WANG X W, LI J.Research on the Raft improved algorithm for anti-Byzantine nodes[J].Journal of Henan Polytechnic University(Natural Science) ,2025,44(2):145-153.

doi:10.16186/j.cnki.1673-9787.2023060005

Received:2023/06/01

Revised:2023/12/14

Published:2025-03-05

Research on the Raft improved algorithm for anti-Byzantine nodes

WANG Xiaowei, LI Jie

Information Management Center， Physical Education College of Zhengzhou University， Zhengzhou 450052， Henan， China

Abstract: Objectives In order to solve the problem of malicious election and log tampering caused by Byzantine nodes in original Raft algorithm， an AntiB-Raft （Anti-Byzantine Raft） algorithm was proposed to resist Byzantine nodes. Methods When the candidate requested to replace the Leader， the heartbeat monitoring threshold mechanism was used to determine whether the candidate can successfully obtain enough votes to become the new Leader， which was agreed that only when more than half of the nodes detected that the current Leader was really down， the candidate can obtain more than half of the votes and become the new Leader， so as to prevent the Byzantine node from maliciously pulling votes when current Leader was not down， which would result in the replacement of the normal Leader. In the log verification phase， iterative hashing algorithm was used to encrypt the logs and appropriate verification time was selected for log verification. It was agreed that log verification was performed once every K transactions or Leader changed to ensure that the synchronized logs were correct. During log verification， if the log verification failed， the binary rollback method was used to locate the faulty log position quickly and retransmit it， which greatly improved the work efficiency. Results To simulate the 100-node election process， the normal Leader was replaced in Raft algorithm because the candidate obtained more than 50% of the votes， while neither RB-Raft nor the proposed algorithm exceeded 50%， thus avoiding malicious canvassing. On the anti-Byzantine， error logs were not identified in Raft algorithm， while the error log recognition rate of AntiB-Raft algorithm can reach 100%， meanwhile the consensus delay was 28% lower than the existing algorithm RB-Raft. Conclusions The proposed algorithm AntiB-Raft had the capability of anti-Byzantine and reduced consensus delay compared with the existing algorithm RB-Raft and its efficiency was improved significantly. 

Key words:Raft;consensus mechanism;Byzantine fault tolerance;iterative hash;heartbeat threshold