Publication

Abstract : Sharing data between connected and automated healthcare sectors without any protection will leak private information. Health information of patient must be carefully handled when shared between various actors in health sector like doctors, lab technicians and data scientist or model owners who executes machine learning model for disease analysis. Patient health related data like X-Ray images must be secured before disease analysis is done. We have taken Pediatric chest x-rays for pneumonia classification. The X-ray images are encrypted before classification. Our work focusses on three encryption algorithm and evaluates the performance based on execution time. After encryption a deep learning model for image classification is built with Convolutional Neural Network (CNN). For experimental results, same dataset is classified using VGG16 transfer learning model which shows a better performance in terms of accuracy and processing cost. We have also demonstrated that the accuracy is further enhanced when the same model is executed in federated distributed learning environment.