Integrating machine learning with proof-of-authority-and-association for dynamic signer selection in blockchain networks

Abstract

Integrating machine learning (ML) into blockchain consensus mechanisms enhances efficiency, scalability, and resilience. This study introduces the PoA${}^2$ algorithm, an ML-enhanced Proof of Authority mechanism that optimizes signer selection for improved transaction processing. Simulations with models including Random Forest, Logistic Regression, SVM, K-Nearest Neighbors, Decision Tree, and Gradient Boosting showed significant gains. Random Forest reduced latency tenfold, achieving nearly 1000 transactions per second, with 93.33% accuracy, 100% precision, 86.67% recall, and a 92.86% F1-score. These results demonstrate ML’s potential to enhance blockchain performance, making hybrid blockchain-ML solutions a promising research direction.