ECG steganography using curvelet transform

https://doi.org/10.1016/j.bspc.2015.07.004Get rights and content

Highlights

  • An attempt is made to hide patient data in ECG signals using steganography.

  • Curvelet transform permit identifying characteristic points of the ECG signal.

  • Modifying coefficients around zero is most favorable.

  • An adaptive n × n sequence scheme is proposed to identify the watermark location.

  • Diagnosability is preserved while achieving imperceptibility and reliability.

Abstract

ECG steganography allows secured transmission of patient data that are tagged to the ECG signals. Signal deterioration leading to loss of diagnosis information and inability to retrieve patient data fully are the major challenges with ECG steganography. In this work, an attempt has been made to use curvelet transforms which permit identifying the coefficients that store the crucial information about diagnosis. The novelty of the proposed approach is the usage of curvelet transform for ECG steganography, adaptive selection of watermark location and a new threshold selection algorithm. It is observed that when coefficients around zero are modified to embed the watermark, the signal deterioration is the least. In order to avoid overlap of watermark, an n × n sequence is used to embed the watermark. The imperceptibility of the watermark is measured using metrics such as Peak Signal to Noise Ratio, Percentage Residual Difference and Kullback-Leibler distance. The ability to extract the patient data is measured by the Bit Error Rate. Performance of the proposed approach is demonstrated on the MIT-BIH database and the results validate that coefficients around zero are ideal for watermarking to minimize deterioration and there is no loss in the data retrieved. For an increased patient data size, the cover signal deteriorates but the Bit Error Rate is zero. Therefore the proposed approach does not affect diagnosability and allows reliable steganography.

Introduction

Modern wearable biomedical devices enable ubiquitous healthcare monitoring. The bio-physiological parameters acquired by these devices can be transmitted over the internet. This allows patients to receive care giver's assistance even remotely. During transmission, patient data such as their personal identity is tagged along with the medical information. Patient data protection needs to be ensured in spite of the threat of unauthenticated access [1], [2], [3]. One way to achieve this is by employing data hiding techniques [4], [5], [6]. Steganography is one such technique where personal data are hidden in biomedical signals [7], [8], [9], [10], [11], [12]. Example of personal data include: patient name, age, gender and past treatment details. In medical domain, the bio-medical signal is the cover signal and personal data to be hidden is referred to as watermark. In this work, ECG is the cover signal and patient data is the watermark.

Steganography causes irreversible deterioration to the cover signal. In ECG signals, the characteristic points that help in diagnosis are the QRS complex, P and T waves. Hence, in ECG steganography it is imperative to minimize the deterioration at these characteristic points. This permits preserving the information needed for diagnosis, despite the deterioration. The extent of deterioration is an usual performance measure of a steganography algorithm. Steganography is usually performed in frequency domain. It consists of cover signal decomposition, digital watermark embedding and watermark retrieval. The signal is usually decomposed using a selected transform. Widely used transformation techniques are Discrete Wavelet Transform (DWT) and Fast Discrete Curvelet Transform (FDCT) [7], [13], [14], [15], [16]. Chen et al. [8] evaluates ECG steganography using DWT, Discrete Fourier Transform (DFT) and Discrete Cosine Transform (DCT). They conclude that ECG steganography in transform domain is efficient and useful.

Watermarking methods such as Least Significant Bit (LSB), coefficient quantization and Singular Value Decomposition (SVD) are used to embed the data into the cover signal [7], [8], [17]. The challenge lies in selecting coefficients that will lead to minimal changes at the characteristic points. In order to avoid overlapping of watermark allocation and modification of pixel value, Hong et al. [18] separates each watermarked position by at least one pixel. Researchers [10], [11], [12] proposed wavelet based watermarking schemes in biomedical images. They demonstrated improvements in robustness, imperceptibility and integrity control capability. Despite promising results, wavelets have limitations in representing curves. Curves are where important phenomenon or information such as characteristic points are present. Hence, it is desirable to develop a steganography framework with the following features: (i) identify the coefficients that store the crucial information for diagnosis (ii) minimally deteriorate the cover signal while preserving the diagnosability information and (iii) retrieve patient data without any loss.

