An enhanced reversible data hiding algorithm using deep neural network for E-healthcare

Abstract

E-healthcare requires communication of patient report to a specialized doctor in a real time scenario. Therefore, any harm to patient medical data can lead to a faulty diagnosis that can be lethal for the patient. Some of the earlier state-of-the-art reversible data hiding methods make use of prediction for embedding data reversibly. To ensure secure and safe communication in encrypted domain, predictors has attained an extraordinary consideration from the research community nowadays. This paper presents an enhanced reversible data hiding method that gives a high embedding capacity by embedding m, ($m\ge 1$) binary bits of electronic patient information (EPI) at embeddable pixels of cover image respectively. In addition to EPI, a fragile watermark has been embedded for observing any tamper to the patient data during transmission phase. In proposed method, deep neural network is used for prediction and data is embedded through generalized prediction error expansion scheme. Through combination of stream ciphers and logistic map, proposed method generated encrypted stego image to improve resistance against various attacks such as cipher-only attack and differential attack respectively. The experimental study reveals that for all types of test images, proposed method has an embedding capacity nearly twice larger than the compared methods, at receiver end precisely recover EPI (Bit Error Rate value is zero) with a PSNR value of $\infty$ dB between the cover image and reconstructed image successfully. For all test images, the proposed method altogether beat all the compared methods in its ability to embed secret information and precisely recover it with maintaining the visual quality of stego images too.