Research On Privacy-Preserving Anonymous Technologies Of Electronic Medical Records Publishing

Abstract:With the development of the society, the amount of health information is exponentially increasing. However, this information is distributed across multiple sites, which held in a variety of paper and electronic formats, and represented as mixtures of narrative and structured data. Electronic Medical Records have been introduced as a method for improving communication among health care providers and improving access to patient data. This application of EMRs has now enabled large and complicated databases of health records to be used for medical and other researches. However, as medical record data become more accessible, protecting patients’ privacy is an increasing concern that should not be overlooked or understated. Since EMRs include patients’ private data, the access should be restricted by researchers. In this thesis, we consider the Electronic Medical Records as a specific data object, and we analyze and research the current anonymization techniques in the process of data distribution. The main research contents and contribution are as follows:Firstly, the preservation of privacy data in Electronic Medical Records and the shortcomings of K-Anonymity model and L-Diversity model are discussed, and then an enhanced privacy model for Electronic Medical Records which combines the advantages of two models is proposed. The new model based on clustering method and a reasonable anonymization cost metric can strengthen the security and availability of the medical data. The experiment results show that the proposed model can not only minimize information loss but also improve the ability to resistant the leakage risks.Then, for achieving the different privacy preservation requirements of each individual, a personalized privacy algorithm for Electronic Medical Records is proposed. Considering the sensitive differences among sensitive attributes, the new algorithm adopts classification of the sensitive attributes to avoid the fact that the same sensitive attribute values of the same sensitivity level appear in the same equivalent groups. At the same time, through the sensitive attribute generalization, it allows the patient to control the protection of the sensitive attributes. The experimental results show that the proposed algorithm can resist privacy attack and is an effective approach. There are7figures,13tables and62references in this thesis.