ZLB: A Blockchain to Tolerate Colluding Majorities

TL;DR

ZLB introduces a novel blockchain protocol capable of tolerating majority adversaries by combining accountable Byzantine agreement with transient disagreement tolerance, outperforming existing systems in speed and fault tolerance.

Contribution

It presents the first blockchain that can tolerate more than half of the system being malicious, leveraging recent theoretical advances to exclude deceitful replicas.

Findings

ZLB tolerates >50% adversarial control.

ZLB outperforms HotStuff in geo-distributed settings.

ZLB is nearly as fast as Red Belly Blockchain.

Abstract

In the general setting, consensus cannot be solved if an adversary controls a third of the system. Yet, blockchain participants typically reach consensus "eventually" despite an adversary controlling a minority of the system. Exceeding this $\frac{1}{3}$ cap is made possible by tolerating transient disagreements, where distinct participants select distinct blocks for the same index, before eventually agreeing to select the same block. Until now, no blockchain could tolerate an attacker controlling a majority of the system. In this paper, we present Zero-Loss Blockchain (ZLB), the first blockchain that tolerates an adversary controlling more than half of the system. ZLB is an open blockchain that combines recent theoretical advances in accountable Byzantine agreement to exclude undeniably deceitful replicas. progressively reduces the portion of deceitful replicas below $\frac{1}{3}$, and reaches consensus. Geo-distributed experiments show that ZLB outperforms HotStuff and is almost as fast as the scalable Red Belly Blockchain that cannot tolerate $n/3$ faults.