Differential Privacy Based Data Privacy Protection And Its Application

In rapid growth of digital information,data have existed in all aspects of people' s lives and played important roles.The cyber-attacks against data have been serious problems.If the data are mishandled or stolen by attackers,it will cause serious privacy leakage.Therefore,the protection of data privacy is a prerequisite for the data applications of data.Data should be obtained prior to applying them.There are two ways to obtain data:(1)Obtaining data from the published dataset;and(2)Collecting data relevant to the research question.Therefore,this work we consider data privacy protection research from three aspects: data publishing,data collection and data mining.The main research work includes the following contents.(1)To exploit the potential value of data and at the same time to protect sensitive information,a method combining k-median clustering and differential privacy to publish mixed data(DP-k-median)is proposed.Firstly,the k-median clustering algorithm is used to reduce the sensitivity of the query function and to improve the availability of the published data.Secondly,the differential privacy is adopted to protect sensitive information.To improve the accuracy and availability of categorical attributes,a global combination of categorical attributes method is proposed to take the correlation among categorical attributes into account,and then apply the exponential mechanism.Theoretical analysis shows that the proposed method satisfies ?-differential privacy.Experimental results show that the proposed DP-k-median method has lower information loss and time overhead than the existing work for same parameters.(2)To enable researchers to obtain data that is more relevant to their research questions,a blockchain-based differential privacy data collection method(DPDCB)is proposed that ensures the legitimacy and fairness of the data collection,the privacy and high availability of data.Firstly,a new architecture of verification before collection is designed to ensure the legitimacy of the data collection and to improve data availability.Secondly,smart contracts are designed and combined with the non-repudiation of blockchain to achieve fairness in data collection.A reasonable reward and punishment mechanism is designed,so that workers are paid in proportion to their efforts.Finally,the DP-k-median method proposed previously is used to add noise to the collected data for the purpose of protecting the privacy of the data owners.Theoretical analysis shows that the proposed method ensures the legality and fairness of data collection while protecting the privacy of the data owner.Experimental results show that the data collection using this method has a low CPU usage and memory consumption,and the collected data offer highly availability.(3)A novel two-way privacy protection method is proposed to solve the problem in the existing medically assisted diagnosis systems which often pay attention to the accuracy of diagnosis,while ignoring the protection of privacy.The proposed method combines decision trees,differential privacy and oblivious transfer(OT)technology.Firstly,differential privacy is used to add noise to the decision tree to ensure database privacy.Secondly,the OT protocol is used to protect the two-way privacy of the client and the server in the query process.In order to effectively combine the decision tree algorithm with the OT technology,a decision tree indexing protocol is proposed that can efficiently digitize the decision tree.The proposed method is the first to apply decision tree and OT technology to a medically assisted diagnosis system.According to the comparison results of simulation experiments,the proposed method has higher query efficiency and query accuracy.