A new approach for ensuring medical data privacy using neural networks

The foremost intention of data mining is to extort unfamiliar patterns from huge set of data. There is a marvellous escalation in the quantity of data anthology with the rapid growth in technology. In recent years, with the advancement in information technology many hospitals started storing various information related to patients in the electronic format commonly referred as Electronic Health Records (EHRs). If used rightly EHRs can precisely identify diseases and add to the effectiveness of healthcare delivery. Quality of treatment given to the patients can be still improved if these records are shared among various hospitals. The key obstruction in data sharing is the sensitive medical information existing in the EHR. The revelation of sensitive data results in privacy breach of individuals, which results in the usage of many privacy preserving techniques. This work is provoked with the necessity to mutually guard private information and facilitate its use for research purpose. In this paper we present a new approach for generating noise to preserve privacy using a neural network. The pseudo identifiers in the data set (EHRs) are used as the input to the network nodes. Output generated by the neural network is added with the original data to generate the sanitized data. We have compared the output of the proposed method with the existing approaches and it seems too superior to the existing methods