数据发布中基于背景知识模型的二值映射隐私风险度量研究  

Measuring the Privacy Risk with Binary Mapping Based Background Knowledge Model in Data Publishing

作　　者：赖特 詹雯 刘思杙 杨子辰 夏晓峰[2] LAI Te;ZHAN Wen;LIU Siyi;YANG Zichen;XIA Xiaofeng

机构地区：[1]国网四川省电力公司技能培训中心,四川成都610072 [2]重庆大学大数据与软件学院,重庆400044

出　　处：《电力系统装备》2025年第1期178-181,共4页Electric Power System Equipment

摘　　要：数据发布是数据挖掘数据集的重要来源之一,数据发布者应在最大程度上匿名化发布,以降低隐私泄漏的风险,而数据挖掘者则应关注发布数据在挖掘中的可用性,即解决隐私风险与数据可用性之间的平衡问题.隐私风险评估和数据可用性评估是解决平衡问题的两个关键步骤.文章提出了一个框架来对隐私保护数据发布中的背景知识建模,基于该模型提出了一种隐私度量方法,该方法满足文章中提出的4个理想属性.在实践中,应用马尔可夫链蒙特卡洛方法从背景知识分布中进行采样,并评估给定数据库的隐私风险.最后,通过在真实数据库上的试验验证了所提出模型和隐私度量方法的有效性.Data publishing is one of the important sources for data mining datasets.Data publishers are concerned with the extent to which published data should be anonymized to reduce the risk of user privacy leakage,while data miners focus on the usability of published data in mining,addressing the balance between privacy risk and data usability.Privacy risk assessment and data usability assessment are two key steps in resolving this balance issue.This paper proposes a framework to model background knowledge in privacy-preserving data publishing.Based on this model,a privacy metric method is introduced,which satisfies the four ideal properties proposed in this paper.In practice,the Markov Chain Monte Carlo method is applied to sample from the background knowledge distribution and assess the privacy risk of a given database.Finally,experiments on real databases validate the effectiveness of the proposed model and privacy metric method.

关 键 词：数据发布 隐私风险 风险度量 背景知识 

分 类 号：TM309[电气工程—电机]