A Machine Learning-Based Defensive Alerting System Against Reckless Driving in Vehicular Networks

Reckless driving severely threatens the safety of innocent people, which accounts for around 33% of all fatalities in major vehicle accidents. However, most existing efforts focus on the detection and adjustment of a vehicle's own driving behavior, whose effectiveness is very limited. In this paper, we develop a defensive alerting system to detect and notify the threats posed by reckless vehicles. By integrating the computation capability of a cloud server with that of vehicles nowadays, we propose a cooperative driving performance rating (CDPR) mechanism to automatically and intelligently detect reckless vehicles based on machine learning algorithms. To support such a defensive alerting system, a three-tier system architecture is developed from existing vehicular networks, which consists of vehicles, road-side units (RSU) and a cloud server. Moreover, given the fact that most vehicles can be trusted to drive safely, to further reduce the transmission load of the CDPR mechanism, we design our scheme in such a way that every vehicle only uploads the data of driving maneuvers with reckless potential to RSUs. Based on the aggregated data, the cloud server globally rates every vehicle's driving performance by using support vector machine (SVM) and decision-tree machine learning models. We finally implement the proposed CDPR mechanism into a popular traffic simulator, Simulation of Urban MObility (SUMO), for evaluation. Simulation results illustrate that our defensive alerting system can accurately detect reckless vehicles and effectively provide timely alerts.