Research Challenges in Privacy-Aware Mobility Data Analysis and in Text Mining with Enriched Data

I propose a research program combining two areas in which I have worked in the last years, i.e. Mobility Data Analysis and Privacy-Preservation Techniques. Mobility data is the data created by moving devices (e.g. cellphones, GPS, wifi) registering their presence, timestamp (and, for GPS enabled devices, their position) with antennas, receivers and routers. Mobility data is ubiquitous and its volume is growing constantly. Its importance for understanding human and animal behaviour is crucial, and therefore there is general interest in collecting and exploring this type of data for a vast range of applications, ranging from traffic and transportation, ecology, epidemiology, to safety and security. The fundamental mobility data concept is a trajectory - a sequence of points where each point consists of a geospatial coordinate set and a time stamp.

The main goal of the proposed research program is to develop Machine Learning methods for the analysis of human mobility at both coarse and fine granularity, making them privacy-preserving whenever this data represents - or can identify - individuals, or breach other confidential information. While it is well known that human mobility data presents enormous privacy challenges, I show that the same applies for ship movements, particularly for smaller recreational and fishing vessels. I list specific research tasks that collectively will provide tools for addressing mobility data in a private manner. These tasks also make realistic and interesting topics of graduate theses for students working with me. Those tasks are: dividing trajectories into semantically meaningful parts (segmentation), prediction of the next point in a trajectory (next move prediction), segment classification, clustering of trajectories and use of clustering as a privacy-oriented data representation, detection of anomalous trajectories, linking and integration of extraneous data with mobility data, and privacy models conducive to the special characteristics of mobility data.

Exploring partnerships of my labs with companies that collect and own large mobility datasets, I will focus on two main types of data: ships tracks on world's oceans available through a GPS-like AIS (Automatic Identification System) platform, and people's traces left with wifi hotspots in an urban environment. I argue that this research will have significant impact. For instance, clustering urban mobility data by speed would identify spatio-temporal cycling patterns and inform the city about the times and routes with the highest likelihood of collisions between cyclists and motorists, enabling solutions (e.g. cyclist-only lanes) at specific times of the day and the year.