Improved HotStuff consensus algorithm based on reputation mechanism grouping

Abstract To address the limitations in reward-punishment mechanisms and leader election in the HotStuff consensus algorithm, we propose an enhanced version called the reputed HotStuff (RHS) consensus algorithm, which introduces reputation-based grouping. In RHS, consensus is driven by high-reputation leaders, with nodes grouped according to their reputation scores. Witness nodes monitor and evaluate node behavior, while consensus nodes elect the leader, who stakes part of their reputation to initiate a new consensus round. Guided by the witness nodes, the reputation-based voting mechanism reduces communication complexity to O(n), enhancing efficiency. The RHS algorithm effectively differentiates Byzantine nodes and prevents them from frequently becoming leaders, resulting in lower consensus latency and higher throughput. Experimental results demonstrate that with 37 nodes, including Byzantine participants, RHS achieves 53.3% higher throughput and 32.6% lower consensus latency compared to HotStuff. Furthermore, when compared to another reputation-based algorithm, reputed practical Byzantine fault tolerance (RPBFT), RHS offers 29.57% higher throughput and reduces consensus latency by 25.97%. The experimental results illustrate that RHS achieves notable advantages in enhancing communication efficiency and addressing system faults.