面向医疗数据的AUR-Tree差分隐私数据发布算法  

AUR-Tree differential privacy data publishing algorithm for medical data

作　　者：张思琪 李晓会[1] 江欣俞 李波[1] Zhang Siqi;Li Xiaohui;Jiang Xinyu;Li Bo(School of Electronics&Information Engineering,Liaoning University of Technology,Jinzhou Liaoning 121000,China)

机构地区：[1]辽宁工业大学电子与信息工程学院,辽宁锦州121000

出　　处：《计算机应用研究》2022年第7期2162-2166,2171,共6页Application Research of Computers

基　　金：国家自然科学基金资助项目(61802161);辽宁省教育厅科学研究经费资助项目(JZL202015402);辽宁省自然科学基金资助项目(2020-MS-292)。

摘　　要：随着医学技术的进步和大数据时代的到来,在数据发布时如何对患者就诊记录中的敏感信息进行隐私保护成为当前的研究热点。针对医疗大数据在发布过程中隐私保护问题,提出了基于属性效用值排序法AUR-Tree(attribute utility value ranking-tree)差分隐私数据发布算法。该算法用属性效用值排序法衡量准标识属性对敏感属性的影响程度,以此作为迭代分割的度量依据,采用基于泛化的自顶向下迭代分割分类树技术,通过类等差法合理地分配隐私预算从而实现在医疗数据发布过程中的隐私保护。实验结果表明:该算法在极大地提高了数据的安全性、有效性和可用性的前提下,还保留了后续数据挖掘的价值。With the advancement of medical technology and big data era,how to protect the privacy of sensitive information in patients’medical records has become a current research focus issue.In order to protect the privacy of medical big data in the publishing process,this paper proposed an AUR-Tree differential privacy data publishing algorithm.The algorithm used the attribute utility value ranking method to measure the influence degree of the quasi-identity attributes on the sensitive attributes,which was as the measurement basis for iterative segmentation.Then it adopted top-down iterative segmentation classification tree technology based on generalization,and reasonably allocated the privacy budget through the class arithmetic method to realize the privacy protection in the medical data release process.The experimental results show that the algorithm retains the value of subsequent data mining and greatly improves the security,effectiveness and usability of the data.

关 键 词：属性效用值排序法 属性层次泛化树 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]