Good-case and Bad-case Latency of Unauthenticated Byzantine Broadcast: A Complete Categorization

TL;DR

This paper provides a complete categorization of good-case and bad-case latency in unauthenticated Byzantine broadcast, establishing tight resilience thresholds and proposing protocols for asynchronous systems.

Contribution

It introduces a comprehensive analysis of latency thresholds, proves impossibility results, and proposes four protocols for asynchronous Byzantine reliable broadcast.

Findings

Resilience threshold of n ≥ 4f separates 2-round and 3-round good-case latency.

Matching upper and lower bounds for bad-case latency with optimal 2-round good-case latency.

Four asynchronous Byzantine reliable broadcast protocols are proposed.

Abstract

This paper studies the {\em good-case latency} of {\em unauthenticated} Byzantine fault-tolerant broadcast, which measures the time it takes for all non-faulty parties to commit given a non-faulty broadcaster. For both asynchrony and synchrony, we show that $n\geq 4f$ is the tight resilience threshold that separates good-case 2 rounds and 3 rounds. For asynchronous Byzantine reliable broadcast (BRB), we also investigate the {\em bad-case latency} for all non-faulty parties to commit when the broadcaster is faulty but some non-faulty party commits. We provide matching upper and lower bounds on the resilience threshold of bad-case latency for BRB protocols with optimal good-case latency of 2 rounds. In particular, we show 2 impossibility results and propose 4 asynchronous BRB protocols.