Algorithm for byzantine fault-tolerant consensus to support dynamic feedback decision-making

The blockchain is popular as a distributed ledger that can record encryptions, ensuring the consistency of transactions in the ledger through consensus algorithms across a network of untrusted nodes. The broadcasting process of voting-based consensus protocols requires a large number of network forwards, with a huge communication overhead that severely affects the on-chain performance. With the increase in the number of nodes, the performance deteriorates dramatically, and the scalability of large-scale nodes is severely constrained. To address the above problems, the consensus process is treated as an optimization problem, and a Byzantine fault-tolerant consensus algorithm using learning automata for voting decision is proposed. Learning automata are embedded into blockchain nodes instead of nodes to complete consensus voting-related operations, reducing the impact of malicious operations of nodes on the system. The consensus decision is made by the master node and its neighboring learning nodes, the master node gives feedback to the learning nodes based on the criterion function and the overall voting result, the learning nodes adjust their voting strategy based on the feedback, and the consensus is reached when the criterion function of the master node converges. The proposed strategy accelerates the convergence of consensus, adjusts the rules of learning automata to reduce the influence of faulty nodes, and uses the reward and punishment mechanism to improve the enthusiasm of normal nodes to participate in the consensus process and reduce the consensus delay. Experimental results show that the consensus algorithm proposed in this paper has a lower complexity compared to the existing algorithms in large-scale node scenarios, and also shows a better fault tolerance in the face of Byzantine nodes, which reduces the impact of the faulty nodes while maintaining the fast transactions and ensures the scalability and fault tolerance of large-scale node networks. © 2024 Science Press. All rights reserved.