基于聚类的差分隐私民航旅客数据发布算法  

Differential privacy civil aviation passenger data release algorithm based on clustering

作　　者：丁建立[1] 杜天天 DING Jian-li;DU Tian-tian(School of Computer Science and Technology,Civil Aviation University of China,Tianjin 300300,China)

机构地区：[1]中国民航大学计算机科学与技术学院,天津300300

出　　处：《计算机工程与设计》2022年第3期608-615,共8页Computer Engineering and Design

基　　金：国家自然科学基金项目(U1833114);民航安全能力基金项目(SA2020280)。

摘　　要：为使数据管理者可以发布数据集供研究人员进行挖掘分析,对数据集采用满足差分隐私的保护算法,但其中会加入大量噪声,破坏数据可用性,因此,提出一种基于聚类的差分隐私民航旅客数据发布算法。改进聚类算法,按照数据类型的不同,对数值型属性和分类型属性分别选用不同的距离计算方法,将更可能相关的记录分为一组,降低差分隐私敏感度,结合聚类结果形成的簇,采用差分隐私保护技术对数据记录进行加噪。实验结果表明,算法能够在降低信息损失的同时防止信息泄露。To enable data managers to release data sets for researchers to conduct mining and analysis,a protection algorithm that satisfies differential privacy is adopted for the data sets,but a large amount of noise will be added to it,thereby destroying data availability.Therefore,a clustering-based differential privacy civil aviation passenger data release algorithm was proposed.The clustering algorithm was improved.According to different data types,different distance calculation methods were selected for the numerical attributes and sub-type attributes,which were more likely to be related.The records were divided into a group to reduce the sensitivity of differential privacy.Combining the clusters formed by the clustering results,the differential privacy protection technology was used to add noise to the data records.Experimental results show that the proposed algorithm can reduce information loss while preventing information leakage.

关 键 词：差分隐私 民航旅客数据 数据发布 聚类 隐私敏感度 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]