基于KNN的具有隐私保护功能的区块链异常交易检测  

Blockchain Abnormal Transaction Detection with Privacy-preserving Based on KNN

作　　者：陈彬杰 魏福山[1,2] 顾纯祥 CHEN Binjie;WEI Fushan;GU Chunxiang(School of Cyberspace Security,Information Engineering University,Zhengzhou 450001,China;Henan Key Laboratory of Network Cryptography Technology,Zhengzhou 450001,China)

机构地区：[1]信息工程大学网络空间安全学院,郑州450001 [2]河南省网络密码技术重点实验室,郑州450001

出　　处：《信息网络安全》2022年第3期78-84,共7页Netinfo Security

基　　金：国家自然科学基金[61772548]。

摘　　要：随着区块链技术的发展,以Hyperledger为代表的联盟链技术得到了广泛应用,其异常交易检测需求也逐渐凸显。但当前的区块链异常交易检测技术大多针对公有链设计,没有考虑联盟链交易的隐私保护需求。为了实现高效异常检测并保证联盟链交易的隐私性,文章提出一种基于KNN的具有隐私保护功能的区块链异常交易检测方案。该方案中记账节点使用矩阵乘法对交易数据进行随机化,云服务器使用KNN对随机化后的交易数据特征进行异常检测,并将结果反馈给联盟链记账节点进行验证。实验结果表明,该方案对联盟链效率的影响很小,同时具有良好的检测效果,召回率、精度、F1值分别可达85.3%、87.7%和86.5%。With the development of blockchain,the consortium blockchain technology represented by Hyperledger has been applied widely,and its abnormal transaction detection needs have become prominent gradually.The current anomaly detection technology focuses on public blockchain,which neglects the privacy protection requirements of the consortium blockchain.In order to realize efficient anomaly detection and privacy protection of the consortium blockchain,this paper proposed a privacy-preserving abnormal transaction detection scheme based on KNN.The accounting nodes of this scheme used the matrix method to randomize the transaction data,and cloud server used KNN to test the randomized transaction data and feed back the result to accounting nodes for validation.The experimental results show that the scheme has little effect on the efficiency of the consortium blockchain,and has a good detection performance.The recall rate,precision and F1 value can reach 85.3%,87.7% and 86.5% respectively.

关 键 词：区块链 联盟链 异常检测 隐私保护 KNN 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]