Dynamic Scalable PBFT Consensus Algorithm Based on Binary K-Means

Addressing issues such as the inability to dynamically add nodes, low consensus efficiency among multiple nodes, and lack of enthusiasm among nodes in existing Practical Byzantine Fault Tolerance (PBFT) consensus algorithms, this paper proposes a dynamic PBFT consensus mechanism based on binary K-means—BK-PBFT. First, a reputation mechanism is introduced to enhance the security of the system, improving node motivation through an improved Fibonacci function, and classifying nodes into three levels based on reputation: high-quality nodes, ordinary nodes, and untrusted nodes. This classification is used to select consensus nodes. Next, the improved binary K-means algorithm is employed to group nodes in the blockchain network, improving the efficiency of consensus when many nodes participate. When new nodes join, they are assigned to the most appropriate group based on their characteristics. Finally, the consensus process of the PBFT algorithm is simplified, adopting an intra-group first consensus followed by inter-group consensus. Experimental results show that compared to the PBFT algorithm, the BK-PBFT consensus mechanism exhibits significant advantages in terms of flexibility in dynamically adding or removing nodes, and achieves superior performance in multiple metrics such as consensus delay, communication costs, and throughput.