Optimising driver profiling through behaviour modelling of in-car sensor and global positioning system data. (May 2021)

• Record Type: Journal Article
• Authors: Ahmadi-Assalemi, Gabriela; al-Khateeb, Haider M.; Maple, Carsten; Epiphaniou, Gregory; Hammoudeh, Mohammad; Jahankhani, Hamid; Pillai, Prashant
• Is Part Of: Computers & electrical engineering. Volume 91(2021)
• Publication Date: 2021-05
• Subjects: Driver identification -- Behaviour profiling -- Classification -- Machine learning -- Connected cars -- Random forest -- GPS -- Cybersecurity threat -- Incident response; Computer engineering -- Periodicals; Electrical engineering -- Periodicals; Electrical engineering -- Data processing -- Periodicals; Ordinateurs -- Conception et construction -- Périodiques; Électrotechnique -- Périodiques; Électrotechnique -- Informatique -- Périodiques; Computer engineering; Electrical engineering; Electrical engineering -- Data processing; Periodicals; Electronic journals; 621.302854
• DOI: 10.1016/j.compeleceng.2021.107047
• Languages: English
• ISSNs: 0045-7906

Abstract

Highlights: Successful classification based on a short segment of driving data. Driver gender was identifiable through profiling driver groups. Downloadable dataset built from actual (non-simulated) experiment. Feature selection optimised to increase the accuracy of the result. Piloting lessons learned shared to facilitate future experiment design. Abstract: Connected cars have a massive impact on the automotive sector, and whilst this catalyst and disruptor technology introduce threats, it brings opportunities to address existing vehicle-related crimes such as carjacking. Connected cars are fitted with sensors, and capable of sophisticated computational processing which can be used to model and differentiate drivers as means of layered security. We generate a dataset collecting 14 h of driving in the city of London. The route was 8.1 miles long and included various road conditions such as roundabouts, traffic lights, and several speed zones. We identify and rank the features from the driving segments, classify our sample using Random Forest, and optimise the learning-based model with 98.84% accuracy (95% confidence) given a small 10 s driving window size. Differences in driving patterns were uncovered to distinguish between female and male drivers especially through variations in longitudinal acceleration, driving speed, torque and revolutions per minute.