一种基于格的隐私保护聚类数据挖掘方法

发布时间：2018-03-30 10:25

本文选题：数据挖掘　切入点：隐私保护　出处：《软件学报》2017年09期

【摘要】：由于云计算的诸多优势,用户倾向于将数据挖掘和数据分析等业务外包到专业的云服务提供商,然而随之而来的是用户的隐私不能得到保证.目前,众多学者关注云环境下敏感数据存储的隐私保护问题,而隐私保护数据分析的相关研究还比较少.但是如果仅仅为了保护数据隐私,而不对大数据进行挖掘分析,大数据也就失去了其潜在的巨大价值.提出了一种云计算环境下基于格的隐私保护数据挖掘方法,利用格加密构建隐私数据的安全同态运算方法,并且在此基础上实现了支持隐私保护的云端密文数据聚类分析数据挖掘服务.为保护用户数据隐私,用户将数据加密之后发布给云服务提供商,云服务提供商利用基于格的同态加密算法实现隐私保护的k-means、隐私保护层次聚类以及隐私保护DBSCAN数据挖掘服务,但云服务提供商并不能直接访问用户数据破坏用户隐私.与现有的隐私数据发布方法相比,隐私数据发布基于格的最接近向量困难问题(CVP)和最短向量困难问题(SVP)具有很高的安全性.同时,有效保持了密文数据间距离的精确性.与现有研究相比,挖掘结果也具有更高的精确性和可用性.对方法的安全性进行了理论分析,并设计实验对提出的隐私保护数据挖掘方法效率进行评估,实验结果表明,提出的基于格的隐私保护数据挖掘算法与现有的方法相比具有更高的数据分析精确性和计算效率.
[Abstract]:Because of the advantages of cloud computing, users tend to outsource services such as data mining and data analysis to professional cloud service providers. However, the privacy of users is not guaranteed. Many scholars have paid close attention to the privacy protection of sensitive data storage in the cloud environment, but the research on privacy protection data analysis is still relatively few. But if big data is not mined and analyzed just to protect data privacy, Big data also lost its potential huge value. He proposed a method of privacy protection data mining based on lattice in cloud computing environment, using lattice encryption to construct a secure homomorphism operation method of privacy data. On this basis, the data mining service of cloud ciphertext clustering analysis supporting privacy protection is implemented. In order to protect the user's data privacy, the user encrypts the data and publishes it to the cloud service provider. Cloud service providers use lattice based homomorphic encryption algorithm to implement privacy protection k-means, privacy protection hierarchical clustering and privacy protection DBSCAN data mining service. However, cloud service providers do not have direct access to user data to undermine their privacy. Privacy data publishing based on the nearest vector difficulty problem (CVP) and the shortest vector difficulty problem (SVP) have high security. At the same time, the accuracy of the distance between ciphertext data is effectively maintained. The results of mining also have higher accuracy and availability. The security of the method is analyzed theoretically, and the efficiency of the proposed privacy data mining method is evaluated by designing experiments. The experimental results show that, Compared with the existing methods, the proposed data mining algorithm of privacy protection based on lattice has higher data analysis accuracy and computational efficiency.
【作者单位】：软件工程国家重点实验室(武汉大学);武汉大学计算机学院;
【基金】：国家自然科学基金(61232002,61572378,61202034) CCF中文信息技术开放课题(CCF2014-01-02) 武汉市创新团队项目(2014070504020237) 武汉大学自主科研项目(2042016gf0020,2016-2017)~~
【分类号】：TP309;TP311.13

【相似文献】