Intrusion Detection System Using Deep Learning for Vehicular Ad Hoc Network

Abstract

In recent years, there has been a lot of interest in study of vehicular ad hoc networks (VANETs). VANETs are essential to the development of self-driving and semi-au tonomous vehicles since they increase comfort and safety. However, the security risks that are either present in ad hoc networks or specific to VANET pose significant difficulties. VANETs could be targets of numerous attacks. To defend the external communication system from intrusions, this study provides an intrusion detection system (IDS) for VANET that makes use of anomaly detection. In this research, we proposed a deep learning-based Attentive Interpretable Tabular Learning (Tab Net) architecture-based intrusion detection system for VANET. Deep learning Tab Net model exhibits strong interpretability and quick learning rates. The proposed system obtains with an accuracy of 98.12%, precision of 98.86% and a False Alarm Rate (FAR) of 0.78%. The principles of the Tab Net architecture are used for VANET security for the first time in research of VANETs. The developed methodology is tested on the real-time CIC-IDS 2018 dataset and the results of the experiment are contrasted with those obtained using other cutting-edge techniques.