RAND Statistics Seminar Series

Secure Logistic Regression with Distributed Databases

Presented by Dr. Aleksandra Slavkovic, Pennsylvania State University
Thursday, October 25, 2007
1:30 P.M. EDT
Conference Room 6202
RAND Corporation, Pittsburgh, PA
Please contact Denise Miller if you would like to attend this seminar.

Abstract

This presentation provides a brief overview of statistical disclosure limitation methods and their link to privacy-preserving data mining techniques. It then specifically addresses the problem of performing a logistic regression in a secure way across partially overlapping data bases, owned by multiple parties. The idea is to have a shared computation for the purposes of model fitting and assessment using the pooled data, in such a fashion that no party's data is directly disclosed to any other party. Examples of parties are government agencies or competing business establishments. We describe the relationship of this problem to other well-formulated statistical ones, and discuss the computational details, the meaning of "secure" computation in this context, and the implications for the protection of privacy, both for data base owners and for the individuals whose data are incorporated into the calculations. This is joint work with Stephen Fienberg, Alan Karr, and Yuval Nardi.

Speaker Bio

Dr. Slavkovic's past and current research interests include usability evaluation methods, human performance in virtual environments, statistical data mining, application of statistics to social sciences, algebraic statistics, and statistical approaches to confidentiality and data disclosure. Her Ph.D. dissertation work focuses on statistical methodologies for disclosure limitation and data confidentiality and presents new theoretical links between disclosure limitation, statistical theory and computational algebraic geometry. Dr. Slavkovic served as a consultant to the National Academy of Sciences/National Research Council Committee to Review the Scientific Evidence of Polygraph in 2001 and part of 2002. In 2003, she received an honorable mention for the best student paper from the Committee on Statisticians in Defense and National Security of the American Statistical Association.



Attending a Seminar

Other Locations/Times: Washington, D.C. Conf. Rm. 4132: 1:30 p.m. ET
Santa Monica Conf. Rm. 1226 and 1228: 10:30 a.m. PT

Visitors are welcome to attend in Pittsburgh and Santa Monica and must RSVP at least one day prior to the seminar. To ensure your attendance please contact Denise Miller at dmiller@rand.org with your name, company (or university) affiliation, and national citizenship (for security purposes).

RAND Pittsburgh is located at 4570 Fifth Avenue, Suite 600, Pittsburgh, PA. For parking and directions please see: http://www.rand.org/about/locations/pittsburgh.html. Visitors should check in at the receptionist's desk on the 6th floor.

For parking and directions to RAND's Santa Monica office, please see: http://www.rand.org/about/locations/santa-monica.html.

For further information and to be added to the mailing list contact Denise Miller at dmiller@rand.org.