Keynote: The End of Anonymity, the Beginning of Privacy: New Directions in Privacy-Preserving Data Analysis

Vitaly Shmatikov (The University of Texas at Austin)

The new Web economy relies on the collection of personal data on an ever-increasing scale.  Information about our tastes, purchases, searches, browsing history, friendships and relationships, health history, genetics, and so forth is shared with advertisers, marketers, and researchers. The aggregated datasets do not exist in isolation; they contain implicit or explicit references to other datasets.  Unsurprisingly, this raises a number of interesting privacy issues.

I will survey several approaches to privacy-preserving data sharing and show that "anonymization," including popular methods based on k-anonymity and similar syntactic properties, fails to provide meaningful privacy guarantees.  I will discuss the subtle relationship between anonymity and privacy, explain several techniques for de-anonymizing large datasets, and present the experimental results which demonstrate how de-anonymization can be carried out on real-world datasets and social networks.

In the second part of the talk, I will describe the ongoing research on Airavat, a system for large-scale, privacy-preserving computation which is being developed at UT Austin.  Airavat is based on the MapReduce framework and includes a novel integration of mandatory access control and differential privacy.  It enables users without security or privacy expertise to carry out computations on sensitive data, while ensuring compliance with the data providers' privacy policies.

Vitaly Shmatikov is an associate professor at the University of Texas at Austin.  His research focuses on security, privacy, and formal verification methods for secure systems and protocols.  Vitaly was the recipient of the 2008 PET Award for Outstanding Research in Privacy Enhancing Technologies.  He will serve as co-chair of ACM CCS 2010 and 2011.  Vitaly received his PhD from Stanford University.

Invited Talk: Privacy in social networks

Elena Zheleva (University of Maryland)

Abstract: With the proliferation of online social networks, there has been a growing interest in understanding the mechanisms which govern complex network formation and evolution. On the flip side, statistical models, which allow learning hidden information in these networks, bring a lot of privacy concerns. In this talk, we will give a brief overview of the research in online social network privacy, and present recent work in which we have shown that in addition to friendship links, group memberships have a strong potential for leaking personal information.

Elena Zheleva is a Ph.D. student in Computer Science at the University of Maryland College Park working with Prof. Lise Getoor. Her research interests lie in understanding the practical and theoretical issues in preserving privacy in network data while allowing the discovery of interesting and useful patterns. She is also working on statistical modeling of the behavior of users in social and affiliation networks, as well as on developing practical classification algorithms in these types of networks.