이민종, 하재철, 김수형,

Vol. 35, No. 2, pp. 359-368, 4월. 2025
10.13089/JKIISC.2025.35.2.359, Full Text:
Keywords: VANET Security, Deep Learning, Anomaly Detection
Abstract

Cooperative-Intelligent Transportation Systems (C-ITS) can leverage real-time vehicle-to-vehicle communications in VANETs to provide transportation safety and efficiency. However, its open nature exposes it to security threats such as malicious nodes. To address this, we propose a PacGAN-based VAEGAN framework for anomalous behavior detection in vehicular networks. By incorporating PacGAN and D2GAN, the proposed model can enhance training stability and effectively capture complex data distributions. It also exploits the spatiotemporal characteristics of BSMs to achieve robust anomaly detection under diverse road and network conditions. According to the experimental results, the proposed framework achieved an F1 score of 93.59% and an accuracy of 90.31%. These results show that it outperforms existing methods. Therefore, our approach contributes to the reliability and safety of VANET-based C-ITS in complex vehicular environments.

Statistics
Cite this article