Research On The Identification And Classification Of Ethereum Accounts Based On Graph Representation Learning

As one of the most popular blockchain platforms,Ethereum can achieve the automation execution of code through smart contracts,providing more flexible and diverse application scenarios.Therefore,Ethereum has attracted a large number of organizations and individuals to join,creating different types of accounts such as exchanges,miners,and token contracts.These accounts play an important role in promoting the development and application of blockchain technology.However,illegal users can also create a large number of accounts to engage in fishing,money laundering,fraud and other activities,which poses a serious threat to the security and stability of the blockchain.Therefore,how to identify various types of accounts in Ethereum is an urgent problem to be solved.Graph representation learning has been applied in Ethereum account identification thanks to its efficient representation capabilities for complex network structures.This article takes anonymous accounts in the Ethereum transaction network as the research object,and comprehensively uses graph representation learning technology to conduct research on various types of accounts in Ethereum.The main research work is as follows:(1)Research on the construction method of the Ethereum account transaction datasetThis paper proposes a method to collect large-scale Ethereum account transaction data automatically.The method includes three collection strategies: node data collection,transaction data collection,and smart contract data collection strategies.Through these three collection strategies,node data,transaction data,and smart contract data of common Ethereum accounts can be automatically collected.The transaction features are then mined to construct Ethereum account transaction datasets,which provide data support for subsequent Ethereum illegal account identification and account classification.(2)Research on the application of feature fusion graph embedding method in identifying illegal Ethereum accountsThis paper proposes a graph embedding method based on feature fusion and applies it to identify illegal Ethereum accounts.The method includes two feature extraction strategies:account feature extraction and transaction feature extraction.Specifically,the BP neural network is used to extract account features,and the random walk strategy is used to extract transaction features.Then,the extracted account features and transaction features are fused to obtain the representation of Ethereum accounts.The experimental results show that this method has better performance indicators than other algorithms and is capable of effectively detecting illegal accounts in Ethereum.(3)Research on feature enhanced graph neural network method and its application in Ethereum account classificationThis paper proposes a feature enhanced graph neural network and applies it to classify Ethereum accounts.The method includes two feature enhancement components: a convolutional layer and an attention layer.The convolutional layer generates a new graph structure and uses multiple candidate adjacency matrices to find the new graph structure.The attention layer uses attention mechanisms to focus on neighboring nodes to calculate the hidden representation of each node.Then,by repeatedly stacking multiple convolutional layers and attention layers to gradually enhance node features.The experimental results indicate that this method has better performance indicators than other algorithms and can effectively classify various types of accounts in Ethereum.This paper conducts research on various types of Ethereum account transaction data collected by synthetically using various graph representation learning methods.Based on this research,the paper proposes a feature fusion graph embedding method and a f feature enhanced graph neural network method,which improve the performance of identifying illegal Ethereum accounts and account classification.