Local high-order graph clustering method based on differential privacy

Abstract

The invention discloses a local high-order graph clustering method based on differential privacy. The method comprises the following specific steps: using the structure of a local high-order network subgraph Motif to convert an original social network into a Motif-based adjacency matrix, setting a certain threshold value for the weight in the generated Motif adjacency matrix based on the diversity of a Motif network subgraph structure, performing Laplacian noise disturbance on the weight within the threshold value range, and achieving privacy protection on a social network subgraph. In order to improve the operation efficiency of the random walk algorithm, an approximate hot kernel page ranking seed algorithm is adopted, random walk is conducted on the disturbed Motif matrix, the cut ratio of a divided set is calculated according to the walk hot kernel vector, and a cluster set is output.

Description

technical field [0001] The invention belongs to the technical field of data security, in particular to a local high-order graph clustering method based on differential privacy. Background technique [0002] With the rise of social media such as blogs and microblogs, social networks with users as nodes and user relationships as edges have grown rapidly. There are multiple communities or clusters in the social network due to the user's interests, behaviors, functions and other relationships. Its efficient clustering method enables users to understand the intimacy of their communities more efficiently, but there is a risk of violating user privacy during the clustering process. On the one hand, users worry that if too much content is included in the clustering process, their private information will be revealed; on the other hand, there are many uncertainties in the community division of social users based on the characteristics of network subgraphs, which limits their ability...

AI Technical Summary

Problems solved by technology

[0004] In view of the privacy protection problems in the above-mentioned prior art, the purpose of the present invention is to propose a local high-order graph clustering method based on differential privacy. The present invention adopts the structural characteristics of social network subgraphs and differential privacy to design a privacy protection The model, based on the random walk principle of thermonuclear page ranking, solves the clustering problem of local high-order graphs of complex social networks by limiting the number of steps of random walks; in the specific improvement, based on the structure of social network subgraphs, firstly The original social network is converted into a Motif weight matrix based on the network subgraph, and then the privacy protection strength of the Motif weight matrix is ​​determined through the constructed Motif weight matrix; finally, the approximate thermonuclear page ranking seed algorithm is used to perform the perturbed Motif weight matrix Random walk enables social networks to still have good clustering results under the premise of privacy protection

Images