Achieving perfect location privacy in Markov models using anonymization
Zarrin Montazeri, Amir Houmansadr, Hossein Pishro-Nik · International Symposium on Information Theory and its Applications · 2016
Our previous work [1] defined the notion of perfect location privacy for location-based services (LBS) and demonstrated how it can be achieved under basic user models. In this paper, we extend our work by studying perfect location privacy within a more complex scenario in which a user's movements are dependent upon the time domain. We model a mobile user's path using Markov chains and show that perfect location privacy is achievable for the user if her pseudonym is changed before o(n 2/|E|-r) observations by the adversary, where |E| is the number of edges in the Markov model and r is the number of all possible locations. We support our results with simulations.