Using Multimodal Biometrics, Data Hiding, and Encryption for Secure Healthcare Imaging System

Abstract

In this digital era, images are the most vital information carrier used for healthcare communication and entertainment. However, the increasing use of images in several applications also poses a risk of their unauthorised usage or modification without proper attribution to the owner. To overcome this issue while ensuring one-time password (OTP)-based system authentication, this study designed a highly secure healthcare imaging system with multimodal biometrics, data hiding and encryption in a deep learning environment. First, we segmented a medical image via a customised, deep neural network to locate the lesion and non-lesion areas. Next, the lesion part was embedded into the non-lesion part via least significant bit (LSB) substitution and timestamp. Furthermore, the marked non-lesion and lesion parts were combined to generate the marked image. Second, encoded multimodal biometric features, i.e., face and iris, and a novel 2D chaotic system were used to encrypt the marked image before transmission over the network. Through simulation findings on security and accuracy of segmentation and feature extraction design, we demonstrated the feasibility and effectiveness of our proposed secure system, highlighting their superior performance compared to existing techniques.