[1]钟可欣,杨 庚.函数回归的差分隐私保护算法[J].计算机技术与发展,2023,33(02):132-137.[doi:10. 3969 / j. issn. 1673-629X. 2023. 02. 020]
　ZHONG Ke-xin,YANG Geng.Differential Privacy Preservation Algorithm in Functional Regression[J].,2023,33(02):132-137.[doi:10. 3969 / j. issn. 1673-629X. 2023. 02. 020]
点击复制

函数回归的差分隐私保护算法()
分享到：

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
33
期数:
2023年02期
页码:
132-137
栏目:
网络空间安全
出版日期:
2023-02-10

文章信息/Info

Title:
Differential Privacy Preservation Algorithm in Functional Regression
文章编号:
1673-629X(2023)02-0132-06
作者:
钟可欣1 ; 杨 庚1; 2
1. 南京邮电大学 计算机学院,江苏 南京 210046;
2. 江苏省大数据安全与智能处理重点实验室,江苏 南京 210023
Author(s):
ZHONG Ke-xin1 ; YANG Geng1; 2
1. School of Computer,Nanjing University of Posts and Telecommunications,Nanjing 210046,China;
2. Jiangsu Province Key Laboratory of Big Data Security and Intelligent Processing,Nanjing 210023,China
关键词:
函数型数据分析; 差分隐私; 函数回归; 数据隐私保护; 隐私预算分配
Keywords:
functional data analysis; differential privacy; functional regression; data privacy preservation; privacy budget allocation
分类号:
TP309
DOI:
10. 3969 / j. issn. 1673-629X. 2023. 02. 020
摘要:
函数型数据回归是一种特殊的回归分析,其响应或协变量包含函数型数据,即样本元素为连续函数的数据。 函数型数据在医疗保健、社交媒体、传感网络等诸多领域都有重要应用,通常包含一些敏感信息,在回归分析的过程中,不加保护会引起隐私的泄露。 针对函数型数据回归分析中缺少隐私保护的问题,提出了一种基于拉普拉斯机制的函数回归的差分隐私保护算法。 首先,对响应数据进行降维,将响应函数建模为相互正交的 B 样条基的张量积,建立函数回归的数学模型;其次,对回归模型的未知参数取值使用惩罚最小二乘法估计,并通过正交基函数的数量控制粗糙度;最后,对估计参数加入服从拉普拉斯分布的噪声扰动,得到最终的回归结果。 理论分析和实验表明,函数回归的差分隐私保护算法满足拉普拉斯机制的差分隐私保护,并且随着隐私预算的减小,算法效率越高,在保证数据安全性的同时达到了良好的可用性。
Abstract:
Functional data regression is a special kind of regression analysis whose responses or covariates contain functional data,that is,data whose sample elements are continuous functions. Functional data has important applications in many fields such as health care,socialmedia,and sensor networks. It usually contains some sensitive information. In the process of regression analysis,if it is not protected,itwill cause privacy leakage. Aiming at the lack of privacy protection in functional data regression analysis,a differential privacy protectionalgorithm based on Laplacian mechanism for functional regression is proposed. Firstly,reduce the dimension of the response data,modelthe response function as a tensor product of mutually orthogonal B - spline bases, and establish a mathematical model of function　regression. Secondly, use the penalized least squares method for the unknown parameters of the regression model, and control theroughness through the number of orthogonal basis functions. Finally,add noise disturbance obeying Laplace distribution to the estimatedparameters to obtain the final regression result. Theoretical analysis and experiments show that the differential privacy protectionalgorithm of functional regression satisfies the differential privacy protection of the Laplacian mechanism,and with the reduction of theprivacy budget,the efficiency of the algorithm is higher,and it achieves excellent usability while ensuring data security.

