Secure multi-party computation protocols often offload most cryptographically and computationally intensive components to an offline procedure. The objective of this offline procedure is to output secure samples from highly structured correlated randomness, for example, Beaver triples. The offline procedure relies on public-key cryptography to achieve this objective and, consequently, is computation and communication intensive. On the other hand, there are diverse inexpensive-to-generate correlated randomness, like joint samples from noise sources, that can also facilitate secure computation via interactive protocols.

A natural approach to increase the efficiency of this offline phase will be to non-interactively and securely transform such correlated randomness into correlations useful for secure computation while incurring low computational overhead. Motivated by this general application for secure computation, we introduce the notion of secure non-interactive simulation (SNIS) – the information-theoretic analog of pseudo-correlation generators, and a cryptographic extension of non-interactive simulation and non-interactive correlation distillation in information theory.

This talk lays the theoretical foundations for SNIS research and, in this context, discusses tight feasibility, rate, and capacity results for some fundamental correlations. Our technical contributions establish deep connections with harmonic analysis, error-correcting codes, and information theory and motivate fundamental new research in these fields.

===========================

https://umd.zoom.us/j/97585901703

Meeting ID: 975 8590 1703