UCLA

Pervasive mobile devices have enabled countless context-and location-based applications that facilitate navigation, life-logging, and more. As we build the next generation of smart cities, it is important to leverage the rich sensing modalities that these numerous devices have to offer. This work demonstrates how mobile devices can be used to accurately track driving patterns based solely on pressure data collected from the devices barometer. Specifically, by correlating pressure time-series data against topographic elevation data and road maps for a given region, a centralized computer can estimate the likely paths through which individual users have driven, providing an exceptionally low-power method for measuring driving patterns of a given individual or for analyzing group behavior across multiple users. This work also brings to bear a more nefarious side effect of pressure-based path estimation: a mobile application can, without consent and without notifying the user, use pressure data to accurately detect an individuals driving behavior, compromising both user privacy and security. We further analyze the ability to predict driving trajectories in terms of the variance in barometer pressure and geographical elevation, demonstrating cases in which more than 80% of paths can be accurately predicted.