A CGAN-based Few-shot Method for Zero-day Attack Detection in the Internet of Vehicles
Bingfeng Xu, Bo Wang, Xinkai Chen, Jincheng Zhao, Gaofeng He · 2023
Due to the limited availability of labeled attack data, the detection of zero-day attacks in the Internet of Vehicles domain often relies on an anomaly-based approach. However, this approach frequently leads to a high false positive rate. In practice, we have observed that the principles behind zero-day attacks and known attacks are similar within the Internet of Vehicles environment. Inspired by the observation, this paper proposes a conditional generative adversarial network-based few-shot method for zero-day attack detection in the Internet of Vehicles environment. Primarily, a conditional adversarial generative network model with multiple generators and multiple discriminators is proposed. With this framework, an adaptive sampling data augmentation method is designed to augment data with known attack samples by optimizing the input samples of this model to reduce the false positive rate. Moreover, to alleviate the data imbalance problem caused by a few input attack samples of this model, a collaborative focus loss function is provided in the discriminators, which focuses on discerning challenging-to-classify data. Lastly, the proposed method’s effectiveness is evaluated through comprehensive experiments carried out on the F2MD vehicle network simulation platform. The experimental results demonstrate the superiority of the proposed method compared to existing approaches, both in terms of detection effectiveness and latency. As a result, this paper presents a practical and promising solution for zero-day attack detection in the Internet of Vehicles domain.