Research On Privacy Protection In Medical Data Analysis

With the popularization of medical information technology,the collection of medical data is more convenient and more massive.There is a trend of linear increase in the quantity and dimension of medical data,which makes the analysis and protection of medical data face great challenges.At present,the specialized service providers are responsible for the identification and analysis.Most of the similarity processing and abnormal behavior identification analysis in medical data are responsible by specialized service providers,and the service mode of online data processing brings risks of disclosure of patient privacy information.If such data and information security is not properly protected,so that criminals steal data and information,it will bring direct or indirect economic losses to medical institutions and patients.How to ensure the security and privacy of sensitive information while processing data is the primary problem in medical data analysis.In this paper,privacy protection schemes for medical data analysis is designed,and the existing problems in the current algorithm are solved.The main work is as follows:(1)In the analysis of medical fraud,aiming at the calculation process is vulnerable,the algorithm is sensitive to cluster merging times and it takes a long time to compare the distance with its points one by one and merging them.A differential local gravity patient privacy protection scheme is proposed.This scheme is composed of DPLG-HAC(Differential Privacy Local Gravity-Hierarchical Clustering)algorithm.On the basis of hierarchical clustering,differential privacy algorithm is combined to add noise to the original data,and local gravity algorithm is used to optimize hierarchical clustering process to reduce the impact of cluster merging times and speed up the judgment of abnormal patients.Experimental simulation shows that DPLG-HAC improves the accuracy of clustering and the privacy of data compared with other algorithms.(2)In doctors’ prescriptions analysis,aiming to the parameters for single clustering caused the result unstable,the reliability cannot be guaranteed and uncertain the number of neighborhood in VNS clustering initialization and merging cause the leak problem of privacy,a differential variable neighborhood privacy protection scheme is proposed.The scheme is composed of DPVNS-EAC(Differential Privacy Variable Neighborhood Search-Evidence assembled Clustering)algorithm,and the variable neighborhood search algorithm is applied to the EAC algorithm.The local optimal solution is determined by VNS to ensure the reliability of the algorithm,and the number of neighbors in the variable neighborhood is fundamentally determined according to the principle of EAC,and the laplace noise is added in the clustering iteration to ensure the clustering results and protect the privacy of doctors’ prescriptions.The simulation results show that the DPVNS-EAC algorithm improves the accuracy of the analysis results and the privacy of the data compared with other algorithms.(3)In health monitoring of high-dimensional data analysis,aiming to uneven clustering effect and easy be partitioned attack,traditional k-means algorithm to deal with high dimensional data high time complexity and density threshold algorithm and fixed boundary parameter sensitive problem.A differential artificial bee colony patient privacy protection scheme is proposed.The scheme by DPCLI-ABC(Differential Privacy Clique-Artificial Bee Colony Clustering)algorithm,the clique subspace clustering with adding noise in monitoring data of the patients,and through the artificial bee colony algorithm tuning,with the result of clustering in the neighborhood to choose the best food source,in order to reduce the density threshold and fixed boundary parameters on the impact of the results of the analysis.Experimental simulation shows that DPCLI-ABC algorithm improves the accuracy of clustering effect and the privacy of data compared with other high-dimensional clustering algorithms.