匿名发布表质量度量问题是微观数据发布匿名模型中的重要内容之一。已有相关研究工作主要从准码属性取值层次变化幅度或泛化等价组中元组记录数角度定义匿名发布表质量度量方法,具有度量结果不精确的缺点。基于信息熵理论,根据泛化前后等价组中准码属性在不同层次取值包含的精确信息量变化情况,结合考虑具体数据分析任务对准码属性敏感程度不同因素为不同准码属性泛化路径设置权重,设计一组细粒度的匿名发布表隐私保护程度和信息损失程度度量方法。实验分析表明,利用该方法能够更加精确地度量泛化匿名表质量。
Measure criterion of published micro data is an important factor of anonymous models.Common measure criteria were studied in attributes hierarchy and record number view,which had obvious defect.Based on entropy theory,this paper designed a set of equation refers to quasi-attributes weight and domain to measure the utility of privacy preserving and information loss.Theoretical analysis and experimental results show the measure criteria is more accurate.