Goal-Oriented Privacy Preservation

This is a newly funded NSF Cybertrust project, carried out with David DeWitt and Jude Shavlik in CS, Larry Hanrahan (Chief Epidemiologist, State of Wisconsin), and Amy Trentham-Dietz (Population Health Sciences). Sophisticated data analysis and mining techniques are necessary tools for understanding complex datasets in an increasing number of application domains, including bioinformatics, health care, e-commerce, fraud detection, network attacks, and homeland security. The increasing use of these techniques has also created a heightened awareness of their potential for compromising privacy, and the challenge of controlling data access in accordance with privacy policies while enabling useful analysis has emerged as a central and ubiquitous problem. To meet this challenge, we need to go well beyond traditional access-control strategies because analysis and mining techniques can potentially be applied to published data to infer data that is intended to be private. We are investigating the trade-off between privacy and security guarantees and the utility of the published data for specific analysis tasks; we call this goal-oriented privacy preservation.

Funding