基于可用性的数据噪声添加方法研究

doi:10.3969/j.issn.1671-1122.2024.11.012

摘要/Abstract

摘要：

随着信息技术的加速发展，数据隐私保护逐渐成为研究热点。如何有效保护个人隐私并充分利用各种数据资源，成为亟待解决的问题。利用数据加噪实现隐私保护是目前较常用的方法之一，但目前缺乏对不同加噪方法对数据可用性影响的研究。文章对反映用户体验的“五星制”评价数据分别添加拉普拉斯噪声和高斯噪声，比较分析加噪后数据的平均绝对误差、均方根误差、方差增长率3个统计指标的变化情况。然后探讨添加两种噪声比例变化的组合噪声以及不同数据量对数据统计指标的影响规律。实验结果表明，当数据量较大时，添加高斯噪声的数据比例越大，加噪后数据的统计性能与原始数据越接近，能在实现个人隐私保护的同时，更好地保证数据的可用性。

关键词: 隐私保护, 可用性, 评价数据, 拉普拉斯噪声, 高斯噪声

Abstract:

As information technology rapidly advances, data privacy protection has become a focal point of interest. Effectively safeguarding personal privacy while maximizing the utility of data resources is an urgent issue to address. Implementing data privacy protection through noise addition is one of the prevalent methods, yet research on the impact of various noise addition techniques on data usability is scarce. This study experimentally introduced laplace noise and gaussian noise to “five-star” rating data indicative of user experience. It compared and analyzed the alterations in three statistical metrics of the noise data: mean absolute error, root mean square error, and variance growth rate. The paper further investigated the effects of combining noise with different ratios and varying data volumes on statistical indicators. The experimental results indicate that with larger data volumes, a higher proportion of Gaussian noise addition results in statistical properties of the noise data that more closely resemble the original data. This approach ensures data usability while achieving personal privacy protection.

Key words: privacy protection, availability, evaluation data, Laplace noise, Gaussian noise

中图分类号:

TP309

顾海艳, 柳琪, 马卓, 朱涛, 钱汉伟. 基于可用性的数据噪声添加方法研究[J]. 信息网络安全, 2024, 24(11): 1731-1738.

GU Haiyan, LIU Qi, MA Zhuo, ZHU Tao, QIAN Hanwei. Research on Data Noise Addition Method Based on Availability[J]. Netinfo Security, 2024, 24(11): 1731-1738.

