Research And Application On Privacy Protection For Data Mining

The rapid development of science and technology has made people's lives more and more convenient.At the same time,the increasing amount of data has increased the difficulty of people to obtain information.The development of data mining technology is conducive to solve the problem of people's information acquisition.However,the wide application of data mining is seriously jeopardizing the protection of people's personal private data.While people are enjoying the services brought by this data mining,they are taking the risk of personal private leakage.Therefore,the proposed privacy protection technology for data mining can make people do not need to worry about the disclosure of their privacy in data mining.Privacy protection technology for data mining is an urgent issue to be studied.The main research works in this thesis are as follows:1.This thesis proposes a differential privacy protection method based on k-means.This method firstly removes the isolated point from the original data set.Secondly,determine the initial cluster center according to the average density of the data set.Thirdly,cluster the data set according to the determined initial cluster center.Finally,add the noise data to the processed data set.The data set processed by clustering anonymity can effectively reduce the sensitivity of query function,which can reduce the amount of Laplacian noise,so clustering anonymity with differential privacy protection can greatly improve the availability of data.2.Based on the DiffGen algorithm,this thesis proposes an improved DiffGen algorithm.The traditional DiffGen algorithm has an unreasonable allocation of privacy budget in the subdivision scheme,which leads to the premature consumption of the privacy budget in the process of DiffGen algorithm.In view of this problem,this thesis adopts the scheme to select the gambling split attribute,and changes the original average allocation privacy budget to adaptive allocation in the subdivision process,which is beneficial to increase the number of subdivisions and improve the algorithm execution efficiency.In addition,this thesis uses the Gini gain as the usability function in the algorithm,which effectively improves the efficiency of the usability function in the exponential mechanism.3.This thesis applies the privacy protection method for data mining to personalized recommendation.Personalized recommendation is also a special data mining method.This thesis combines the privacy protection method for data mining with the collaborative filtering algorithm,which shows the feasibility of the proposed method in practical application.This method enables people to enjoy personalized recommendation technology services while ensuring the security of personal privacy data.