基于节点聚类的数据起源安全方法研究  

Data provenance security method based on node clustering

作　　者：孙连山[1] 马胜天 王荔 陈秀婷 SUN Lianshan;MA Shengtian;WANG Li;CHEN Xiuting(School of Electronic Information and Artificial Intelligence,Shaanxi University of Science and Technology,Xi’an 710021,China)

机构地区：[1]陕西科技大学电子信息与人工智能学院,陕西西安710021

出　　处：《南京邮电大学学报（自然科学版）》2021年第5期67-76,共10页Journal of Nanjing University of Posts and Telecommunications：Natural Science Edition

基　　金：国家自然科学基金(61202019);陕西省自然科学基础研究计划(2019JM⁃354)资助项目。

摘　　要：数据起源的公开发布具有时序性,通常不是一次性完成的,因此先发布的起源图可能会对后发布的起源图造成安全威胁。针对这一场景,提出一种基于节点聚类的数据起源安全保护方法。首先获取已公开起源图,采用聚类的方法构建已公开起源图节点与当前起源图节点的聚类层次树;其次根据数据起源发布者的安全需求以及节点间的关联关系,对敏感节点及其关联节点进行泛化;最后,计算敏感节点泄露概率,满足数据起源发布者的期望后便停止泛化,得到过滤视图。通过3组测试数据验证,与现有方法相比,该方法能够在达到敏感节点隐私保护的同时,更大程度地保持溯源效用,并且满足不同数据起源发布者的偏好。The public release of the data provenance is time⁃sequential,and is usually not completed at once,thus the provenance graph which has been released earlier can pose a security threat to the later.For this scenario,a data provenance privacy protection method is proposed based on the node clustering.Firstly,the published provenance graph is obtained and the clustering method clustering is used to construct a clustering hierarchy tree of nodes in the published provenance graph and the current provenance graph.Secondly,according to security requirements of the data provenance publisher and the relationship between the nodes,it is necessary to generalize the sensitive nodes and their associated nodes.Finally,it is the time to stop generalization and obtains the sanitized provenance when the leakage probability of sensitive nodes meets data provenance publisher expectations.Through the verification of three groups of test data,compared with the existed methods,the provenance privacy protection method can maintain the traceability to the greatest extent and meet the preferences of different data provenance publishers.

关 键 词：数据起源 聚类 泛化 隐私保护 

分 类 号：TP309.2[自动化与计算机技术—计算机系统结构]