图/表 13

图1

图2

表1

图3

表2

图4

表3

图5

表4

表5

图6

表6

图7

参考文献 19

[1]	SHAO Guosong. Psycho-Graphic Profiling: What Can Be Learned from Cambridge Analytica[J]. Media Observer, 2021, 455(11): 45-49.
	邵国松. 心理绘图与用户画像:剑桥分析的兴亡启示录[J]. 传媒观察, 2021, 455(11): 45-49.
[2]	QIAN Wenjun, SHEN Qingni, WU Pengfei, et al. Research Progress on Privacy-Preserving Techniques in Big Data Computing Environment[J]. Chinese Journal of Computers, 2022, 45(4): 669-701.
	钱文君, 沈晴霓, 吴鹏飞, 等. 大数据计算环境下的隐私保护技术研究进展[J]. 计算机学报, 2022, 45(4): 669-701.
[3]	LI Hongtao, REN Xiaoyu, WANG Jie, et al. Continuous Location Privacy Protection Mechanism Based on Differential Privacy[J]. Journal on Communications, 2021, 42(8): 164-175. doi: 10.11959/j.issn.1000-436x.2021123
	李洪涛, 任晓宇, 王洁, 等. 基于差分隐私的连续位置隐私保护机制[J]. 通信学报, 2021, 42(8): 164-175. doi: 10.11959/j.issn.1000-436x.2021123
[4]	DWORK C. Differential Privacy[C]// Springer. International Colloquium on Automata, Languages, and Programming. Heidelberg: Springer, 2006: 1-12.
[5]	ZHANG Xuejun, YANG Haoying, LI Zhen, et al. Differentially Private Location Privacy-Preserving Scheme with Semantic Location[J]. Computer Science, 2021, 48(8): 300-308. doi: 10.11896/jsjkx.200900198
	张学军, 杨昊英, 李桢, 等. 融合语义位置的差分私有位置隐私保护方法[J]. 计算机科学, 2021, 48(8): 300-308. doi: 10.11896/jsjkx.200900198
[6]	HUANG Haiping, WANG Kai, TANG Xiong, et al. Differential Privacy Protection Scheme Based on Edge Betweenness Model[J]. Journal on Communications, 2019, 40(5): 88-97. doi: 10.11959/j.issn.1000-436x.2019095
	黄海平, 王凯, 汤雄, 等. 基于边介数模型的差分隐私保护方案[J]. 通信学报, 2019, 40(5): 88-97. doi: 10.11959/j.issn.1000-436x.2019095
[7]	WANG Li’e, LI Xiaocong, LIU Hongyi. News Recommendation Method with Knowledge Graph and Differential Privacy[J]. Journal of Computer Applications, 2022, 42(5): 1339-1346. doi: 10.11772/j.issn.1001-9081.2021030527
	王利娥, 李小聪, 刘红翼. 融合知识图谱和差分隐私的新闻推荐方法[J]. 计算机应用, 2022, 42(5): 1339-1346. doi: 10.11772/j.issn.1001-9081.2021030527
[8]	LIU Kai, HAN Yiliang, GUO Kaiyang, et al. Differential Privacy Trajectory Protection Mechanism Based on DBSCAN Clustering Algorithm[J]. Science Technology and Engineering, 2022, 22(25): 11091-11096.
	刘凯, 韩益亮, 郭凯阳, 等. 基于密度的噪声应用空间聚类算法的差分隐私轨迹保护机制[J]. 科学技术与工程, 2022, 22(25): 11091-11096.
[9]	WEI Kang, LI Jun, DING Ming, et al. Federated Learning with Differential Privacy: Algorithms and Performance Analysis[J]. IEEE Transactions on Information Forensics and Security, 2020, 15: 3454-3469.
[10]	SONG Cheng, CHENG Daochen, NI Shuiping. k Anonymous Trajectory Privacy Protection Scheme of Personalized Differential Privacy[J]. Journal of Beijing University of Posts and Telecommunications, 2023, 46(3): 109-114.
	宋成, 程道晨, 倪水平. 个性化差分隐私的k匿名轨迹隐私保护方案[J]. 北京邮电大学学报, 2023, 46(3): 109-114.
[11]	LI Kejia, HU Xuexian, CHEN Yue, et al. Differential Privacy Linear Regression Algorithm Based on Principal Component Analysis and Functional Mechanism[J]. Computer Science, 2023, 50(8): 342-351. doi: 10.11896/jsjkx.220800255
	李可佳, 胡学先, 陈越, 等. 基于主成分分析和函数机制的差分隐私线性回归算法[J]. 计算机科学, 2023, 50(8): 342-351. doi: 10.11896/jsjkx.220800255
[12]	XU Yahong, YANG Geng, BAI Yunlu, et al. Differential Privacy Data Publish Algorithm for Principal Component Analysis[J]. Cyberspace Security, 2018, 9(10): 74-82.
	徐亚红, 杨庚, 白云璐, 等. 面向主成分分析的差分隐私数据发布算法[J]. 网络空间安全, 2018, 9(10): 74-82.
[13]	YANG Jun, ZHOU Juling. Restore of Mathematical Detail: The Process of Gauss Deriving the Probability Density Function of Normal Distribution[J]. Journal of Statistics and Information, 2019, 34(6): 17-21.
	杨军, 周菊玲. 还原数学细节:高斯推导正态分布概率密度函数的过程[J]. 统计与信息论坛, 2019, 34(6): 17-21.
[14]	ROY I, SETTY S T V, KILZER A, et al. Airavat:Security and Privacy for MapReduce[C]// USENIX. Proceedings of the 7th USENIX Symposium on Networked Systems Design and Implementation. Berkeley: USENIX, 2010: 297-312.
[15]	BAO Ting, XU Lei, ZHU Liehuang, et al. Optimized Setting of Privacy Budget in a Recommendation System with Local Differential Privacy[J]. Scientia Sinica (Informationis), 2022, 52(8): 1481-1499.
	暴婷, 徐蕾, 祝烈煌, 等. 满足本地化差分隐私的推荐系统中隐私预算的优化设置[J]. 中国科学:信息科学, 2022, 52(8): 1481-1499.
[16]	YAN Yan, ZHANG Xiong, FENG Tao. Proportional Differential Privacy Budget Allocation Method for Partition and Publishing of Statistical Big Data[J]. Netinfo Security, 2022, 22(11): 24-35.
	晏燕, 张雄, 冯涛. 大数据统计划分发布的等比差分隐私预算分配方法[J]. 信息网络安全, 2022, 22(11): 24-35.
[17]	ZHANG Xing, ZHANG Xing, WANG Qingyang. DP-IMKP: Data Publishing Protection Method for Personalized Differential Privacy[J]. Computer Engineering and Applications, 2023, 59(10): 288-298. doi: 10.3778/j.issn.1002-8331.2201-0457
	张星, 张兴, 王晴阳. DP-IMKP:满足个性化差分隐私的数据发布保护方法[J]. 计算机工程与应用, 2023, 59(10): 288-298. doi: 10.3778/j.issn.1002-8331.2201-0457
[18]	WANG Chao. Personalized Recommendation Techniques and Applications Based on User Decision-Making Behaviors[D]. Hefei: University of Science and Technology of China, 2022.
	王超. 基于用户决策行为的个性化推荐方法及应用研究[D]. 合肥: 中国科学技术大学, 2022.
[19]	DWORK C, ROTH A. The Algorithmic Foundations of Differential Privacy[J]. Foundations and Trends in Theoretical Computer Science, 2013(9): 211-407.

