We are looking for an excellent candidate for a Master’s project in trajectory anonymisation. The pervasive use of mobile technologies, GPS-equipped vehicles, etc., has resulted in the availability of massive amounts of trajectory data. Trajectory databases offer a vast application potential for researchers, enterprises, public administrations, among others. For example, transport authorities can use trajectory databases for designing better public transportation networks, optimizing routes and resource consumption, which in turn benefits the environment by reducing pollution and improves the quality of life of the population by reducing traffic jams, commute times, etc. Despite their usefulness, the use of these databases raise important privacy concerns due to the sensitive nature of location information. For example, most persons will probably have no objection with others knowing that they pass by the central train station on their every-day commutes, but most would prefer not to disclose information regarding whether, and how often, they visit some place of worship, or health institution, etc.
In view of these concerns, it is important to properly sanitise trajectory data before its use. Numerous anonymisation techniques have been proposed for trajectory data. Within this project, we will focus on techniques based on the notion of k-anonymity. A dataset satisfying k-anonymity guarantees that every record is indistinguishable from at least other k - 1 different records, ensuring that the probability of an entry being uniquely identifiable is at most 1/k. We will focus on techniques based on clustering trajectories by similarity, aligning sets of points along clustered trajectories, and applying micro-aggregation operators on these sets of points. In particular, we will explore two main issues. First, we will assess the applicability of the so-called Fréchet distance for comparing and aligning trajectories. Secondly, we will test the hypothesis that the Weber location, a popular tool for facility location planning in Operations Research, is a good aggregation operator, in opposition to the more widely used centroid operator.