基于信誉机制的改进PBFT共识算法

Improved PBFT consensus algorithm based on reputation mechanism

针对实用拜占庭容错共识算法(practical Byzantine fault tolerant,PBFT)通信开销大和缺乏奖惩机制的问题,提出一种基于信誉机制的改进PBFT共识算法RPBFT(reputed practical Byzantine fault tolerance)。首先,引入信誉机制对节点评分,将参与共识的节点分为收集器节点和普通共识节点,并对恶意节点进行惩罚。其次,收集器节点负责收集普通共识节点的投票消息,避免普通共识节点之间的通信,从而降低通信开销。最后,当普通共识节点中的拜占庭节点均无恶意行为时,通过增加收集所需的投票数量,减少一次投票收集过程,实现快速共识。实验结果表明,RPBFT能够有效地发现恶意节点并对其作出惩罚,同时具有更低的通信开销、平均共识时延以及更高的共识吞吐量。当节点总数为37时,与SBFT相比,RPBFT将平均共识时延降低25.2%以上,并将共识吞吐量提高39%以上。

Aiming at the problem that PBFT had high communication overhead and lacks reward and punishment mechanism,this paper proposed a reputed PBFT consensus algorithm(RPBFT)based on reputation mechanism.Firstly,the algorithm introduced a reputation mechanism to score nodes,dividing the nodes participating in the consensus into collector nodes and ordinary consensus nodes,and punishing malicious nodes.Secondly,the collector node was responsible for collecting the voting messages of ordinary consensus nodes to avoid communication between ordinary consensus nodes,thus reducing communication overhead.Finally,when the Byzantine nodes in the ordinary consensus nodes had no malicious behavior,the algorithm achieved a fast consensus by increasing the number of votes required for collection and reducing the process of collecting votes once.The experimental results show that RPBFT can effectively detect and punish malicious nodes,while having lower communication overhead,average consensus latency,and higher consensus throughput.When the total number of nodes is 37,RPBFT reduced the average consensus latency by more than 25.2%and increased consensus throughput by more than 39%compared to SBFT.