I am a a lecturer (eq. Assistant Professor) at Imperial College London's new Data Science Institute (joint with the Department of Computing). I was previously a postdoctoral researcher at Harvard IQSS working with Prof. Latanya Sweeney and Prof. Gary King and I received my PhD from MIT under the supervision of Prof. Alex "Sandy" Pentland (250-words bio).
I am currently looking for talented PhD students to join our computational privacy research group at Imperial College.
My research aims at understanding how the unicity of human behavior impacts the privacy of individuals in large-scale metadata datasets. My work has been covered in The New York Times, BBC News, CNN, Wall Street Journal, Harvard Business Review, Le Monde, Die Spiegel, Die Zeit, El Pais, and in reports of the World Economic Forum, United Nations, OECD, FTC, and the European Commission, as well as in my talks at TEDxLLN and TEDxULg. I recently wrote a white paper for Brookings on the use and privacy metadata as well as op-eds for the World Economic Forum, Christian Science Monitor, and Le Monde. I worked for the Boston Consulting Group and acted as an expert for the Bill and Melinda Gates Foundation and the United Nations. I was recently named an Innovator under 35 for Belgium (TR35). I am a fellow of the ID³ Foundation, the B.A.E.F. Foundation, and a research associate at Data-Pop. I am organizing NetMob, the International Conference on the Analysis of Mobile Phone Datasets, and I am serving on numerous program committee.
I approach the privacy of metadata; mobile phone, credit cards, or browsing metadata; from two perspectives. First, I developed the concept of unicity to study the risks of re-identification of large-scale metadata datasets. I showed that in 4 spatio-temporal points are enough to uniquely identify 95% of people in a mobile phone database of 1.5M people and to identify 90% of people in a credit card database of 1M people. I furthermore showed that, in both cases, even coarse or blurred datasets provide little anonymity. Second, I use machine learning techniques to study what can be inferred from metadata about individuals. For example, using behavioral indicators computed from metadata using the Bandicoot toolbox, we were able to predict people's personality up to 1.7x better than random. Ultimately, however, I am convinced of the amazing potential of this data for good, allowing us to better understand human behavior and their societies at scale. The goal of my research is to help make metadata more available but in a privacy-conscientious way through online systems such as openPDS/SafeAnwers or conscientious anonymization such as in the D4D Challenge with Orange.