Data-Driven Intrusion Detection for Intelligent Internet of Vehicles: A Deep Convolutional Neural Network-Based Method

• Published in: IEEE transactions on network science and engineering Vol. 7; no. 4; pp. 2219 - 2230
• Main Authors: Nie, Laisen, Ning, Zhaolong, Wang, Xiaojie, Hu, Xiping, Cheng, Jun, Li, Yongkang
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 01.10.2020; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; Artificial neural networks; Back propagation; Back propagation networks; Computer architecture; Convergence; convolutional neural network; Convolutional neural networks; Cybersecurity; data-driven; Deep learning; Feature extraction; Industrial applications; Intelligent transportation systems; Internet of Things; Internet of Vehicles; Intrusion detection; Intrusion detection systems; Machine learning; Malware; Neural networks; Roadsides; Smart cities; Training; Transportation networks; Vehicular ad hoc networks
• ISSN: 2327-4697; 2334-329X
• DOI: 10.1109/TNSE.2020.2990984

As an industrial application of Internet of Things (IoT), Internet of Vehicles (IoV) is one of the most crucial techniques for Intelligent Transportation System (ITS), which is a basic element of smart cities. The primary issue for the deployment of ITS based on IoV is the security for both users and infrastructures. The Intrusion Detection System (IDS) is important for IoV users to keep them away from various attacks via the malware and ensure the security of users and infrastructures. In this paper, we design a data-driven IDS by analyzing the link load behaviors of the Road Side Unit (RSU) in the IoV against various attacks leading to the irregular fluctuations of traffic flows. A deep learning architecture based on the Convolutional Neural Network (CNN) is designed to extract the features of link loads, and detect the intrusion aiming at RSUs. The proposed architecture is composed of a traditional CNN and a fundamental error term in view of the convergence of the backpropagation algorithm. Meanwhile, a theoretical analysis of the convergence is provided by the probabilistic representation for the proposed CNN-based deep architecture. We finally evaluate the accuracy of our method by way of implementing it over the testbed.