An unsupervised approach for automotive driver identification Poster Abstract

The adoption of on-vehicle monitoring devices allows different entities to gather valuable data about driving styles, which can be further used to infer a variety of information for different purposes, such as fraud detection and driver profiling. In this paper, we focus on the identification of the number of people usually driving the same vehicle, proposing a data analytic work-flow specifically designed to address this problem. Our approach is based on unsupervised learning algorithms working on non-invasive data gathered from a specialized embedded device. In addition, we present a preliminary evaluation of our approach, showing promising driver identification capabilities and a limited computational effort.