On the Internet, a huge number of privacy-sensitive transactions are being performed, and it is imperative to have mechanisms available that guarantee privacy in various contexts. In this talk, I will discuss recent work on generic secure multi-party computation (MPC), which enables any number of parties to jointly evaluate an arbitrary function of their inputs while preserving security properties such as privacy and correctness.
In the first part, I will describe work on the generic MPC solution for boolean circuits. We implemented the first scalable protocol for boolean circuits with optimal resilience. We also applied our solution to the design of a distributed, privacy-preserving marketplace, for which our solution runs at least 10 times faster than previous works.
In the second part, I will discuss the security of the very effective technique recently introduced by Kolesnikov and Schneider, which provides a 40% efficiency improvement for generic two-party computation. We pinned down the cryptographic assumptions needed to prove security of this technique.