Overview
Multi-Party Computation (MPC) is arguably one of the most fundamental problems in the field of
Cryptography and Distributed Computing. It allows a set of parties to perform a computation over
private data, without revealing any sensitive information. Since its inception more than thirty years ago,
a significant line of work has been done, both in terms of theoretical work and feasibility in practice.
Today, MPC is a technology that is becoming more and more practical, and there have been a number
of efforts to deploy such systems in the real world.
Despite the efforts, the vast majority of improvements with respect to the efficiency of MPC consider
an idealized communication network that is so-called ‘synchronous’, where parties have access to
synchronized clocks and an upper bound on the network communication delay is assumed. While this
model is theoretically interesting and allows to study clean protocols, it fails to capture real-world
network behaviors such as the Internet, which typically have an unpredictable delay and are
asynchronous.
The state of the art of the efficiency of asynchronous MPC protocols is far from its synchronous
counterpart. In particular, the most efficient synchronous MPC protocols up to date achieve a
communication efficiency that grows linearly with the number of parties involved and the computation
size, and make use of very light computation tools (many solutions are even information-theoretic and
the computation only consists of linear operations on the local data). On the other hand, current
asynchronous MPC protocols grow at least quadratically in the number of parties and make use of
cryptographic tools that require more computation (such as homomorphic encryption).
The goal of this project is to develop new techniques that allow for the design of efficient MPC
protocols that can be deployed over the Internet. In particular, we envision the application of MPC-as-
a-service with strong security guarantees (such as fairness and robustness). Since we aim for a large-
scale application involving thousands of users performing complex computations, our focus is on
solutions that use less computation-heavy cryptographic primitives as possible.
