Ethereum Analysis via Node Clustering

Abstract

As an open source public blockchain with the capabilities of running smart contract, Ethereum provides decentralized Ethernet virtual machines to handle peer-to-peer contracts through its dedicated cryptocurrency Ether. And as the second largest blockchain, the amount of transaction data in Ethereum grows fast. Analysis of these data can help researchers better understand Ethereum and find attackers among the users. However, the analysis of Ethereum data at the present stage is mostly based on the statistical characteristics of Ethereum nodes and lacks analysis of the transaction behavior between them. In this paper, we apply machine learning in Ethereum analysis for the first time and cluster users and smart contract into groups by using transaction information in existing blocks. The clustering results are analyzed by using the identity information of the available Ethereum users and smart contracts. Based on the clustering results, we propose a new way of user identity discrimination and malicious user detection.