摘要
在图数据发布过程中,应用节点差分隐私约束能够有效保障用户的隐私信息安全。节点差分隐私下图数据发布的挑战主要来自查询的高敏感度,为此提出一种基于紧密度和边介数排序的边添加投影方法(EACB),基于节点度和紧密度得到有序节点集,结合边介数对边序列排序,最后基于边序列生成图。在此基础上提出一种基于K-means++聚类的直方图发布方法(EACB_Hist)进行差分隐私扰动发布。通过模拟仿真实验对比,基于EACB生成的图比原始图误差更低,有效降低全局敏感度,更大程度地保留原始图信息。度直方图发布实验结果表明,该方法在保证隐私保护性的同时比其他方法更接近原始图的度分布,提高发布数据的可用性和准确性。
In the process of publishing graph data,applying node differential privacy constraints can effectively ensure the security of users'privacy information.The challenge of node differential privacy in graph data publishing mainly comes from the high sensitivity of queries.Therefore,an Edge Addition Projection Method(EACB)based on compactness and edge betweenness centrality sorting is proposed.This method obtains an ordered set of nodes based on node compactness and closeness,sorts the edge sequence based on edge betweenness centrality,and finally generates a graph based on the edge sequence.On this basis,a histogram publishing method based on K-means++clustering(EACB_Hist)is proposed for differential privacy perturbation publishing.Through simulation experiments,it was found that the images generated based on the EACB method have lower errors compared to the original images,effectively reducing global sensitivity and preserving the original image information to a greater extent.The experimental results of the degree histogram publication show that this method is closer to the degree distribution of the original graph than other methods while ensuring privacy protection,improving the usability and accuracy of the published data.
作者
马静文
李晓晔
孙振龙
MA Jing-wen;LI Xiao-ye;SUN Zhen-ong(College of Computer and Control Engineering,Qiqihar University,Heilongjiang Qiqihar 161006,China)
出处
《齐齐哈尔大学学报(自然科学版)》
2025年第5期24-32,共9页
Journal of Qiqihar University(Natural Science Edition)
基金
黑龙江省省属高等学校基本科研业务费科研项目(145209124)。
关键词
隐私保护
图数据
节点差分隐私
度分布
直方图发布
privacy preserving
graph data
node differential privacy
degree distribution
histogram publishing
