An Improved Practical Byzantine Fault-Tolerant Algorithm Based on XGBoost Grouping for Consortium Chains

In response to the challenges presented by the unreliable identity of the master node, high communication overhead, and limited network support size within the Practical Byzantine Fault-Tolerant (PBFT) algorithm for consortium chains, we propose an improved PBFT algorithm based on XGBoost grouping called XG-PBFT in this paper. XG-PBFT constructs a dataset by training important parameters that affect node performance, which are used as classification indexes for nodes. The XGBoost algorithm then is employed to train the dataset, and nodes joining the system will be grouped according to the trained grouping model. Among them, the nodes with higher parameter indexes will be assigned to the consensus group to participate in the consensus, and the rest of the nodes will be assigned to the general group to receive the consensus results. In order to reduce the resource waste of the system, XG-PBFT optimizes the consensus protocol for the problem of high complexity of PBFT communication. Finally, we evaluate the performance of XG-PBFT. The experimental results show that XG-PBFT can significantly improve the performance of throughput, consensus delay and communication complexity compared to the original PBFT algorithm, and the performance enhancement is significant compared to other algorithms in the case of a larger number of nodes. The results demonstrate that the XG-PBFT algorithm is more suitable for large-scale consortium chains.