With the popularization of intelligent connected cars, vehicle data communication becomes increasingly frequent, and controller area network bus anomaly detection becomes more and more important. The study presents the gradient boosting decision tree algorithm as a solution to the low accuracy of controller area network bus anomaly detection. For the problem that its parameters are difficult to determine, an improved genetic algorithm with adaptive crossover probability and mutation probability is used to optimize it. At the same time, principal component analysis method is used to reduce its dimensionality. Finally, a gradient boosting decision tree algorithm optimized by an improved genetic algorithm is combined with the principal component analysis method to propose a vehicle controller area network bus anomaly detection model. The proposed models are compared and analyzed. The results found that the accuracy, precision, F1 value, and average error detection rate of this model were 98.23%, 97.82%, 98.62%, and 2.51% respectively. In the application effect experiment, the detection accuracy of the proposed model for four attack types was 98.9%, 99.3%, 98.4%, and 98.6%. In conclusion, the detection model proposed in the study is effective and practical, and can provide a theoretical basis for anomaly detection in smart car networks.
Zhao et al. (Fri,) studied this question.