基于交易采样子图的以太坊钓鱼检测方法  

Ethereum phishing detection method based on transaction sampling subgraph network

作　　者：雷帮春 孟坤[1] 宋登科 陈大兴 LEI Bang-chun;MENG Kun;SONG Deng-ke;CHEN Da-xing(Joint Laboratory of Perception and Intelligence,Beijing Information Science and Technology University,Beijing 100101,China)

机构地区：[1]北京信息科技大学感知与智能联合实验室,北京100101

出　　处：《计算机工程与设计》2025年第3期690-697,共8页Computer Engineering and Design

基　　金：北京市教委科研计划基金项目(KM201911232002);广东省绿算技术有限公司烽烟研究院科技基金项目(9152223405)。

摘　　要：以太坊提供了图灵完备的去中心网络应用开发和部署环境,所充斥的钓鱼诈骗类应用已成为以太坊的最大安全威胁。为提高以太坊钓鱼检测的效率问题,提出一种基于交易采样子图网络的以太坊钓鱼检测方法。构建目标地址的原始交易子图;设计分层网络采样方法和时序映射机制构建它的4种交易采样子图并提取图特征;利用去重特征融合来消除冗余,输出融合特征来训练随机森林分类器。实验验证了该方法的合理性和有效性,获得了较高的F1值。Ethereum provides a Turing-complete decentralized network application development and deployment environment,and the phishing and scamming applications have become the biggest security threat to Ethereum.To improve the efficiency of Ethereum phishing detection,an Ethereum phishing detection method based on transaction sampling subgraph network was proposed.The original transaction subgraph of the target address was constructed.A hierarchical network sampling method and a timing mapping mechanism were designed to construct its four transaction sampling subgraphs and extract graph features.The deduplication feature fusion was used to eliminate redundancy,and the fused features were output to train the random forest classifier.The experiments verify the rationality and effectiveness of the method,and a high F1 value is obtained.

关 键 词：以太坊 钓鱼检测方法 交易采样子图网络 网络采样方法 图特征方法 去重特征融合 随机森林分类器 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]