(IJACSA) International Journal of Advanced Computer Science and Applications, Vol. 13, No. 12, 2022 566 | Page www.ijacsa.thesai.org A Novel Compound Feature based Driver Identification Md. Abbas Ali Khan 1 , Mohammad Hanif Ali 2 , AKM Fazlul Haque 3 , Md. Iktidar Islam 4 , Mohammad Monirul Islam 5 Computer Science and Engineering, Daffodil International University, Dhaka, Bangladesh 1, 3, 4, 5 Computer Science and Engineering, Jahangirnagar University, Dhaka, Bangladesh 1, 2 Abstract—In today's world, it is time to identify the driver through technology. At present, it is possible to find out the driving style of the drivers from every car through controller area network (CAN-BUS) sensor data which was not possible through the conventional car. Many researchers did their work and their main purpose was to find out the driver driving style from end-to-end analysis of CAN-BUS sensor data. So, it is potential to identify each driver individually based on the driver's driving style. We propose a novel compound feature- based driver identification to reduce the number of input attributes based on some mathematical operation. Now, the role of machine learning in the field of any type of data analysis is incomparable and significant. The state-of-the-art algorithms have been applied in different fields. Occasionally these are tested in a similar domain. As a result, we have used some prominent algorithms of machine learning, which show different results in the field of aspiration of the model. The other goal of this study is to compare the conspicuous classification algorithms in the index of performance metrics in driver behavior identification. Hence, we compare the performance of SVM, Naïve Bayes, Logistic Regression, k-NN, Random Forest, Decision tree, Gradient boosting. Keywords—Compound feature; driver behavior identification; engine speed; fuel consumption; vehicle I. INTRODUCTION Every driver has their driving style; therefore, the driver can be identified according to exploration through the driving pattern analysis. It is to be considered as a fingerprint of the driver's manner like acceleration, speed, and braking habits that vary from driver to driver. Driver fingerprinting could lead to important privacy compromises [1]. Today we cannot consider just a vehicle as a modern car, as it is a fully decorated smart device with various functions like multimedia, security system, and different sensors [2]. The sensors were very simple because the driver was informed regarding the features of the engine and the amount of fuel through the magnetoelectric and light display devices [2]. Using state-of-the-art technology in real-time all the microcomputers are communicated with each other through CAN-BUS (Controller Area Network) [3]. To make a car more efficient a good number of technologies are used in the modern engine. To improve engine performance direct injection technology was introduced in the modern car [4]. According to a survey, the researcher predicted that the number of sales of connected cars will reach 76.3 million in the next 2023 [5]. Through state-of-the-art technology, modern engines use less fuel and besides get more power [6]. Most of the cars have partnered with other components which are highly technology-based, such as traffic lights, garage doors, and services [7]. Not only the driving style there is a discount policy on insurance services but also real-time monitoring, maintenance, pathfinding, driving style development, and also consumption of fuel [8]. Vulnerabilities of connected cars will increase the auto-theft which is one of the threats [9]. Top-of-the-range vehicles are targeted by thieves who simply drive off after bypassing security devices by hacking onboard computers [10]. Penny [11] introduced a man-in-the-middle attack or relay attack, to do this radio signals are passed between two devices. Pekaric I et al. (2021) [12] described other attacks such as GPS spoofing and message injection attacks. BMW Connected Drive [13] seamlessly integrates mobile devices, smart home technology, and vehicle's intelligent interfaces into a complete driver's environment. Even though in 2021 they introduced a remote door unlock system through a signal to the driver’s door to unlock [13]. CAN-BUS is likely a nervous system used to allow configuration, data logging, and communication among electronic control units (ECU) e.g., ECU is like a part of the body and interconnected through CAN, by which information sensed by one part can be shared with another [14]. Up to 70 ECUs have a modern car e.g., the engine control unit, airbags, audio system, acceleration, fuel unit, etc. [15]. Normally, multi-sensor data is made up of in vehicle’s CAN data. The in- vehicle CAN data such as steering wheel, vehicle speed, engine speed, amount of fuel, etc. Several researchers previously proposed a driver identification method based on in-vehicle CAN-BUS data. But direct connectivity is difficult to get data, so onboard diagnostics (ODB-II) is used. (OBD-II, ISO 15765) is a self-diagnostic and reporting capability that e.g., mechanics use to identify car issues, OBD-II specifies diagnostic trouble codes (DTCs) and real-time data (e.g., speed, revolution per minute RPM), which can be recorded via OBD-II loggers from CAN-BUS. Many authors described the problem of CAN-BUS data for identifying the driver [9], [16], [17]. Since this is a big dataset and there are 51 features with 10 labels. Moreover, for analyzing the whole dataset we need more time. To reduce the time complexity we have explained compound feature selection process. In this paper, our objective is to identify driver behavior through telemetric data using machine learning algorithms. We analyze the data in terms of training, testing, and validation to get model accuracy that helps us with driver identification.