In the recent past, Candes et al. [19] introduced a new member in the family of wavelet transforms called the curvelet transforms. They were developed to address the limitations of traditional multiscale representations. They also provide optimal sparse representation of edges. Curvelet coefficients are computed from different scale, orientation and translation analysis of an image. Therefore, curvelets are able to represent curved edges well. Hien et al. [13] evaluated watermarking in curvelet coefficients of an image using thresholding algorithm. A high capacity image steganography using curvelet transform is presented by Al-Ataby et al. [14]. Here, the threshold value is the mean of curvelet coefficients. Curvelet coefficients whose values are less than the threshold value are chosen for watermarking. Feature point based image watermarking using curvelet transform is proposed by Huang et al. [15] to resist the geometric distortion. Leung et al. [16] conducts an extensive study of watermarking methods using curvelet transform. They introduced a blind watermarking scheme which uses secret keys to determine the watermark position. The advantage of blind watermarking scheme is that the cover image is not necessary during watermark extraction. These studies show that curvelet transform can be used in image steganography successfully and we extend the idea to ECG steganography.

In this study, FDCT based ECG steganography is investigated. FDCT allows identifying the coefficients that represent curves. An adaptive thresholding algorithm is proposed to minimize signal deterioration. A signal from MIT-BIH normal sinus rhythm database [20] is used for demonstration purpose. The proposed approach improves the imperceptibility of watermark and reduces error rate in extracted patient data. The performance of steganography algorithm is measured using Peak Signal to Noise Ratio (PSNR) and Bit Error Rate (BER), respectively. Percentage Residual Difference and Kullback-Leibler divergence (KL) which provides the distance between the original and the watermarked signal are also provided. Imperceptibility of watermark is also measured by the watermarking capacity referred to as watermark size. We present the performance of the proposed approach for different watermark sizes as well.

The rest of the paper is organized as follows: In methodology section, the general framework of steganography is presented along with discussions on curvelet and quantization approach. The description of the proposed approach is presented next following which discussions on adaptive watermarking and threshold selection are presented. The performance of the proposed approach is presented in results and discussions.

Section snippets

Methodology

The basic architecture of steganography using biomedical signal consists of two major components: watermark embedding and watermark extraction, as shown in Fig. 1. The biomedical signal when subjected to a transformation results in coefficients. The watermark embedding algorithm hides the patient data into these coefficients. The inverse transform is then applied to construct the watermarked signal. This watermarked signal upon transmission is received by the health care provider. Here, the

Results and discussion

In order to construct 2D ECG image from 1D ECG data, it is subjected to bandpass filtering followed by differentiation which allows distinguishing the QRS complexes. The signal is then subjected to a point-by-point squaring which makes the entire data positive ahead of the subsequent integration. The fiducial mark for temporal location of QRS complex is the maximum slope of QRS complex or peak of R value. Since the sampling rate of 1D ECG signal is 128 Hz, 64 points are taken at both sides of

Conclusion

Performance of FDCT based ECG steganography algorithm using adaptive n × n sequence watermarking technique was studied in this work. FDCT of cover signal results in a set of coefficients and patient data can be hidden in these coefficients. When the coefficients are modified, the cover signal deteriorates and affects the diagnosability. The focus of this work is to minimize the deterioration while preserving the diagnosability. We study the effect of modifying coefficients at three different

References (33)

  • X. Kong et al.

    Watermarking medical signals for telemedicine

    IEEE Trans. Inf. Technol. Biomed.

    (2001)
  • A. Giakoumaki et al.

    Secure and efficient health data management through multiple watermarking on medical images

    Med. Biol. Eng. Comput.

    (2006)
  • R. Acharya et al.

    Simultaneous storage of patient information with medical images in the frequency domain

    Comput. Methods Programs Biomed.

    (2004)
  • R.C. Raúl et al.

    Data hiding scheme for medical images

  • T.D.H.T.D. Hien et al.

    Digital watermarking based on curvelet transform

  • A.A. Al-Ataby et al.

    High capacity image steganography based on curvelet transform

  • Cited by (59)

    • An efficient data hiding for ECG signals based on the integer wavelet transform and block standard deviation

      2022, Journal of King Saud University - Computer and Information Sciences
    View all citing articles on Scopus
    View full text