Preserving privacy when sharing sensitive semantic locations

With recent mobile technological advances, people location can be tracked both indoor and outdoor spaces with a spatial accuracy of a few meters. Along with this advance, location based services usage has increased remarkably and location sharing through these services brought versatile privacy issues. From the shared coordinates one can easily infer the respective semantic location such as hospital, nigh club and restaurant. This thesis investigates two privacy problems concerning the sharing of semantic sensitive locations. In the former, user identity is assumed to be known by service provider but the semantic location is asked to be obfuscated with respect to user privacy profile specification under road network and velocity based-attack constraints. Algorithms solving the problem in a few settings have been developed and extensive experimental evaluations are performed. Unlike the former where privacy is service-centric issue, the latter problem addresses data-centric privacy where collected semantic location trajectory database needs to be published in a privacy-preserving manner. The problem asks providing anonymity given semantic location traces of users. The solution approach uses generalization and order relaxation techniques to achieve anonymity.