A Study On The Attribute-Correlated Differential Privacy Protection Mechanism

Internet plus era,network technology and big data have had a significant impact on all aspects of people's daily life.The user information involved in science,technology and product update is getting wider and wider.The risk of data exposure is further increased,causing the user's personal information to leak out easily.To find a balance of information sharing and privacy protection has become a hot research area for today's scholars.This paper is mainly a study of the problem of the privacy protection resulting from the attribute correlated data release.It points out that the present model has the flaw of large loss of data and poor availability of the data,through the analysis of the present situation of privacy protection method of data release.On this basis,the paper puts forward the attribute-correlated differential privacy model and related algorithms in order to enhance the utility of data.The content of this paper is as follows:(1)In the light of the attribute-correlated data release,the attribute-correlated differential privacy protection model is put forward to solve two problems,one of which is privacy leakage because of inference attack when the complex correlation data are released and the other of which is to solve the problem of destroying the utility of anonymous data because of adding excessive noises.(2)Two methods of attribute-correlated differential privacy data release algorithm are designed based on the idea of projection transformation.The addition of noises can be reduced by taking advantage of the relevance between attributes.The protection strength will gradually be strengthened and the usability of data release will also be improved by using the two algorithm methods.(3)Based on the census data set in the United States,the differential privacy protection model depending on attributes correction is validated.According to the experimental results,the method can greatly reduce data added noises and greatly enhance the usability of data under the same degree of privacy protection.