Hybrid BPSO-XGBoost Framework for Anomaly Detection in Connected and Automated Vehicles

Abstract

This paper introduces a hybrid compact model BPSO-XGBoost for the detection of anomalies in connected and automated vehicles (CAVs). Current deep learning approaches based on DSRC do not scale well, are susceptible to insider attacks, and are not capable of generalizing to high-frequency anomalies. To tackle the above challenges, we combine feature selection to improve performance, and the fast and powerful classifier, XGBoost. When tested on the perturbed SPMD data set, we can achieve 98% precision, 98% sensitivity, 97% precision and an F1 score of 0.98 - outperforming that of the CNN-LSTM by 6.52% in sensitivity and 8.99% in accuracy - with the possibility of operating in real time.