Graph data anonymization method and device based on deep neural network, and storage medium

Abstract

The invention provides a graph data anonymization method and device based on a deep neural network and a storage medium. According to the method, an anonymous graph with high anonymity and high data availability can be generated for data release, and the method comprises the following steps: 1, sampling by using a random walk strategy of a node2vec algorithm to obtain a real node sequence in graphdata, and representing nodes in the node sequence as vectors in an One-hot manner; 2, constructing a learning model of graph data features based on a deep neural network, training the learning modelby taking a real node sequence as input, and optimizing model parameters to obtain a trained learning model; 3, inputting a real node sequence into the trained learning model, and outputting a randomwalk sequence of the simulated nodes; 4, adding noise meeting a differential privacy mechanism to the obtained random walk sequence of the simulated node, and then performing synthesis to obtain an anonymous graph.

Description

technical field [0001] The invention relates to the field of neural networks in machine learning, in particular to a method, device and storage medium for anonymizing graph data based on deep neural networks. Background technique [0002] Today, with the rapid development of Internet big data research and application, it is often necessary to share a large number of data sets between different companies, research institutions, etc., and even many data sets are provided to researchers in the form of public release to fully mine data in scientific research. , public services and commercial applications. The content of the data involves all aspects of social life, including transaction data, cooperative network data, social network data, location trajectory data, medical data, telephone communication data, commodity purchase data, etc. The most common of these data sets are graph data sets that exist in the form of networks. Nodes and edges contain rich user attributes and ass...

AI Technical Summary

Problems solved by technology

However, the potential feature dimension of graph data is very large, and the manually specified features are limited. Attackers can easily find potential correlation features by using advanced mining methods such as machine learning. In order to achieve a good anonymity effect, a large amount of noise information needs to be added, sacrificing usable value of data

For the anonymity of relational structures of graph-structured data such as social networks, existing anonymity technologies can only specify feature dimensions for privacy protection, and cannot resist multi-dimensional association attacks

Moreover, it is difficult to achieve a good trade-off between data anonymity and usability when existing technologies perform operations such as adding noise, feature generalization, and perturbation to the features in the graph.

That is, when anonymity is high, data availability is often low

Images