11:30 am - 12:30 pm, Wednesday, December 02, 2009
Prof. Geetha Jagannathan
Title: Practical Differential Privacy
Abstract: Recent work in private data analysis by Dwork et al. has radically changed the research landscape by defining a model with a strong definition of privacy that addresses how much privacy loss an individual might incur by being in the database. In the most common setting under this new framework of differential privacy, the data owner makes the data available through a statistical database on which only aggregate queries are permitted. The goal is to answer queries while preserving the privacy of every individual in the database, irrespective of any auxiliary information that may be available to the database client. Using existing differential privacy results, it is possible to create high-level structures such as decision trees using multiple low-level differentially private queries. However, for many databases and high-level structures, acceptable levels of privacy in the end result via these methods can only be obtained by sacrificing utility in the high-level structure. In this talk, I will address the problem of constructing private classifiers using decision trees, within the framework of differential privacy preserving privacy without compromising utility.
Prof. Geetha Jagannathan has a Ph.D. in Mathematics from the Indian Institute of Technology, Madras. She is currently wrapping up her Ph.D. in Computer Science from Rutgers University. Prior to this she was a senior developer at RightFreight, Inc., a startup that created logistic systems. Her previous experience includes assistant professorships at SV College of Enginnering and Indian Institute of Technology, Madras, and a stint as a post doctoral researcher in Physics at Hofstra. Her research interests are in cryptography and privacy. For more information about this colloquium, please contact Habib M. Ammari at firstname.lastname@example.org