在没有诚实参与者的情况下实现状态机复制
Achieving State Machine Replication without Honest Players
摘要:
用于令牌化状态机复制协议中的玩家表征的现有标准取决于将始终遵循协议的诚实玩家,而不管可能的令牌增加以用于偏离。鉴于这些令牌化协议的市值不断增加,诚实变得越来越昂贵和不切实际。因此,这种过时的玩家特征必须被删除,以提供真正的安全性和活性保证,从而朝着普遍信任状态机复制协议和新的采用规模迈出重要一步。由于所有当前的状态机复制协议都是基于这些传统标准构建的,因此必须识别并利用新的参与者模型来反映令牌化协议中参与者的真实性质。为此,我们提出了状态机复制协议的ByRa播放器模型。在ByRa模型中,玩家要么试图最大化他们的代币化奖励,要么表现得敌对。这融合了博弈论和分布式系统的领域,其中存在令牌化状态机复制协议,但很少进行形式化。在ByRa模型中,我们确定了所有协议必须满足的期望和公平性的强激励兼容性,以实现状态机复制。然后,我们提供Tenderstake,一个协议,可证明满足这些属性,并通过这样做,实现状态机复制的ByRa模型。
英文摘要:
Existing standards for player characterisation in tokenised state machine replication protocols depend on honest players who will always follow the protocol, regardless of possible token increases for deviating. Given the ever-increasing market capitalisation of these tokenised protocols, honesty is becoming more expensive and more unrealistic. As such, this out-dated player characterisation must be removed to provide true guarantees of safety and liveness in a major stride towards universal trust in state machine replication protocols and a new scale of adoption. As all current state machine replication protocols are built on these legacy standards, it is imperative that a new player model is identified and utilised to reflect the true nature of players in tokenised protocols, now and into the future. To this effect, we propose the ByRa player model for state machine replication protocols. In the ByRa model, players either attempt to maximise their tokenised rewards, or behave adversarially. This merges the fields of game theory and distributed systems, an intersection in which tokenised state machine replication protocols exist, but on which little formalisation has been carried out. In the ByRa model, we identify the properties of strong incentive compatibility in expectation and fairness that all protocols must satisfy in order to achieve state machine replication. We then provide Tenderstake, a protocol which provably satisfies these properties, and by doing so, achieves state machine replication in the ByRa model.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。