Driver Identification with OBD-II Public Data

Abstract

The problem of driver identification is an urgent one, and many publications propose various methods to solve it. Most of these papers use sensor data from the car's CAN-Bus to get a driver's fingerprint. However, at the moment, not all the sensors mentioned in these works are available without knowing the protocol details of a particular vehicle model, and this fact severely limits practical use of the existing methods in many applications. This paper explores the possibility of identifying the driver using information only from sensors listed in SAE J1979 specification. It turns out, that accuracy decreases by 15% compared to methods, in which the driver is identified on the full set of sensors (including the ones specific for a particular car), and is 79% versus 91% on the full set of sensors.

Keywords

• driver identification
• can bus
• obd-ii
• machine learning