Name : Adam Smith

Institution : Pennsylvania State University


Adam Smith is an assistant professor in the Department of Computer Science and Engineering at Penn State University. His research interests lie in cryptography, privacy and their connections to information theory, quantum computing and statistics. He received his Ph.D. from MIT in 2004 and was subsequently a visiting scholar at the Weizmann Institute of Science and UCLA.



Publications :


        S. R. Ganta, S. P. Kasiviswanathan and A. Smith. Composition Attacks and Auxiliary Information in Data Privacy. In 14th ACM International Conference on Knowledge Discovery and Data Mining (KDD), August 2008.

        S. P. Kasiviswanathan, H. Lee, K. Nissim, S. Raskhodnikova and A. Smith. What Can We Learn Privately? In 48th Annual Symposium on Foundations of Computer Science (FOCS), October 2008.



Title of Project : Rigorous Foundations for Privacy in Statistical Databases


The ubiquity of collections of personal and sensitive data (census surveys, online social networks, and public health data, to name a few) has created a host of new problems stemming from conflicts between data access and privacy. An important challenge for these collections is to discover and release global characteristics of the database without compromising the privacy of the individuals whose data they contain. The problem has been studied extensively in such diverse fields as statistics, databases and data mining. However, the approaches proposed in the literature, until very recently, had either no formal privacy guarantees or ensured security only against limited types of attacks. This project seeks to lay a firm conceptual foundation for the field of privacy in statistical databases, taking into account realistic, sophisticated adversarial attacks and bringing together ideas from several different sub-disciplines of statistics and computer science.


The research is centered around three themes:

1.      Formulating realistic models and definitions of privacy that provide resistance against strong, even active, attacks;


2.      Understanding the types of information that can, and cannot, be revealed while retaining privacy according to the definitions discussed above;


3.      Investigating techniques which "break" anonymization protocols, in order to inform protocol design in the same way that cryptanalysis informs modern cryptography.


The research is closely tied to questions of resilience, stability and robustness in machine learning and statistics, and raises fundamental new questions in these areas.