CVNNs-IDS: Complex-Valued Neural Network Based In-Vehicle Intrusion Detection System
The bus of the Controller area Network (CAN) in the vehicle is frequently attacked under the environment of efficient communication. This paper explores ways to hide features of the intrusion detection system (IDS) and obtain a high-precision during an at
- PDF / 2,494,209 Bytes
- 15 Pages / 439.37 x 666.142 pts Page_size
- 60 Downloads / 146 Views
Abstract. The bus of the Controller area Network (CAN) in the vehicle is frequently attacked under the environment of efficient communication. This paper explores ways to hide features of the intrusion detection system (IDS) and obtain a high-precision during an attack on the Internet of vehicle (IoV). To protect the privacy features of the hidden layer with regard to anomaly detection, we proposed the CVNNs-IDS. The system converts the data into an image in real-time using the encoder and then maps it into the complex domain whiles it rotates it to reconstruct the real features to achieve the purpose of system protection. Available researches show that features from random angles are obtained by attackers, making it impossible to distinguish between the real or fake feature. The accuracy of the proposed method CVNNs-IDS is 98%. Results obtained represents that our proposed method performed better than the traditional techniques with regard to performance and security.
Keywords: In-vehicle security protection
1
· IDS · CAN · CVNNs · Privacy
Introduction
Although the current rapid development of 5G communication technology has provided sufficient in-vehicle communication guarantee for the development of auto-pilot, the Internet of vehicle (IoV) and cloud service platform connection, however, it has limitations with the maturity of the technology. Without the support of V2X, which ensures communication between exterior devices such as mobiles, vehicles, the Internet, etc., the maturity of the technology cannot be achieved [1]. Hence, it is with regard to the communication process with external devices that the security of IoV is threatened. Hackers employ the use of the Internet attacks(e.g., dos attack, fuzzy attack, malfunction attack, etc.) Supported by the Innovation Plan for Postgraduate Research of Jiangsu Province in 2014 under Grant KYLX1057. National Science Foundation of China under Grant 61902156. Natural Science Foundation of Jiangsu Province under Grant BK20180860. c Springer Nature Singapore Pte Ltd. 2020 S. Yu et al. (Eds.): SPDE 2020, CCIS 1268, pp. 263–277, 2020. https://doi.org/10.1007/978-981-15-9129-7_19
264
M. Han et al.
to attack the electronic control units (ECUs) of the vehicle. It is within the ECUs that various ways of protecting the CAN bus are applied, including the most effective one intrusion detection. Apart from identifying malicious attacks with high accuracy, it also reduces the computational cost as compared to other methods. Most of the existing methods will need to improve the physical layer or generate MAC authentication message based on cryptography or modify the CAN controller in order to be able to protect the security of the CAN bus. However, the existing method increases the cost and computational complexity of the process or experiment. Research available indicates that IDS based on deep learning (DL) is one of the best in detecting an anomaly because it forms a security barrier between the external devices and the internal CAN bus by placing it at the gateway. This h
Data Loading...
