This paper proposes an anomaly detection algorithm based on Generative Adversarial Networks (GANs) for intrusion detection in Intelligent Connected Vehicles (ICVs). To address the imbalance between normal network traffic and attack traffic in vehicular networks, as well as the difficulty in collecting attack data, we train the neural network solely with normal traffic. We employ the Gini index to perform data dimensionality reduction, selecting 34 features with the highest purity of information and enhancing model training speed via normalization functions. Our algorithm combines the generative capabilities of GANs and the reconstructive power of autoencoders, detecting abnormal network data by learning the distribution of normal network data, thus eliminating the need to analyze packet payloads and safeguarding user privacy. We also develop a vehicular network traffic data collection scheme to collect network traffic data from ICVs. Experimental results indicate that our algorithm demonstrates high accuracy in detecting anomalous network traffic in ICVs, with an accuracy rate of 94.9%.
GAN-IVDS: An Intrusion Detection System for Intelligent Connected Vehicles Based on Generative Adversarial Networks
18.08.2023
792797 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Crack Detection Based on Generative Adversarial Networks and Deep Learning
Springer Verlag | 2022
|