Publishing Correlated Social network Data with Differential Privacy
Abstract— The application of social network collects a large amount of user data and sensitive data, which
may reveal potential privacy information through analysis. At present, the differential privacy protection
model gives a rigorous and quantitative representation and proof of privacy disclosure risk, which greatly
guarantees the availability of data. Recently, differential privacy is very popular. However, differential privacy
assumes that data sets are independent. In real life, few data sets are completely independent. In social
networks, nodes have edges that are related. This paper proposes a solution to use differential privacy on
correlation social network data and designs a mechanism for correlation social network data publishing.
Reduce the large amount of noise added when the association graph data is published with differential privacy.
Considering the degree of correlation between nodes in the graph data, this paper first proposed the degree of
association between nodes and calculated the degree of association between each node to calculate the
sensitivity of association. The correlation sensitivity is used to determine the noise level in the implementation
of differential privacy. Then the hierarchical random graph model is used to add noise satisfying differential
privacy to edge connection probability to generate the pending layout. Finally, the feasibility and effectiveness
of the method are verified by the statistical analysis of degree distribution, aggregation coefficient and onedimensional
Index Terms— Differential privacy, correlation data, social networks, publishing graph generation model
Siyu Li, Dongran Yu, Xuebo Han, Jie Li, Peng Liu, Xianxian Li
Guangxi Key Lab of Multi-source Information Mining & Security, Guangxi Normal University, CHINA
Cite: Siyu Li, Dongran Yu, Xuebo Han, Jie Li, Peng Liu, Xianxian Li, "Publishing Correlated Social network Data with Differential Privacy," Proceedings of 2019 the 9th International Workshop on Computer Science and Engineering, pp. 767-775, Hong Kong, 15-17 June, 2019.