With the wide propagation of handheld devices, more and more mobile sensors are being used by end users on a daily basis. Those sensors could be leveraged to gather useful mobility data for city planners, business analysts and researches. However, gathering and exploiting mobility data raises many privacy threats. Sensitive information such as one’s home or workplace, hobbies, religious beliefs, political or sexual preferences can be inferred from the gathered data.
In the last decade, Location Privacy Protection Mechanisms (LPPMs) have been proposed to protect user data privacy.
They alter data mobility to enforce formal guarantees (e.g., k-anonymity or differential privacy), hide sensitive information (e.g., erase points of interests) or act as countermeasures for particular attacks.
In this thesis, we focus on the threat of re-identification which aims at re-linking an anonymous mobility trace to the know past mobility of its user.
First, we propose re-identification attacks (AP-Attack and ILL-Attack) that find vulnerabilities and stress current state-of-the-art LPPMs to quantify their effectiveness. We also propose a new protection mechanism HMC that uses heat maps to guide the transformation of mobility data to change the behaviour of a user, in order to make her look similar to someone else rather than her past self which preserves her from re-identification attacks. This alteration of mobility trace is constrained with the control of the utility of the data to minimize the distortion in the quality of the analysis realized on this data.