Dude, Who’s Driving My Car? Detecting Familiar and Unfamiliar Drivers Using CAN Bus Data

Driver behavior modeling (DBM) plays a crucial role in improving vehicle security, fleet management, and road safety. This study explores the use of machine learning techniques to assess driver familiarity with a vehicle using raw Controller Area Network (CAN) bus data. Unlike traditional approaches focused on driver identification, this work aims to determine whether a driver is familiar or unfamiliar with a specific vehicle by analyzing driving patterns. The dataset includes CAN bus signals collected from three different vehicles driven by both familiar and unfamiliar drivers. Several machine learning models were evaluated, including classical methods such as Random Forest, Support Vector Machines (SVM), and k-Nearest Neighbors (KNN), as well as deep learning models like Long Short-Term Memory (LSTM) networks and Fully Connected Neural Networks (FNN). Additionally, the ROCKET algorithm was applied to analyze time-series data directly. Results indicate that Random Forest achieved the highest accuracy across all vehicles, demonstrating its effectiveness in distinguishing between familiar and unfamiliar drivers. The ROCKET model also performed well, highlighting its potential for real-world applications. The findings suggest that machine learning-based driver familiarity detection can enhance Advanced Driver Assistance Systems (ADAS), optimize fleet management, and prevent unauthorized vehicle use. Future research should focus on larger datasets, additional contextual factors, and improved model interpretability to ensure real-world applicability and scalability..