Secure Distributed Computing Beyond Classical Honesty Assumptions by Dr Tiantian Gong

Abstract

Secure Multi-Party Computation (MPC) and consensus are foundational primitives in secure distributed computing (SDC), with MPC enabling mutually distrustful parties to compute functions on private inputs and consensus allowing parties to agree on outputs. Classical SDC protocols, however, rely on strong honesty assumptions. For example, both information-theoretic MPC (with guaranteed output delivery) and partially synchronous Byzantine agreement requires an honest supermajority.

In this talk, I present new techniques for preserving privacy even when the number of honest parties falls below traditional thresholds in secret-sharing–style protocols, which are major topics in MPC. These results address longstanding feasibility barriers by combining tools from cryptography, information theory, algorithmic game theory, and coding theory, and they open new pathways toward MPC deployment in real-world adversarial settings. I will then discuss recent progress on correctness guarantees in partially synchronous Byzantine agreement under excessive fault regimes. I introduce the notion of recoverability as a new lens for analyzing Byzantine agreement. I will conclude by outlining future research directions aimed at advancing secure distributed computing beyond classical honesty assumptions, including the design of general SDC protocols with relaxed honesty requirements and the translation of these theoretic advances into practical distributed systems.