Deep Belief Network Based Intrusion Detection in Vehicle Ad Hoc Network

Abstract

Vehicular Ad Hoc Networks (VANET) have garnered a great deal of attention and development in recent years. These networks are essential to the development of fully and partially autonomous vehicles. But there are a lot of obstacles because of security risks that are either exclusive to VANET or inherent in ad hoc networks. Enhancing driving efficiency, comfort, and safety is VANET's primary goal. These vehicles rely significantly on external communication through data control and cooperative awareness message exchange with the surrounding environment. This study presents the implementation of a Deep Belief Network (DBN) model for VANET intrusion detection. Initially, the DBN is employed for intrusion detection and the dataset is pre-processed for class balance. The DBN model that was put into practice performs better than expected; it achieves 96.32% accuracy, 96.21% recall, 96.87% F1-Score, and 92.95% precision. The suggested DBN system is distinct from other models such as the Optimal Weighted Extreme Learning Machine (OWELM), Gated Recurrent Unit (GRU), and Technique for Oder Preference by Similarity to Ideal (TOPSIS).