A Secure Machine Learning-Based Framework for Aggressive Driving Behavior Analysis in IoV System

The increase in the number of road accidents is one of the major global concerns nowadays. The aggressive and abnormal driving behavior of the drivers is the primary cause of this issue. Such behavior can be observed frequently in urban traffic environments, leading to severe consequences including injuries, fatalities, and property damage. Therefore, it is very important to have a certain intelligent mechanism that can monitor driving behavior in real time and take appropriate action if it is found to be aggressive or abnormal. This can be possible by integrating the Internet of Things (IoT), digital twins, and machine learning technology. In this paper, we have proposed a secure machine learning-based framework for the detection of aggressive driving behavior in Internet of Vehicles (IoV) systems. It is assumed that smart IoT computing devices, such as sensors and cameras, are installed on roads and in moving vehicles to capture data such as sudden acceleration, harsh braking, sharp turns, tailgating, etc. This data is sent to the cloud servers in a secure manner, incorporating mutual authentication and key establishment between communicating entities. Machine learning models that are deployed on cloud servers are trained and tested against such data to analyze driving patterns. The machine learning classifiers that are utilized in the proposed work are Random Forest, K-nearest neighbor, AdaBoost, and Multi-layer perceptron (MLP) classifier, where Random Forest has achieved the maximum accuracy and F1-score. Finally, the proposed scheme is also compared with other existing schemes based on different parameters and outperforms those schemes.