Increasing numbers of vehicles are equipped with telematics devices and they are able to transmit vehicle CAN bus information remotely. We examine the possibility of identifying individual drivers from their on-road driving behaviors. This study collected vehicle telematics data from a small fleet of Ford Fiesta vehicles over 6 months in London, UK. The collected variables included vehicle speed, acceleration pedal position, brake pedal pressure, steering angle position, gear position, and engine RPM. A list of driving metrics were developed to quantify driver behaviors, such as mean brake pedal pressure and turning speed. The Random Forest (RF) machine learning algorithm is used to predict driver IDs based on the developed driving metrics. The RF model is also used to rank the importance of each driving metric on driver identification. In conclusion, this paper demonstrates the possibility of identifying drivers from their on-road driving behaviors.