满足差分隐私的dK序列合成图发布  

Publishing dK series composition graph satisfying differential privacy

作　　者：周楠楠[1,2] 龙士工 刘海[1,2,3] Zhou Nannan;Long Shigong;Liu Hai(Guizhou Provincial Key Laboratory of Public Big Data,Guizhou University,Guiyang 550025,China;College of Computer Science&Technology,Guizhou University,Guiyang 550025,China;School of Information,Guizhou University of Finance&Economics,Guiyang 550025,China)

机构地区：[1]贵州大学公共大数据国家重点实验室,贵阳550025 [2]贵州大学计算机科学与技术学院,贵阳550025 [3]贵州财经大学信息学院,贵阳550025

出　　处：《计算机应用研究》2023年第5期1528-1534,共7页Application Research of Computers

基　　金：国家自然科学基金资助项目(62062020,62062017);贵州省自然科学基金资助项目([2020]1Y265)。

摘　　要：为了保护用户隐私,社交网络图数据发布前通常对其进行匿名化操作。然而,现有的各种匿名技术不能很好地保护用户隐私并且由于改变太大影响社交网络数据可用性。提出一种满足差分隐私的社交网络图数据发布模型(differential privacy perturbation graph, DPPM)。该模型将一个图的结构信息提取到dK度的相关统计中,将噪声引入到数据集中,并生成一个社交网络图。从理论上证明了该方法满足差分隐私。使用三个真实的社交网络数据集来评估所提方法的有效性。To protect user privacy,social network graph data is usually anonymized before publishing.However,the existing anonymous technologies can not protect user privacy well and affect the availability of social network data due to too much change.This paper proposed a data publishing model(DPPM)for social network graph satisfying differential privacy.This mo-del extracted the structure information of a graph into dK degree correlation statistics,introduced noise into the dataset,and generated a social network graph.It was proved theoretically that the method satisfies differential privacy.This paper used three real social network datasets to evaluate the effectiveness of the proposed method.

关 键 词：社交网络 差分隐私 dK序列 

分 类 号：TP309.2[自动化与计算机技术—计算机系统结构]