Research On Privacy Protection Technology Of Data Publishing Based On Anonymization

With the development of information technology,the data have seen explosive growth.People collect and analyze these data to obtain useful information for themselves and enjoy the benefits brought by big data.Personal privacy,however,faces huge challenges at the same time.Since many of the collected data contain personal sensitive information,it is highly likely that personal privacy will be revealed if these data are not processed.Therefore,how to retain certain availability of data under the premise of protecting data privacy is a research hotspot in the field of data publishing at present.Compared with other privacy protection technologies,anonymization technology,with the characteristics of high degree of protection and low degree of information loss,can achieve a balance between data security and availability so that it has attracted great attention from researchers.Based on anonymization technology,this thesis studies the privacy protection of data with single-sensitive and multi-sensitive attributes.The main work of this thesis are as follows:(1)Summarize the current research status of anonymization technology at home and abroad.By comparing other privacy protection technologies,this thesis demonstrates the superiority of anonymization.It also introduces some related technologies and some familiar anonymization models for different attack types.(2)Aiming at the similar attacks and personalized anonymity problems of single sensitive attribute data,According to the sensitivity of sensitive values,this thesis sets frequency constraints for different sensitive values to limit the probability of their occurrences in the equivalence class,so as to achieve personalized anonymity requirement.At the same time,the semantic relationship among sensitive values is considered through the semantic hierarchy tree,and the number of the same semantic sensitive values is limited in the equivalence class to resist similarity attacks of sensitive attributes.the safety of this method is analyzed,and experiments are conducted to compare it with other methods.(3)Aiming at the problem of similar attacks and association attacks on multi-sensitive attribute data,data mining algorithms are used to mine association rules betweenquasi-identifier attributes and sensitive attributes to determine their association.Divide the unrelated quasi-identifier attributes separately,and they do not need to be considered when generalizing and clustering,which can reduce the information loss of data.At the same time,sensitive attributes are divided into different attribute groups,which solves the problem of associative attacks among sensitive attributes.And establish a multi-dimensional semantic bucket by analyzing the semantics of each sensitive attribute value to limit the number of the same semantic sensitive value of each attribute in the equivalence class,so as to resist similarity attacks of sensitive attributes.the safety of this method is analyzed,and experiments are conducted to compare it with other methods.The innovations of this thesis are as follows:(1)A personalized anonymity method that can resist similarity attacks is proposed.This method not only meets the requirement of personalized anonymity,but solves the problem of similarity attacks among sensitive attributes.In the process of dividing equivalence classes,the idea of clustering is used to reduce the information loss of data in the course of anonymity.Experimental results and security analysis show that this method has low information loss and high security,and can resist more types of attacks.(2)A privacy protection method of multi-sensitive attributes is proposed based on association rule slicing.This method not only solves the problem of association attacks among sensitive attributes,but also reduces the information loss of anonymous data by dividing attributes through association rules.At the same time,this method also analyzes similarity attacks of multi-sensitive attributes,which can resist more types of attacks.Experimental results show that this method achieves the purpose of privacy protection with lower information loss.