基于依赖差分隐私的高斯机制算法
Gaussian mechanism algorithm based on dependency differential privacy
欧阳恒 1陈洪超1
作者信息
- 1. 贵州轻工职业技术学院,贵州贵阳 550025
- 折叠
摘要
[目的/意义]差分隐私(DP)是一种数据隐私保护框架,能够保证查询结果在概率上不可区分.然而,研究表明,DP不适用具有元组相关性的数据集,元组间的相关性将带来隐私泄露的风险.[方法/过程]根据依赖差分隐私的定义,给出依赖差分隐私敏感度的计算方式,提出依赖差分隐私-高斯机制(Dependent Differential Privacy-Gaussian Mechanism,DDP-GM),实现关联数据集的隐私保护.[结果/结论]通过实验表明,DDP-GM在依赖数据的隐私保护方面有较高的可用性.
Abstract
[Purpose/Significance]differential privacy is a data privacy protection framework,which ensures that query results are indistinguishable in probability.However,the research shows that DP is not suitable for data sets with tuple correlation,and the correlation between tuples will bring the risk of privacy disclosure.[Method/Process]according to the definition of dependent differential privacy,this paper gives the calculation method of dependent differential privacy sensitivity,and proposes Dependent Differential Privacy Gaussian Mechanism to protect the privacy of correlated datasets.[Results/Conclusion]extensive experiments show that DDP-GM has high availability in data dependent privacy protection.
关键词
差分隐私/依赖差分隐私/高斯机制/关联数据集/数据隐私保护/数据安全治理Key words
differential privacy/dependence on differential privacy/gaussian mechanism/associated dataset/data privacy protection/data security governance引用本文复制引用
基金项目
贵州轻工职业技术学院院级课题(23QY16)
出版年
2024
NETL
NSTL
万方数据