Research On The Interactive Differential Privacy Protection Model And Algorithm

With the development of information technological,open sharing of data is a key part of data analysis applications.However,diverse data mining techniques and attack methods have impacted the strength of privacy protection algorithms.If the privacy level of the privacy protection mechanism can be estimated,it can effectively provide data publishers with a basis for decision-making.In addition,high privacy protection leads to low data utility,so solving the balance between privacy and utility in the privacy protection mechanism is crucial.Based on the interactive framework,this paper discusses the interactive differential privacy protection model and its quantitative model research,aiming to solve the problems of privacy leakage and data utility quantification in the interactive differential privacy protection mechanism.First,based on the information entropy measurement model,an interactive differential privacy measurement model based on graph theory and mutual information is proposed;further,under the interactive differential privacy and utility measurement model,the overall data utility of the published data set is quantified based on Hamming distance;Considering the balance between privacy and utility,the balance optimization problem of privacy and utility in the interactive differential privacy protection mechanism is studied.The specific research work is as follows:(1)An interactive differential privacy measurement model based on graph theory and mutual information is proposed.Based on the information entropy measurement model,the interactive differential privacy protection framework is reconstructed,and the amount of mutual information is introduced to quantify the privacy leakage between the published data and the original data.Further,a method for calculating the maximum amount of mutual information privacy leakage is proposed based on the Hamming diagram.Experiments show that the metric calculation method has fewer constraints and holds true for any original data set probability distribution.(2)An interactive differential privacy data utility measurement model based on Hamming distance is proposed.An information theory framework based on differential privacy,introducing the Hamming distance to quantify the data utility between the actual results and the published results;based on the structural properties of the Hamming graph,considering the upper bound of the data utility to solve the problem,analysis and comparison show that the interaction based on Hamming distance The differential privacy utility measurement model can effectively quantify the data utility in the differential privacy protection mechanism.(3)Based on the above research content,considering the balance of data and utility in the interactive differential privacy protection mechanism,the privacy and utility optimization problem is described based on the KKT condition,and the most effective use of mutual information privacy leakage under constraint is solved based on the Blahut iterative idea.Through experimental and comparison analysis,the model of trade off between utility and privacy can effectively solve the balance optimization problem of privacy and utility in the interactive differential privacy data publishing scenario.