大数据环境中交互式查询差分隐私保护模型

Interactive queries differential privacy protection model in big data environment

随着大数据时代的到来,数据挖掘技术被广泛应用,而线性查询作为该技术中最基础和最频繁的操作,其隐私保护在数据分析和数据发布隐私保护中占有极其重要的位置。交互式线性查询的交互增加了数据的处理量,运用传统的隐私保护模型效率较低。为了解决大数据环境中交互式查询差分隐私保护问题,模型针对大规模数据集中交互式线性查询差分隐私保护的特点,通过数据关联性分析减少冗余信息,采用交替方向乘子法对查询负载矩阵进行分解,利用自适应加噪技术产生差分隐私保护所需要的合理数量的噪声,设计并行处理方法实现该模型的计算。实验将提出的模型与以往模型进行对比,结果表明所提出的模型在提升隐私保护精度的同时也极大地提高了算法性能,因此模型切实可行。

With the arrival of the era of big data,data mining technology is widely used,and the most basic and frequent ope- ration of the technology,linear query,whose privacy protection occupies an extremely important position in data analysis and data release privacy protection.The data processed become more when querying in an interactive linear queries way,and it is less efficient when using the traditional privacy protection models.In order to solve the problem of differential privacy protection for interactive queries in big data environment,the model reduced the redundant information through data correlation analysis,decomposed the query load matrix by adopting alternating direction multiplier method,produced a reasonable amount of noise required for differential privacy protection using the adaptive noise injection technology,and designed a parallel processing method calculated it against the characteristics of interactive linear query differential privacy protection for large-scale data set.Experiment compared the model proposed to previous works.The result shows that the proposed model promotes the accuracy of privacy protection and algorithm performance greatly.Therefore,the model is feasible.