MAPBFT: multilevel adaptive PBFT algorithm based on discourse and reputation models

Abstract

The traditional practical Byzantine fault tolerance (PBFT) consensus algorithm has limitations due to its failure to consider node credibility and its static structure, leading to reduced adaptability and increased communication overhead, particularly when dealing with Byzantine nodes. To address these issues, we propose a multilevel adaptive practical Byzantine fault tolerance algorithm (MAPBFT) based on PBFT and incorporating adaptive techniques, including a discourse power mechanism. MAPBFT initially uses a reputation model to evaluate node parameters such as past performance, reliability, availability, and response delay, providing predictive insights for the adaptive algorithm. The adaptive algorithm then employs a multi-layer perceptron to predict the reputation scores of nodes. This enables the selection of high-reputation nodes for consensus participation, narrowing the consensus scope, and reducing communication overhead. Finally, discourse power is distributed differentially based on node reputation scores. The consensus advancement criterion is modified to a threshold achieved through accumulated discourse power, focusing the system on high-reputation nodes and enhancing consensus efficiency and resistance against malicious nodes. We conducted experiments to validate MAPBFT’s performance and compared it with PBFT and APBFT. Experimental results demonstrate that MAPBFT enhances throughput, reduces response delay and communication overhead, and improves security, outperforming the other protocols.