用户编号	原始评分	添加噪声后的评分	添加噪声后数据的偏差
1	4	3.909	0.091
2	4	4.007	-0.007
3	4	4.194	-0.194
…	…	…	…
99	5	4.866	0.134
100	4	4.401	-0.401

用户编号	原始评分	添加噪声后的评分	添加噪声后数据的误差
1	4	4.048	-0.048
2	4	4.122	-0.122
3	4	4.074	-0.074
…	…	…	…
99	5	5.252	-0.252
100	4	3.979	0.021

数据类型	原始数据	添加拉普拉斯噪声后的数据	添加高斯噪声后的数据
平均值	4.340	4.357	4.346
标准差	0.790	0.784	0.800
均方差	0.624	0.615	0.640
平均绝对误差	—	0.126	0.095
均方根误差	—	0.171	0.117
方差增长率	—	1.44%	2.56%

数据类型	原始数据	添加拉普拉斯噪声的数据	添加高斯噪声的数据	添加组合噪声的数据
平均值	4.340	4.357	4.346	4.351
标准差	0.790	0.784	0.800	0.798
均方差	0.624	0.615	0.640	0.637
平均绝对误差	—	0.126	0.095	0.092
均方根误差	—	0.171	0.117	0.118
方差增长率	—	1.44%	2.56%	2.08%

添加高斯噪声的比例η	平均绝对误差	均方根误差	方差增长率
0	0.121	0.172	3.06%
5%	0.122	0.167	3.01%
10%	0.120	0.165	2.69%
15%	0.119	0.164	2.52%
20%	0.118	0.162	3.04%
25%	0.118	0.162	4.48%
30%	0.115	0.159	5.94%
35%	0.115	0.157	5.41%
40%	0.116	0.158	5.70%
45%	0.112	0.152	2.19%
50%	0.113	0.153	2.80%
55%	0.108	0.149	4.07%
60%	0.106	0.146	3.13%
65%	0.100	0.136	1.97%
70%	0.097	0.132	1.54%
75%	0.099	0.134	1.83%
80%	0.096	0.125	0.47%
85%	0.100	0.128	0.59%
90%	0.100	0.127	0.91%
95%	0.096	0.120	1.60%
100%	0.090	0.117	0.96%