Misbehavior Detection based on Deep Learning for VANETs
Xiangyu Liu
Abstract
With the development of intelligent driving technology, the reliability of inter-vehicle communication has been put on the agenda. The safety and efficiency of vehicular communication depend on the correctness of data exchange among vehicles. As the Internet of Things (IoT) is increasingly used in Vehicular Ad Hoc Networks (VANETs), the number of connected vehicles will grow exponentially in the coming years, meaning the number of communication interfaces will increase and so there will be the potential for cyber security attacks. Therefore, it is necessary to add anomaly detection schemes to the existing security schemes to distinguish normal vehicle data from malicious and erroneous data. We use the VeReMi data set to detect the abnormal behavior of common location errors on roads and classify the types of location anomalies. The deep learning neural network model and LSTM model, on the basis of traditional machine learning, can make us once find improper behavior of communication and is possible to categorize it. We can use more accurate and specific steps to deal with these unusual projects to improve the safety of the vehicle system and the ability to quickly recover from the abnormal behavior.