PhD Seminar • Systems and Networking — Synkletos: A New Approach to Blockchain Consensus Incentive Design

Friday, January 24, 2020 1:30 pm - 1:30 pm EST (GMT -05:00)

Yuhao Dong, PhD candidate
David R. Cheriton School of Computer Science

Consensus algorithms used to secure public blockchains differ significantly in their design constraints from those driving traditional fault-tolerant distributed systems. This is because blockchains must rigorously model trust within a game-theoretical model, rather than simply making assumptions about fault tolerance. To truly achieve incentive-compatible security, we cannot rely on the typical approach of considering ideal honest behavior and then positing an adversary with certain powers. Unfortunately, game-theoretical analysis of multi-party coordination problems, of which blockchain consensus is an instance, tends to be pernicuously difficult, leading to most consensus algorithms in use lacking formal analysis.

Synkletos is a blockchain consensus algorithm designed through a novel approach that drastically simplifies incentive analysis. Instead of modeling the ideal blockchain as decentralized parties participating in a coordination game to produce a certain optimal behavior, we start by proposing an ideal monopoly blockchain, where blockchain rules are such that even a selfish monopoly will behave in a “faulty” way. We then design an simple incentive structure and consensus algorithm based on a variation of proof of stake that incentivizes any number of uncoordinated or coordinated parties to simulate such a monopoly as a whole. We argue that such a mechanism, though it is slightly less economically efficient compared to systems such as Nakamoto consensus that rely on noncoordination assumptions, is much easier to analyze and far more robust to game-theoretical attacks.