摘要
比特币作为第一个去中心化的加密货币,由于具有匿名性这一特点,被大量用在各种交易服务中,如博彩、跨境支付等,同时也被恶意交易所利用。目前比特币用户地址分类主要通过启发式聚类方法实现,受到比特币协议的变化影响,该方法对出现的新输出地址、单输入地址以及参与混币交易的地址无法分类,因此仅适用于现有比特币地址中很小一部分。针对该问题,本文提出一种基于改进随机森林的比特币地址分类方法,对区块链原始区块数据进行解析,直接提取用于机器学习的地址特征,进而学习一个随机森林分类器,能对任何一个没有标签的比特币地址进行分类,同时为了降低特征集中的冗余,对传统的随机森林进行改进以获取最终有助于地址分类的重要特征。实验结果表明,该方法可以准确地对比特币用户地址进行分类,仅仅需要14个重要特征。
As the first decentralized cryptocurrency,Bitcoin is widely used in various public services,such as gambling,cross-border payments,etc.,and it is also used by illegal exchanges due to its anonymity.At present,Bitcoin user address classifi⁃cation is mainly realized through heuristic clustering method,which is affected by the changes of Bitcoin protocol.This meth⁃od cannot classify new output addresses,single input addresses and addresses participating in mixed currency transactions that have not appeared before,therefore,it only applies to a small part of the existing Bitcoin addresses.In response to this problem,this paper proposes a Bitcoin address classification method based on improved random forest,which analyzes the original block data of the blockchain,directly extracts the address features for machine learning,and then learns a random for⁃est classifier which can classify any unlabeled Bitcoin address.At the same time,in order to reduce the redundancy of feature set,we improve the traditional random forest to obtain important features that ultimately help address classification.Experi⁃mental results show that this method can accurately classify Bitcoin user addresses,and only requires 14 important features.
作者
陶峰
王劲松
吕垛斌
赵泽宁
张洪玮
石凯
TAO Feng;WANG Jinsong;LV Duobin;ZHAO Zening;ZHANG Hongwei;SHI Kai(School of Computer Science and Engineering,Tianjin University of Technology,Tianjin300457,China;Tianjin Key Laboratory of Intelligent Computing and New Software Technology,Tianjin University of Technology,Tianjin300457,China;National Engineering Laboratory for Computer Virus Prevention and Control Technology,Tianjin University of Technology,Tianjin300457,China)
出处
《天津理工大学学报》
2022年第1期53-58,共6页
Journal of Tianjin University of Technology
基金
天津市新一代人工智能科技重大专项(19ZXZNGX00080)
天津市研究生科研创新项目(2019YJSS047)
天津市科学技术普及项目重点项目(20KPZDRC00040)。
关键词
比特币
随机森林
特征选择
匿名性
bitcoin
random forest
feature selection
anonymity
