K-means‖隐私保护聚类算法  

K-means‖privacy protection clustering algorithm

作　　者：郑剑[1] 冷碧玉 ZHENG Jian;LENG Bi-yu(School of Information Engineering,Jiangxi University of Science and Technology,Ganzhou 341000,China)

机构地区：[1]江西理工大学信息工程学院,江西赣州341000

出　　处：《计算机工程与设计》2022年第1期26-33,共8页Computer Engineering and Design

基　　金：国家自然科学基金项目(61462034);江西省教育厅科学技术研究基金项目(GJJ170517)。

摘　　要：针对异常离群点对k-means‖算法的聚类精确度影响较大且在确定中心点过程中会泄露聚类数据隐私的问题,提出DPk-means‖算法。标记离群点,降低离群点对k-means‖算法聚类精确度的影响,将差分隐私应用于k-means‖聚类算法中保护聚类数据隐私。在选择聚类初始中心点和迭代求取均值中心点的过程中,应用拉普拉斯机制注入噪声,解决数据隐私泄露的问题。通过隐私预算动态变化对聚类结果准确性的影响及同类算法对比实验分析验证,DPk-means‖算法能够提供更高的隐私保护水平且保证聚类结果的准确性。Aiming at the problems that abnormal outliers have a greater impact on the clustering accuracy of the k-means‖algorithm and the privacy of the clustering data may be leaked in the process of determining the center point,the DPk-means‖algorithm was proposed,which marked the outliers to reduce the influence of outliers on the clustering accuracy of k-means‖algorithm and applied differential privacy to k-means‖clustering algorithm to protect the privacy of clustering data.In the process of selecting the initial center point of clustering and iteratively obtaining the mean center point,the Laplace mechanism was used to inject noise,and the problem of data privacy leakage was solved.Through experimental analysis of the influence of the dynamic change of privacy budget on the accuracy of clustering results and the comparison of similar algorithms,it is verified that the DPk-means‖algorithm can provide a higher level of privacy protection and ensure the accuracy of clustering results.

关 键 词：聚类精确度 并行化k均值 离群点 拉普拉斯机制 差分隐私 

分 类 号：TP309.2[自动化与计算机技术—计算机系统结构]