Beyond the thin client model for location privacy

Location privacy in emerging location-based applications has been the center of extensive research in the past decade. While a number of efficient models and algorithms have been proposed over the years, majority of them were designed assuming a thin client model of computing. As a result, the dependence on third party systems, or the requirement for significant upgrades to the application architecture, could not be eliminated. The state-of-the-art in current mobile devices is now comparable to traditional desktop systems five years ago, presenting us the opportunity to move beyond the thin client model. Motivated by this observation, we propose a novel architecture for performing location-based points-of-interest search, where the client device can locally determine the top results using a small amount of metadata from the server. Precise location data of the user is never transmitted outside the device. We demonstrate that the computational power required to efficiently execute the algorithms is within the capabilities of current mobile devices.