Microaggregation is a statistical disclosure control (SDC) algorithm widely used by national statistical agencies to enforce k-anonymity on numerical datasets. In this seminar, we will introduce the general SDC scenarios and their main disclosure risks. Later, we will show that microaggregation algorithms are flexible enough to cover most of those disclosure risks. Afterwards, we will present some performance tricks used in kNN literature, to speed up microaggregation for very large datasets. Finally, we will present some results concerning classifiers performance on microaggregated datasets.

Bio: Jordi Nin holds a Ph.D. (2008) in Computer Science from the Autonomous University of Barcelona (UAB). He works as a tenure-track Lecturer in the Technical University of Catalonia (UPC). Previously, from 2009 to 2010 he was a Marie Curie postdoc researcher in the LAAS-CNRS in Toulouse, France. From 2007 to 2009 he worked as pre-doctoral researcher at the IIIA-CSIC. His fields of interests are privacy technologies and machine learning. He has been involved in several research projects funded by the Catalan and Spanish governments and the European Community. His research has been published in specialized journals and major conferences (more than 50 papers accepted).