (IoV). Using different Neural Network models, the Intrusion Detection System is being compared based on accuracy, f1 score, precision, and recall. This helps us to understand which model will perform better under real-time attack scenarios and why it is better than the other models. All the models are trained on a standard input dataset (CICIDS2017) and processed to give an accuracy table. This dataset uses 81 attributes like packet size, packet rate, and other factors to consider for an intrusion attack. The project’s accuracy is considered by how efficiently it identifies an intrusion in the system based on the input. To improve road safety and traffic efficiency, modern vehicular networks, which are essential to the Internet of Vehicles (IoV), allow communication between vehicles and infrastructure. However, the increased connectivity creates a lot of openings for cyberattacks like Distributed Denial-of-Service (DDoS) attacks, data injection, and message tampering. The high mobility and dynamic topology of vehicular networks present challenges for traditional IDS. That is why these challenges are recognized using this comparative study of conventional and new models for Intrusion Detection.