相似文献/References:

[1]史武超,吴振强,刘海. 一种基于VCG机制的差分式隐私服务定价机制[J].计算机技术与发展,2017,27(06):119.
　SHI Wu-chao,WU Zhen-qiang,LIU Hai. A Pricing Mechanism of Differential Privacy Service with VCG Mechanism[J].,2017,27(02):119.
[2]高瑜[][],田丰[],吴振强[][]. 基于差分隐私保护的DPk-medoids聚类算法[J].计算机技术与发展,2017,27(10):117.
　GAO Yu[][],TIAN Feng[],WU Zhen-qiang[][]. A DPk-medoids Clustering Algorithm with Differential Privacy Protection[J].,2017,27(02):117.
[3]陈晓宇,韩斌,黄树成.基于差分隐私的数据匿名化隐私保护方法[J].计算机技术与发展,2018,28(07):99.[doi:10.3969/ j. issn.1673-629X.2018.07.021]
　CHEN Xiao-yu,HAN Bin,HUANG Shu-cheng.An Anonymized Data Privacy Protection Method Based on Differential Privacy[J].,2018,28(02):99.[doi:10.3969/ j. issn.1673-629X.2018.07.021]
[4]刘中锋.基于局部学习的差分隐私集成特征选择算法[J].计算机技术与发展,2018,28(10):79.[doi:10.3969/ j. issn.1673-629X.2018.10.016]
　LIU Zhong-feng.An Ensemble Feature Selection Algorithm with Differential Privacy Based on Local Learning[J].,2018,28(02):79.[doi:10.3969/ j. issn.1673-629X.2018.10.016]
[5]刘欢,吴桂兴.众包环境下的隐私保护研究[J].计算机技术与发展,2018,28(12):111.[doi:10．3969/j． issn．1673－629X．2018．12．024]
　LIU Huan,WU Guixing.Research on Privacy Protection in Crowdsourcing[J].,2018,28(02):111.[doi:10．3969/j． issn．1673－629X．2018．12．024]
[6]华雯丽,黄 刚,唐 震.一种融合差分隐私的随机游走算法[J].计算机技术与发展,2021,31(09):112.[doi:10. 3969 / j. issn. 1673-629X. 2021. 09. 019]
　HUA Wen-li,HUANG Gang,TANG Zhen.A PersonalRank Walk Algorithm Fusing Differential Privacy[J].,2021,31(02):112.[doi:10. 3969 / j. issn. 1673-629X. 2021. 09. 019]
[7]郑 剑,杨立聪.基于奇异值分解的大型社交网络差分隐私算法[J].计算机技术与发展,2022,32(03):126.[doi:10. 3969 / j. issn. 1673-629X. 2022. 03. 021]
　ZHENG Jian,YANG Li-cong.Differential Privacy Algorithm for Large Social Networks Based on Singular Value Decomposition[J].,2022,32(02):126.[doi:10. 3969 / j. issn. 1673-629X. 2022. 03. 021]
[8]徐 捷,杨 庚,白云璐.Top-k 频繁子图挖掘的差分隐私保护算法[J].计算机技术与发展,2022,32(05):80.[doi:10. 3969 / j. issn. 1673-629X. 2022. 05. 014]
　XU Jie,YANG Geng,BAI Yun-lu.A Differential Privacy Protection Algorithm for Mining Top-k Frequent Subgraphs[J].,2022,32(02):80.[doi:10. 3969 / j. issn. 1673-629X. 2022. 05. 014]
[9]李玉伟,杨 庚.满足差分隐私的一种频繁序列挖掘算法[J].计算机技术与发展,2022,32(05):99.[doi:10. 3969 / j. issn. 1673-629X. 2022. 05. 017]
　LI Yu-wei,YANG Geng.An Algorithm for Mining Frequent Sequence under Differential Privacy[J].,2022,32(02):99.[doi:10. 3969 / j. issn. 1673-629X. 2022. 05. 017]
[10]谢雅琪,杨庚.多项式回归的差分隐私保护算法[J].计算机技术与发展,2022,32(08):103.[doi:10. 3969 / j. issn. 1673-629X. 2022. 08. 017]
　XIE Ya-qi,YANG Geng.Differential Privacy Preservation in Polynomial Regression Analysis[J].,2022,32(02):103.[doi:10. 3969 / j. issn. 1673-629X. 2022. 08. 017]

更新日期/Last Update: 2023-02-10