An Improved PBFT Algorithm Based on Reputation Mechanism and eBPF Network Acceleration

Abstract Practical Byzantine Fault Tolerance (PBFT) is a widely adopted consensus algorithm in blockchain systems, demonstrating excellent consensus efficiency in small-scale networks. However, as the number of nodes increases, its communication complexity increases exponentially, becoming a significant performance bottleneck for blockchain consensus. To address the issue of reduced consensus efficiency of PBFT in large-scale node environments, this paper proposes e-PBFT, a blockchain grouping consensus algorithm that integrates a reputation mechanism with eBPF-based network acceleration. The algorithm partitions the network into a consensus committee and an ordinary synchronization group, where the consensus committee is responsible for rapid agreement, and the ordinary synchronization group serves to mitigate the overhead caused by redundant communications. Additionally, a threshold signature scheme is employed to shorten the inter-group synchronization verification time, and eBPF is utilized to minimize unnecessary context switches. Experimental results demonstrate that under large-scale node conditions, the e-PBFT algorithm achieves a throughput improvement exceeding 14% compared to the PBFT algorithm, alongside a reduction in latency of over 39%, indicating a significant improvement in consensus efficiency. Furthermore, the algorithm enhances the system’s robustness against malicious nodes.