Research On Privacy Protection Methods For Sensitive Attributes In Data Publishing

With the rapid development of the network information,data publishing has made data mining and information sharing more convenient.Though,it has provided great convenience for people from all walks of life,it is accompanied by a large number of sensitive data leakage problems.How to make the published data both highly available and protect the sensitive information from being leaked,this issue has become an important issue in the field of data publishing privacy protection.This paper focuses on the privacy protection methods of publishing sensitive attribute data.The main work and innovation are as follows:(1)Research on privacy protection model and privacy attack methods in data publishing.Firstly,analyze common attacks that cause private data leakage and illustrate them with examples.Secondly,research and summary on existing typical privacy protection method models and application scenarios,and analysis on the advantages and disadvantages of these models.Thirdly,conclusion on relevant principles of information metrics in privacy protection methods.(2)Research on Privacy Protection Method of Single-Dimensional Sensitive Attribute(SSA).as for the problem of privacy leakage of single-dimensional sensitive attributes caused by similarity attacks,(SD,K)-anonymity model for simultaneous protection of sensitive values and sensitivity is given in this paper.This model introduces the concept of sensitive property measure,and combines the sensitive knowledge of classification tree to measure the sensitivity.the degree of similarity between sensible values is quantified,and the quantified values are used to classify the degree of sensitive attributes.Finally,the attribute values in equivalence classes contain not only the diversity of sensitive values but also the diversity of sensitivity levels.In addition,a corresponding algorithm is designed based on this model.Experiments show that the proposed single-dimensional sensitive attribute protection model can reduce the number of sensitive level attack records,reduce the risk of privacy leakage,balance a small amount of information loss and execution time issues,and better protect the user's sensitive attributes.(3)Research on Privacy Protection Method of Multidimensional Sensitive Attribute(MSA).In order to solve the problem of chained attack leakage caused by the mapping relationship between multi-dimensional sensitive attributes,this paper establishes(LI,L2)-diversity model with semi-constraint of primary sensitive attributes.The model introduces statistical ideas and statistically publishes sensitive values corresponding to primary sensitive attributes in equivalence classes.To cut off one-to-one correspondence with other dimensionally sensitive attributes.Combined with the idea of multi-dimensionalbuckets,we propose a candidate set of primary sensitive attributes and design an improved implementation algorithm MP+HC(L1,L2)-diversity.Experiments show that compared with the multi-dimensional bucket grouping algorithm,the algorithm has a low hiding rate and a small increase in time consumption.On the whole,this method effectively protects the sensitive information of multiple sensitive attributes and improves the data validity at the expense of less time.(4)Design of anonymized Data Publishing System for Electronic Medical Records.As for the privacy and security problems of electronic medical record data,an electronic data record anonymization data publishing system was designed based on the algorithm proposed in this paper.