Artificial Intelligence-based intrusion detection system for V2V communication in vehicular adhoc networks

Vehicular Adhoc Networks (VANETs) enable vehicle-infrastructure communication, enhancing Intelligent Transportation Systems (ITS). Vehicles interconnect wirelessly to share information. Still, this communication is vulnerable to various attacks, particularly in Vehicle-to-Vehicle (V2V) scenarios. This study proposes an Artificial Intelligence-based Intrusion Detection System (IDS) in an edge-envisioned environment, combining edge computing and deep learning. The approach uses the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) for edge node selection to perform packet forwarding and employs a Bidirectional Generative Adversarial Network (BiGAN) for intrusion detection. This poses a challenge since regular traffic samples are typically more prevalent than anomalous traffic, and the dataset is unstructured and highly imbalanced. Our BiGAN-based model overcomes synchronization challenges between generator and discriminator networks. Training iterations increase independently until cross-entropy loss conditions are met. The trained encoder-discriminator network is a single-class classifier, distinguishing normal and abnormal data. Experimental results demonstrate superior performance on the NSL-KDD dataset compared to similar methods.