Quantifying Location Privacy: The Case of Sporadic Location Exposure

Mobile users expose their location to potentially untrusted entities by using location-based services. Based on the frequency of location exposure in these applications, we divide them into two main types: Continuous and Sporadic. These two location exposure types lead to different threats. For example, in the continuous case, the adversary can track users over time and space, whereas in the sporadic case, his focus is more on localizing users at certain points in time. We propose a systematic way to quantify users' location privacy by modeling both the location-based applications and the location-privacy preserving mechanisms (LPPMs), and by considering a well-defined adversary model. This framework enables us to customize the LPPMs to the employed location-based application, in order to provide higher location privacy for the users. In this paper, we formalize localization attacks for the case of sporadic location exposure, using Bayesian inference for Hidden Markov Processes. We also quantify user location privacy with respect to the adversaries with two different forms of background knowledge: Those who only know the geographical distribution of users over the considered regions, and those who also know how users move between the regions (i.e., their mobility pattern). Using the Location-Privacy Meter tool, we examine the effectiveness of the following techniques in increasing the expected error of the adversary in the localization attack: Location obfuscation and fake location injection mechanisms for anonymous traces.