CS Talk - Mahnush Movahedi
Host: Mariana Raykova
Resource-Efficient and Robust Distributed Computing
Abstract:
There has been a tremendous growth in the size of distributed systems in the past three decades. Today, distributed systems, have become so large that they require highly scalable algorithms; algorithms that have asymptotically-small communication, computation, and latency costs with respect to the network size. Moreover, systems with thousands or even millions of parties distributed throughout the world is likely in danger of faults from untrusted parties or channels. In this talk, we study scalable and secure distributed algorithms that can tolerate such faults.
Throughout this talk, we balance two important and often conflicting characteristics of distributed protocols: security and efficiency. Our first result is a protocol that solves the MPC problem in polylogarithmic communication and computation cost and is secure against an adversary than can corrupt a third of the parties. We adapted our synchronous MPC protocol to the asynchronous setting when the fraction of the corrupted parties are less than 1/8.
Furthermore, we presented a protocol that can solve the interactive communication problem over a noisy channel when the noise rate in unknown. In this problem, we have focused on the cost of the protocol in the resource-competitive analysis model. Unlike classic models, resource-competitive models consider the cost that the adversary must pay to succeed in corrupting the protocol.