基于防预测分组的PBFT算法改进方案

Improvement of PBFT Algorithm Based on Unpredictable Grouping

针对PBFT算法(实用拜占庭容错算法)由于多次全网广播所表现出的通信开销大和共识效率低下的问题,多种改进算法通过网络分组的方式减小单次共识规模,普遍取得较好的效果。这类改进算法基于某种规则进行网络分组,如果分组结果被获知或预测,则攻击者通过有倾向性地攻击部分节点即可瘫痪整个网络的共识能力。论文提出一种具备随机性、不相干性和局部性的防预测分组算法来避免上述问题,然后结合两阶段PBFT改进算法提出RS-PBFT算法。实验证明,算法在通信开销、共识效率方面优于PBFT算法,并在容错能力上极大地保留PBFT算法的性能。

Aiming at the problems of high communication overhead and low algorithm caused by PBFT(practical byzantine fault tolerance algorithm),some algorithms based on grouping have better performance by reducing the size of consensus nodes.But they are based on the regular rules to group nodes.If the grouping result is known or predicted,the attacker can disable the consensus capability of the network by attacking some nodes.This paper proposes a grouping algorithm with randomness,irrelevancy and locality is proposed to avoid the above problems.A consensus algorithm called RS-PBFT is proposed combining with the two-level PBFT algorithm.Experimental results show that RS-PBFT has better performance in terms of communication overhead and consensus delay,and retains the performance of PBFT in fault tolerance.