Research On Privacy Protection Data Publishing Algorithm Based On Cloud Computing

While driving the rapid development of all industries,big data cloud computing also brings serious challenges to personal privacy security.In order to better promote the rapid development of various industries,more and more data needs to be shared,the data which contained some sensitive privacy information,if the information which was directed release without any protection will result in the disclosure of private information,threatened the life property safety of the data owner,even threat the information security of the country,so the research on data privacy protection technology becomes very important.At present,privacy protection methods for data publishing are mainly based on disruption,encryption and anonymity.This paper analyzed the existing privacy protection model based on anonymous technology,and proposed an anonymous privacy protection model for single sensitive attributes based on the number of sensitive attributes--(?i,k,?)-anonymity protection model and an anonymous privacy protection model for multi-sensitive attributes--Multi-(?i,k,?)-anonymity protection model.The main research contents of this paper are as follows:1.This paper analyzed and compared the three mainstream models of data privacy protection,considered the advantages and disadvantages of these three mainstream models,chose the anonymity protection model which took both practicality and security as the research focus,and introduced the research status and related basic knowledge of the model in detail.2.The anonymous protection model of single sensitive attribute is studied.The advantage of the(?i,k)-anonymous model is that it can resist homogeneous attack,but it has the risk of semantic similarity attack.In order to resist this risk,a(?i,k,?)-anonymous model was proposed to resist homogeneous and similarity attacks.The model classified the values of sensitive attributes and calculated the constraint ?i of each level according to the distribution of sensitive attributes in the data set.Through semantic analysis of sensitive attribute values,the semantic hierarchy tree of sensitive attribute is constructed and equivalence class dissimilarity constraint is introduced to resist similarity attack.At the same time,in order to preserve the correlation between attributes and reduce the generalization loss as much as possible,analytic hierarchy process is applied to calculate the correlation between quasi-identifier attributes and sensitive attributes,and the correlation between attributes is applied to the generalization loss solution to reduce the generalization loss of information in the anonymous process.The simulation results showed that the model can effectively reduce generalization loss,resist multiple attack models and protect data privacy.3.The anonymous protection model of multiple sensitive attributes is studied.In the process of data publishing,multiple sensitive attributes often exist together.Therefore,the(?i,k,?)-anonymous protection model is improved and a Multi-(?i,k,?)-anonymous protection model is proposed,which is suitable for the privacy protection of multiple sensitive attributes.In order to reduce generalization loss better,this model uses information gaining method to reduce dimensionality of multi-sensitive attributes,which can not only preserve the correlation between sensitive attributes,but also reduce generalization loss.At the same time,considering the existing in various attributes of numeric attribute problem,put forward the numeric attributes into classification attribute,the method of classification of attributes instead of numerical attributes was applied to calculating correlation between attributes,but in the equivalence class generalization when using numeric attribute generalization,still retains the correlation between numerical properties and classification,and will not increase the generalization.The simulation results showed that this model is applicable to the privacy protection of multiple sensitive attributes,can well retain the correlation between attributes,has a small degree of generalization,can resist various attack models,and has a high degree of data privacy protection.