Differential Privacy Preservation Method for Linked Social Network Data

Abstract

The invention discloses a differential privacy protection method for associated social network data. On the basis of analyzing the characteristics of the social network, the original social network graph is classified into multiple subgraphs according to the node labels, so as to ensure that the distribution of the inner edges of the subgraphs is relatively concentrated, and the subgraphs The connections between the edges are weak. Secondly, for each subgraph after classification, use the hierarchical random graph to infer its structure, and add noise that satisfies differential privacy protection to the node probability. Finally, construct the subgraph to be released according to the hierarchical random graph, and then reorganize it into a graph to be released. The invention can solve the impact of the correlation between the social network data on the differential privacy protection, thereby greatly improving the security of the published social network graph.

Description

Technical field [0001] The present invention relates to the field of social network privacy protection technology, which involves a differential privacy protection method of associated social network data. Background technique [0002] Social networks collect a large number of user data and sensitive data while helping people build social network application services. Direct analysis of social network data will cause sensitive information disclosure to make a threat to user privacy. Traditional data-free privacy protection technology seems to be in the face of continuous improved background attacks. In this regard, differential privacy is in the privacy protection that can be introduced into a social network as a rigor-defined magic technology. However, existing differential privacy methods are based on all records independently assumptions, while in the context of network data, this assumption is not always established. Given that a large number of social network data is associa...

AI Technical Summary

Problems solved by technology

However, existing differential privacy methods are based on the assumption that all records are independent of each other, which does not always hold in the context of network data.

Given that a large amount of social network data is correlated, recent studies have shown that differential privacy is susceptible to data correlation, which hinders the application of differential privacy to correlated network data

Images