Encryption by Heart (EbH) for Secured Data Transmission and CNN Based EKG Signal Classification of Arrhythmia with Normal Data

Abstract

Remote healthcare monitoring systems are commonly used to manage patient diagnostic data. These systems are subjected to data privacy and security, reliability, etc. A new technique is introduced in this paper to solve privacy and security issues. Using Discrete Wavelet Transforms (DWT), EKG steganography technique is implemented in the proposed method. This method is based on the techniques of encryption and decryption. Encryption is used to hide the EKG signal within an image and to extract the EKG signal from the encrypted image, decryption is used. Subsequently, a prominent amount of raw EKG time series signal information is given as inputs for convolution neural networks (CNN). The representative and key characteristics used to classify the module autonomously are learned. Thus, the features are learned directly from the prominent time domain EKG signals by using a CNN. Trained characteristics can efficiently substitute the hand-crafted characteristics of the time-consuming user and traditional ad hoc characteristics. Using GoogLeNet CNN we have achieved an accuracy of 0.90625 .