Research On Publishing And Management Of Multiple Correlated Data Based On Differential Privacy

With the development of the era of big data,personal data collection becomes more and more convenient.On one hand,data brings people a convenient and efficient life,but on the other hand,there also exists more risks of privacy leakage.In nowadays post-NCP era,people's locations and social network relationships are no longer individual privacy due to the needs of prevention and control of NCP.In this case,data holders must carry out strict data protection methods to avoid the risk of privacy leakage during daily data collection and publishing.Especially for multiple correlated data like personal trajectory data,due to the existence of association inference,the multiple correlation contained in the data has a greater risk of privacy leakage.Therefore,exploring the release and management of multiple correlated data under privacy protection is of great significance to personal privacy protection under the daily situation of NCP prevention and control.In recent years,the issue of privacy protection has continued to attract the attention of many sides.Thanks to its strict mathematical definition,differential privacy stands out among privacy protection models and methods.It mainly realizes privacy protection by adding noise to disturb the original data and retain the data statistical characteristics to guarantee the data utility.Some applications have already been obtained in private data publishing.However,from the perspective of concrete application,the current differential privacy research lacks a unified standard for setting the privacy level to publish data,which is the pain point of differential privacy from theory to application,and most of the existing correlated differential privacy research assumes that there is only one correlation in the data,which may cause unexpected privacy leakage during data release.What's more,many query attacks on differential privacy are emerging,indicating the loophole of data protection.To solve the existing problems of differential privacy mentioned above,this article will focus on optimizing the application of differential privacy in the release among multiple correlated data to realize multiple correlated data publishing under differential privacy.By exploring the relationship between the privacy level and the characteristics of noise distribution,social privacy preferences and other factors,the privacy protection parameter can be determined by proposing a selection model.A data perturbation mechanism is designed by quantifying and constraining the privacy leakage of multiple correlated data under differentially private release.Then a basic framework for interactive data publishing is established.What's more,in order to deal with potential attacks,malicious query detection layer is added to optimize the basic framework,by proposing a malicious query detection model.Finally,experiments based on real data sets are conducted,and the results show that the multiple correlated data publishing framework proposed in this article has achieved both data privacy and high data utility,which can meet the target privacy level,avoid unexpected privacy leakage,and enhance the data protection level by adding the detection layer.This article includes four innovations.1)It is the first combination of attack model and economic model in the research of privacy protection parameter choice,among which,data value is utilized to quantify social privacy preference to set the attack success probability threshold in the attack model;2)Expand the existing attack model to establish a DP-Geometric attack model that is more suitable for the release of multiple correlated data;3)Combine and optimize the existing correlation differential privacy mechanism,and the multiple correlated differential privacy MCDP is proposed to solve the limitations of differential privacy in the multiple correlated data publishing;4)From the perspective of active defense data management,this article solves the potential privacy vulnerabilities of differential privacy by adding a query detection layer,and provides new ideas for the existing research that only optimize noise mechanism to resist attacks.These not only contribute new research ideas and methods for the publishing and management of protecting multiple correlated privacy data,but also provide a theory and basis with reference value for my country to promote and improve the specific implementation plan of personal privacy protection.