机器学习算法在同态加密数据集上的应用  

Machine learning algorithm for a homomorphic encrypted data set

作　　者：贾春福 王雅飞 陈阳 孙梦洁 葛凤仪 JIA Chunfu;WANG Yafei;CHEN Yang;SUN Mengjie;GE Fengyi(College of Cyberspace Security,Nankai University,Tiainjin 300350,China;Tiainjin Key Laboratory of Network and Data Security Technology,Tiainjin 300350,China)

机构地区：[1]南开大学网络空间安全学院,天津300350 [2]天津市网络与数据安全技术重点实验室,天津300350

出　　处：《清华大学学报（自然科学版）》2020年第6期456-463,共8页Journal of Tsinghua University(Science and Technology)

基　　金：国家自然科学基金资助项目(61972215,61702399,61972073);国家重点研发计划项目(2018YFA0704703);天津市自然科学基金资助项目(17JCZDJC30500)。

摘　　要：大数据时代要求数据在云端进行存储和计算,这导致敏感数据隐私泄露的问题。该文提出了一种在同态加密数据集上应用机器学习分类算法的方案:首先对明文进行预处理,保证其满足对数据进行同态加密的要求;然后在加密数据集上通过协议实现比较、排序等操作;最后获取分类结果。客户端将加密数据上传,可以保证服务器端不会获取任何敏感信息;选取同态加密算法,能够保证服务器端仍可对密文执行相关操作。实验结果表明:该方案适用于Bayes、超平面和决策树分类器,其经过修正具有良好的适用性能,准确率高。The continuous development of big data requires that data be stored and analyzed in the cloud,which leads to privacy leakage of sensitive data.This paper presents a machine learning classification algorithm for homomorphic encrypted data sets.Firstly,preprocess the data set to meet the requirements of homomofphic encryption.The encrypted data set is then sorted by protocol and classified.Finally,the classification results are obtained.The client can then upload encrypted data and ensure that the server will not get any sensitive information.A homomorphic encryption algorithm is used to ensure that the server can still perform required operations on the ciphertext.Tests show that this scheme can provide accurate,useful results with Bayes,hyperplane and decision tree classifiers.

关 键 词：隐私保护 同态加密 分类器算法 加密数据运算 

分 类 号：TP309.2[自动化与计算机技术—计算机系统结构]