Explaining Machine Learning Based Speed Anomaly Detection System Using eXplainable Artificial Intelligence

The goal of combating road accidents through the application of advanced communication and network technologies in transportation system called intelligent transportation system (ITS) has recorded a remarkable success. Nevertheless, the complex nature of the ITS makes cyberattacks a major concern even with cryptographic techniques employed for authentication of users, as they cannot prevent attacks from authenticated users (insider attackers). To solve this problem, several data-centric approaches for anomaly detection using machine learning (ML) are proposed by the research community. However, those approaches are limited in interpretability and explainability due to the complexity of some ML algorithms and black-box nature of neural networks which makes threat intelligence and analysis difficult. In this paper, we proposed an eXplainable artificial intelligence (XAI)-based speed anomaly detection system using trusted vehicles method with more than 99% accuracy, precision, recall and F1-Score where Shapley Additive exPlanation (SHAP) is further employed to explain the predictions made by the models.