深入敏感值评估的隐私度量分级模型  

Comprehensive privacy metric grading model for in-depth sensitivity assessment

作　　者：谢思琪 田秀霞 XIE Si-qi;TIAN Xiu-xia(School of Computer Science and Technology,Shanghai University of Electric Power,Shanghai 201306,China)

机构地区：[1]上海电力大学计算机科学与技术学院,上海201306

出　　处：《计算机工程与设计》2025年第3期756-761,共6页Computer Engineering and Design

基　　金：国家自然科学基金面上基金项目(61772327);上海市大数据管理系统工程研究中心开放课题基金项目(H2020-216);2022年CCF-华为胡杨林基金-数据库专项基金项目(CCF-HuaweiDB202209);国网甘肃省电力公司电力科学研究院基金项目(H2019-275)。

摘　　要：为对数据的隐私等级进行细致化评定,探求同一类型敏感属性下不同敏感值之间的差异,提出一种深入敏感值评估的隐私度量分级模型。以信息熵为度量,通过考虑敏感值的泛化程度、语义敏感度、句法结构3方面的影响,对数据集中各个敏感因素进行定权计算得到记录的隐私度量向量,输入到分类器中进行分级。实验结果表明,该模型在结构化数据集和短文本数据集中都能实现无需预定义权重的敏感数据分级,准确度分别为94.17%和87.84%。To intricately evaluate data privacy levels,investigate distinctions among various sensitivity values within the same category of sensitive attributes,a comprehensive privacy metric grading model for in-depth sensitivity assessment was proposed.The information entropy was utilized as a metric and the influences of generalization degree,semantic sensitivity,and syntactic structure on sensitivity values were considered,a privacy measurement vector for records was derived through the weighted calculation of various sensitive factors in the dataset.This vector was inputted into a classifier for grading.Experimental results demonstrate that the proposed model achieves accurate sensitive data grading without the need for predefined weights in both structured datasets(94.17% accuracy) and short text datasets(87.84% accuracy).

关 键 词：敏感值评估 数据分级 信息熵 敏感信息识别 敏感数据管理 信息安全 依存句法分析 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]