Automatic Proofs of Privacy of Secure Multi-party Computation Protocols against Active Adversaries

Abstract

We describe an automatic analysis to check secure multi-party computation protocols against privacy leaks. The analysis is sound - a protocol that is deemed private does not leak anything about its private inputs, even if active attacks are performed against it. Privacy against active adversaries is an essential ingredient in constructions aiming to provide security (privacy + correctness) in adversarial models of intermediate (between passive and active) strength. Using our analysis we are able to show that the protocols used by the Sharemind secure multi-party computation platform are actively private.