In-Device Spatial Cloaking for Mobile User Privacy Assisted by the Cloud

Abstract

Spatial cloaking has been proposed and studied to protect mobile user privacy when using location based services (LBS). Traditional spatial cloaking methods are carried out by a trusted proxy known as location trusted server (LTS) to generate a region that contains at least k users for every request. The LTS is assumed to know the location of all users at all times, and perform the cloaking for all user requests. There are a number of disadvantages of relying on a single service for privacy preservation, including the scalability concern and the appropriate worry that this service “knows too much”. To ameliorate this single-service problem, in-device spatial cloaking may be more desirable. However, the sticky problem is that the device does not know, at the time of the request, the locations of all other users, which are necessary to obtain an appropriate cloaked region. With cloud services, it may be appropriate to assume that user-density information is available from cloud servers. These servers may collect user location information for different regions, or may use sophisticated method to estimate user densities for different places. When a request needs to be anonymized, the device goes to the cloud to acquire appropriate user density information to perform spatial cloaking. In this operating environment, traditional spatial cloaking methods such as Casper [9] need to be modified in order to guarantee safety. This paper proposes and studies a new algorithm and reports performance evaluation of the new algorithm and its optimized version, aiming at provably safe cloaking with minimized communication cost. Experimental results show that the new algorithms work well in the realistic evaluation environment.