Research on Vehicular External Network Intrusion Detection System Based on Ensemble Learning

With the development of the Internet of Vehicles(IoV), the combination of networks and vehicles has brought great convenience to people. However, hackers can attack vehicles through technical loopholes in the external network, which seriously affects the driving safety of vehicles. In this paper, we propose a new intrusion detection system (IDS) framework named the Leading Model and Confidence Decision Method (LMCDM). The LMCDM is built on three distinct machine learning models(Extreme Gradient Boosting (XGBoost), random forest (RF), and decision tree (DT)) to select the model exhibiting the best identification performance across attack types. Furthermore, the LMCDM combines prediction confidence to facilitate precise predictions. Finally, the LMCDM model was assessed using the CICIDS2017 dataset, and relative to other IDS systems, F1 scores demonstrated an improvement of no less than 0.88%. The experimental results show the effectiveness of LMCDM in external network intrusion detection.

I. Introduction

Internet of Vehicles (IoV) refers to the communication between vehicles using onboard devices as a medium, including Vehicle to Vehicle (V2V), Vehicle to Road (V2R), and Vehicle to Infrastructure (V2I) [1]. In IoV, vehicles transmit various parameter data, such as vehicle speed, angle, message timestamp, and actual or simulated vehicle identity. These parameter data can be received by the server to assist in autonomous driving and improve road safety and efficiency.