Andreas Haeberlen 

Department of Computer Information Science

"How to have your cake and eat it too"

 

Abstract

A wealth of data is constantly accumulating in various databases - patient records, link graphs of social networks, mobility traces in cellular networks, etc. - and there are many good uses to which it can potentially be put. However, much of this data contains private information about individuals, and revealing that information would be an enormous risk. Hence, it appears that we must make a choice: we can give up privacy, and gain many interesting opportunities, or we can protect privacy and avoid the risks. In other words, we can either have our cake, or we can eat it, but not both.

In this talk, I will show that it is not always necessary to make this choice. I will discuss two example scenarios. In the first scenario, I will show how to verify whether an Internet network is keeping a promise it has made about its routing behavior, without revealing what that behavior actually is. In the second scenario, I will show how we can join information from multiple databases (say, a travel agent's reservation database and a hospital's medical records) to answer questions about the combination (say, how many patients went to region X and were then treated for disease Y) while still protecting the privacy of each individual. In both scenarios, our solutions can give strong, provable privacy guarantees.

 

Refreshments will be served on the

2nd Floor Mezzanine Level

outside Wu & Chen

immediately following the talk.

____________________________________________________________________________