How to compute group statistics while preserving individual privacy
Speaker: Eleanor Rieffel , Senior Research Scientist, FXPAL
Date: November 4 2011
Time: 10:30AM to 12:00PM
Location: 32-G449 Patil/Kiva
Host: Ron Rivest, CSAIL, MIT
Contact: Be Blackburn, 3-6098, be@csail.mit.edu
Relevant URL: As sensors become ever more prevalent, more and more information will be collected about each of us. A long-term research question is how best to support beneficial uses while preserving individual privacy. Aggregate statistics are sufficient for many beneficial uses in public health and medical research, smart metering, and user analysis, and other areas. W
We introduce "aggregator oblivious" protocols for time series data in which an individual has full access to her own data, a third party processes the data without learning anything about the data values, and analysts learn only statistical information over the whole group at each time step. We describe two families of non-interactive aggregator oblivious protocols. We also show how these schemes can be combined with differential privacy mechanisms to obtain stronger privacy guarantees, and how encryption makes it possible to obtain a given level of differential privacy with less noise.
Joint work with: Jacob Biehl, Hubert Chan, Richard Chow, Adam J. Lee, Elaine Shi, Dawn Song, William van Melle
See other events that are part of CIS Seminars 2011/2012
See other events happening in November 2011
