Performance Analyses for Applying Machine Learning on Bitcoin Miners

Abstract

Bitcoin and cryptocurrency rely on peer-to-peer (P2P) networking. Incorporating intelligence on Bitcoin miners to analyze and control networking can improve the information delivery and defend against networking threats. However, applying machine learning (ML) for building intelligence introduces a challenge because miners participate in the resource-intensive distributed consensus protocol and the ML application can consume much computing resources. In this paper, we study the feasibility and the interplay between ML algorithms and mining operations. Our prototype-based experiments measure and compare the performance of the ML algorithms to evaluate the implementation overhead and efficiencies of the ML algorithms and their impacts on mining operations, i.e., the mining reduction when the ML algorithm is running in parallel.