Abstract
In this talk, we will look at secure multiparty computation from the point of view of its practical effectiveness, and detail the secure computation paradigm in the correlated randomness model. In this paradigm, participants have access to a secure source producing long correlated random strings. This approach reduces the problem of efficient secure computation to two sub-questions : how can the use of long correlated random strings speed up secure computation protocols ? How can long correlated random chains be generated efficiently and securely? We will first illustrate the answer to the first question using a Beaver protocol, and show how the use of unconscious transfer extension allows us, starting from a simple pseudo-random generator, to answer the second question in the context of Boolean circuit computation with passive security. At the end of the talk, we'll show how the recent notion of pseudo-random correlation generators makes it possible to achieve this extension of unconscious transfers with minimal communication.
