TOC Colloquium featuring:


Martin J. Strauss

University of Michigan

Secure Multiparty Computation of Approximations: A Survey

Tuesday, April 29, 4:15pm

32-155

[refreshments at 3:45pm in 32-G5]




Abstract

Suppose two or more parties have pieces of a large database and they want to do joint datamining. Typically, they will want their computation to be efficient, at least approximately correct (often exact computations cannot be performed efficiently), and private, in the sense that the players do not want to reveal more information than necessary.

Although approximation algorithms and private multiparty computation (as well as a stronger notion of secure multiparty computation) have been studied for decades, it is not straightforward to combine existing techniques. We discuss the issues, give a general framework for secure approximations, and survey some results in the area.


Host: Piotr Indyk