The complexity of information theoretic secure computation

Abstract

Summary form only given. A protocol for secure computation allows two or more parties to perform a distributed computation on their local inputs while hiding the inputs from each other. In the so-called “information theoretic” setting for secure computation, the parties are assumed to communicate over secure channels and the inputs should remain hidden even from computationally unbounded parties. It is known that every computation can done securely when there is a majority of honest parties, or alternatively when the parties are given access to certain types of correlated secret randomness. However, the true cost of such secure computations remains wide open. The talk will survey some recent progress and open questions in this area.