A differential privacy publishing method and system for associated tuple data

Abstract

The invention provides a differential privacy publishing method and system for associated tuple data, and relates to the fields of data mining and privacy protection. Initialization noise is first generated and initialized query results are disturbed, and then noise satisfying a specific auto-covariance matrix is ​​generated according to a new query function. And perturb the query results, and finally use an iterative mechanism to process all queries until the query function sequence is processed and output and the perturbed query results are released. The application of the generalized Laplacian noise generation method and the practical and efficient iterative and update mechanism provided by the present invention solves the problem that the tuple data to be protected is related and different in the existing publishing method using differential privacy to protect the associated tuple data. The problem of privacy protection strength reduction caused by the independence of privacy-generated noise.

Description

technical field [0001] The present invention relates to the fields of data mining and privacy protection, and more specifically, to a method and system for publishing differentially private associated tuple data, which is used to solve the problems faced by existing publishing methods for using differential privacy to protect associated tuple data. The protection of tuple data is correlated and the noise generated by differential privacy is independent, which leads to the reduction of privacy protection strength. Background technique [0002] In data-driven application scenarios, such as location-based services (LBS), disease monitoring, and social networking, tuple data is a widely-existing data type, and tuple data sharing is better for data owners. Service is very necessary. For example, in location-based applications, users upload their precise location data to service providers to obtain better navigation services; in disease monitoring applications, sharing personal b...

AI Technical Summary

Problems solved by technology

In these two models, the method based on the correlation model can offset the reduction of privacy by increasing the size of the noise, but larger noise will destroy the availability of data; although the method based on the transformation model overcomes the inefficiency of the method based on the correlation model With disadvantages, but the distribution after the inverse transformation is not in Laplacian form, so the transformation-based method may not strictly satisfy the definition of differential privacy

Therefore, the existing methods do not completely solve the problem of differential privacy protection for associated tuple data, and still face the problems of low data availability and the inability to strictly meet the definition of differential privacy

Images