Privacy-preserving data mining is an emerging field that studies how data mining algorithms can affect the privacy of data, and tries to find and analyze new algorithms that will preserve this privacy. In this talk, I will present an approach that allows two or more participants to construct a boosting classifier without explicitly sharing their data sets. Although they are willing to collaborate for the accomplishment of this task of mutual benefit, the participants wish at the same time to preserve the privacy of their data. For this purpose, I will describe BiBoost (Bipartite Boosting) and MultBoost (Multiparty Boosting), two distributed privacy-preserving algorithms that inherit the excellent generalization performance of AdaBoost.

This is joint work with Esma Aïmeur and Balázs Kégl, and Gilles Brassard for an earlier version of the bipartite algorithm.