DIFFERENTIAL PRIVACY AND TRAJECTORIES

Abstract

Publishing meaningful datasets that don't jeopardize the privacy of the partecipants is still a great challenge for the database community. It has been demonstrated that the recent differentially private approach guarantees a good privacy protection mantaining accurate query results. With the increasing popularity of devices which generate spatio temporal data, trajectories related to human movements has been collected and stored. Trajectory data are extremely useful for data mining tasks but they can expose the partecipants of the datasets to privacy breaches. This thesis aims to offer a differentially private protection to trajectory datasets, in order to allow their safe publication for data mining tasks. A survey of the most recent researches on the topic is presented, followed by the description of potential solutions to the problem.