Research On Intelligent Recognition Algorithm And Adaptive Protection Model Of Sensitive Data

In the stage of data opening and sharing,it has become a consensus to analyze and apply of data.However,the data release process is usually accompanied by privacy disclosure.How to automatically identify and classify sensitive attributes in structured data sets is still a difficult issue in privacy protection.Based on the measurement of attribute sensitivity and the relationship between attributes,this paper discusses the intelligent identification algorithm and adaptive privacy protection scheme of sensitive data.Firstly,use the relevant definition of entropy to quantify the sensitivity of attributes,and sensitive attributes are identified through clustering of sensitivity and association rules between attributes;Secondly,by mining the explicit mutual information correlation and implicit association between sensitive attributes,the sensitive attributes are classified,under the framework of fuzzy comprehensive evaluation method,the sensitive attributes can be graded;Finally,considering the difference of attribute sensitivity,a noise adding method based on attribute sensitivity is proposed,so as to propose an adaptive privacy protection scheme for sensitive attributes.The specific research work is as follows:(1)An intelligent identification algorithm for sensitive attributes of structured data sets is proposed.Introduce information entropy and maximum discrete entropy as means to quantify attribute sensitivity,and use k-means algorithm to cluster analysis of attribute sensitivity,further use the Aprior algorithm to mine association rules between attributes to realize the identification of sensitive attributes.Experiments show that the algorithm can identify the sensitive attributes of any structured data set without predicting the characteristics of the attributes,and the efficiency and accuracy are higher than other algorithms.(2)A classification and grading method for sensitive attributes of structured datasets is proposed.Introduce mutual information to measure the correlation between sensitive attributes,and at the same time consider the correlation between sensitive attributes to realize the classification of sensitive attributes;under the framework of fuzzy comprehensive evaluation method,the sensitive level of attributes is evaluated.Comparative analysis shows that this method takes into account the correlation and association between attributes,and can better solve the classification problem of sensitive attributes.(3)Based on the above research results,an adaptive privacy protection scheme with sensitive attributes is designed.This scheme is based on the research of the Laplace noise-adding mechanism in differential privacy,according to the difference in the sensitivity of attributes in the data set,reasonable noise is assigned to different types of sensitive attributes,thus an adaptive privacy protection scheme for sensitive attributes is proposed.Through experimental comparison and analysis,the scheme is better than the traditional differential privacy noise addition method in terms of data utility and privacy protection.This scheme has more advantages than others.