Design And Simulation Of GNSS And CAN Anomaly Detection Algorithms For Intelligent Connected Vehicles

With the booming development of automobile industry,automobile intellectualization has become a trend that OEMs and suppliers pay more and more attention to.Excluding battery,motor and electric control technology,the level of intelligence shows the ability of differentiated competition of automobile brands.One of the major aspects of vehicle intelligence is the advanced assistance driving system,which can assist the driver to drive the vehicle better and may evolve into autonomous driving in the future.The advanced driver assistance system needs to use sensor data to schedule the controller LAN.Vehicles value safety more than ordinary consumer electronics.Only for advanced driver assistance system,sensor data and controller LAN security protection,is very important for the safety of driver assistance.This paper focuses on GNSS and CAN network anomaly detection,and use data analysis to complete the two links of vehicle safety.Traffic travel is closely related to everyone,vehicle safety is an important part of traffic safety,but also an important lifeline of the future development of the vehicle,the work is a part of the vehicle safety,has a certain application and practice value.Main contributions of this paper are as follows:· GNSS anomaly detection algorithm based on Gaussian mixture model An anomaly detection algorithm based on Gaussian mixture model is proposed.Firstly,k-means algorithm is used to simplify signal sequence and remove outliers.Then gaussian mixture model is used to model the signal sequence.Then use EM algorithm to train the model,get the corresponding parameters,put them in the corresponding regression expression,get the prediction result,and compare the predicted value with the actual value to judge whether the anomaly occurs.· GNSS anomaly detection algorithm based on multi-decision tree integration model A GNSS anomaly detection algorithm based on multi-decision tree integration model is proposed.PCA method is first used to reduce the dimension of vehicle data,and then Pearson coefficient is used for feature selection.Then,the multi-decision tree in-tegration model is constructed to determine the leaf splitting growth strategy,and the bayesian optimization model parameters are used to train the multi-decision tree integration model.The obtained model is used for prediction,and the predicted value is compared with the actual value to detect anomalies.· CAN anomaly detection algorithm based on stacked LSTM residual network A CAN anomaly detection algorithm based on stacked LSTM residual network is proposed.Firstly,the corresponding features are extracted from CAN messages,normalized and smooth processing is carried out,the data is converted into supervised learning data,the model parameters are optimized with genetic algorithm,the stacked LSTM residual network is trained,and the unknown data is predicted to be abnormal or not using the trained network.In this paper,for the anomaly detection of GNSS and CAN networks,which is a problem in the field of in-vehicle information security,three methods are proposed: GNSS anomaly detection algorithm based on Gaussian mixture model,GNSS anomaly detection algorithm based on multidecision tree integration model and CAN anomaly detection algorithm based on stacked LSTM residual network,and simulation experiments are conducted on the collected data sets,which achieve better results which have certain reference value for practical applications.