Research On Privacy Protection Method Based On Local Data

With the rapid development of the mobile technology,the risk of personal privacy data disclosure will increase when the user use the service of the LBS.The localized differential privacy that has emerged in recent years has been widely used to protect sensitive personal information of users because it inherits the centralized privacy protection method.At present,there are three methods for publishing data based on local differential privacy,namely data perturbation,data compression and data distortion.The RAPPOR method,Randomized Aggregatable Privacy-Preserving Ordinal Response,is the common method,which is more applicable to the crowdsourcing application.Because the RAPPOR method has certain limitations,it cannot meet the data analysis and mining work in some cases.At the same time,due to the limitations of its own applicable conditions,RAPPOR is increasingly unable to meet the complex query analysis needs.It will not only affect the availability of data,but also cause users to distrust the data analyst.Our goal is to share meaningful data while ensuring the privacy needs of our users.Based on this,we propose a privacy protection method based on the local variable user data set.This method can perform data mining and analysis on a changeable user set under the requirement of localized differential privacy.We designed a new release method ELRAPPOR based on the traditional RAPPOR method.First,under the premise of satisfying the local differential privacy protection,a new data distribution method based on variable user set is proposed.ELRAPPOR uses a method of extending the fixed number of user sets of the traditional RAPPOR method.Second,the concept of feature string is proposed.Through the user's feature string set information,perform classification analysis and restoration of incremental set data on variable user data sets,and analyze user information of incremental set users in detail.Third,According to the analysis of the incremental set we implemented,we can construct a new user set.On this new set of users,we perform frequent pattern mining,mainly k-frequent item data analysis.Compared with the traditional local differential privacy protection method,our method first breaks through the limitations of the immutable number of users,and our method can also achieve more complex data analysis work,and has better results.Finally,the validity of the ELRAPPOR method was verified by experimental results.