Procedia Technology 10 ( 2013 ) 105 – 111

2212-0173 © 2013 The Authors. Published by Elsevier Ltd. Open access under CC BY-NC-ND license.Selection and peer-review under responsibility of the University of Kalyani, Department of Computer Science & Engineeringdoi: 10.1016/j.protcy.2013.12.342

ScienceDirect

International Conference on Computational Intelligence: Modeling Techniques and Applications (CIMTA) 2013

A Genetic Algorithm based Steganography using Discrete Cosine Transformation (GASDCT)

Amrita Khamruia, J K Mandalb

aDept. of Computer Science and Engineering Future Institute of Engineering & Management, Kolkata,

Sonarpur Station Road, Sonarpur, Kolkata-150, West Bengal e-mail: khamruiamrita@gmail.com

bDept. of Computer Science and Engineering, University of Kalyani,

Kalyani, Nadia-741235, West Bengal, India e-mail:- jkm.cse@gmail.com.

Abstract

In this paper a Genetic Algorithm based steganographic technique in frequency domain using discrete cosine transform has been proposed. A 2×2 sub mask of the source image is taken in row major order and Discrete Cosine Transformation is applied on it to generate four frequency components. Two bits of the authenticating image are embedded into each transformed coefficients except the first one. In each coefficient second and third positions from LSB are chosen for embedding in the transform domain. Stego sub intermediate image is generated through reverse transform. Sub mask from this intermediate image is taken as initial population. New Generation followed by Crossover is applied on initial population to enhance a layer of security. New Generation is applied to initial population. Rightmost three bits of each byte are taken, a consecutive bitwise XOR is applied on it in three steps which generates a triangular form. The first bit of each intermediate step is taken as the output and Crossover is performed on two consecutive pixels where two LSB bits of two consecutive bytes are swapped. The dimension of the hidden image is embedded followed by the content. Reverse process is followed during decoding. The proposed scheme obtains high image fidelity, PSNR and high capacity of embedding in stego images compared Chang Chin et al [1].

© 2013 The Authors. Published by Elsevier Ltd. Selection and peer-review under responsibility of the University of Kalyani, Department of Computer Science & Engineering.

Keywords: Frequency Domain, Steganography; peak signal to noise ratio (PSNR); Least significant bit (LSB); Image fidelity (IF); mean square error (mse); Discrete Cosine Transformation (DCT)

Available online at www.sciencedirect.com

© 2013 The Authors. Published by Elsevier Ltd. Open access under CC BY-NC-ND license.Selection and peer-review under responsibility of the University of Kalyani, Department of Computer Science & Engineering

106 Amrita Khamrui and J.K. Mandal / Procedia Technology 10 ( 2013 ) 105 – 111

1. Introduction

Steganography offers an essential alternative to image integrity and authenticity problem. It is a kind of data hiding technique that provides another way of security protection for digital image data. Unlike utilizing a particular cipher algorithm to protect secret data from illicit access, the purpose of steganography is to embed secret data in preselected meaningful images, called cover images, without creating visually perceptible changes to keep an invader unaware of the existence of the secret. Generally, a steganographic message may be picture, video, sound file [8], [7]. A message may be hidden by using algorithms like invisible ink between the visible lines of innocuous documents to ensure the security which is a big concern in modern day image trafficking across the network. Steganography can be achieved in two ways. One is spatial domain steganography [5] and another is frequency domain steganography [6]. In spatial domain steganography the hidden information is directly embedded into image pixels. In frequency domain steganography the image pixels are first transformed into frequency domain using discrete fourier transformation [4]/ discrete cosine transformation [1]/ discrete wavelet transformation [6] etc. Then the information is embedded on it. Genetic Algorithm [3] in conjunction with steganography has been incorporated in the research work to add another layer of security for more sensitive application like military people, research institute and medical diagnosis etc. Data hiding refers to the nearly invisible [9], [14], [15], [16] embedding of information within a host data set as message, image or video. A classic example of steganography is that of a prisoner communicating with the outside world under the supervision of a warden. The data hiding represents a useful alternative to the construction of a hypermedia document or image, which is very less convenient to manipulate. The goal of steganography is to hide the message/image in the source image by some key techniques and cryptography is a process to hide the message content. The motive is to hide a message inside an image keeping its visible properties [10] the source image as close to the original. The most common methods to make these alteration is usage of the least-significant bit (LSB) developed through [2], [10] masking, filtering and transformations on the source image [7]. Present proposal would facilitate secure message transmission through block based data hiding. Most of the works [13], [12], [11] used minimum bits of the hidden image for embedding in spatial domain, but the proposed algorithm embeds in transformed domain with a bare minimum distortion of visual property.

Rest of the paper is organized as follows. Section 2 deals with the proposed technique. Results and comparisons are given in section 3. For generating result and compare it with existing approach some benchmark image is taken from image database [17]. Concluding remarks are presented in section 4 and references are drawn at end.

2. The Technique

The process of embedding is divided into four steps, transformation into frequency domain using Discrete Cosine transformation(DCT), embedding secret information, re-transformed into spatial domain using inverse Discrete cosine Transformation (IDCT) and applying Genetic Algorithm. In GASDCT, insertion is made by choosing 2×2 non overlapping sub mask from the source gray scale image in row major order. Discrete cosine transformation is performed on it to generate four frequency components. Like other transforms, the Discrete Cosine Transform (DCT) attempts to de-correlate the image data. The most common DCT definition of a 1-D sequence of length N is

for u = 0,1,2, ..., N − 1. Similarly, the inverse transformation is defined as

for x = 0,1, 2, ..., N − 1. In both equations (1) and (2) α(u) is defined in equation (3)

(1)

(2)

107 Amrita Khamrui and J.K. Mandal / Procedia Technology 10 ( 2013 ) 105 – 111

The first transform coefficient is the average value of the sample sequence. In literature, this value is referred to as the DC Coefficient. All other transform coefficients are called the AC Coefficients. The 2-D DCT is a direct extension of the 1-D case and is given by

for u, v = 0,1,2, ..., N − 1, where α(u) and α(v) are defined in (3). The inverse transform is defined as

for x , y = 0,1, 2, ..., N − 1. The 2-D basis functions can be generated by multiplying the horizontally oriented 1-D basis functions with vertically oriented set of the same functions. Two bits of the secret information are embedded onto the second and third position of each lower frequency component i.e. AC coefficients excluding the highest frequency one i.e. DC coefficient. The position of embedding is chosen in such a way that there is no loss of precession during reverse transformation. The dimension of the hidden image is also embedded on to the first thirty two bit position.

The resultant image sub mask is converted back to spatial domain by IDCT using equation (5). From this image in spatial domain image mask of size 32 bits are taken as initial population. New Generation followed by Crossover is applied on this initial population. New Generation is performed on rightmost three bits of each byte by consecutive bitwise XOR operation on three steps, taking the MSB of the intermediate stream generated in each step. Crossover is applied on the consecutive two pixels in row major order. As a result rightmost two bits of two consecutive pixels are swapped. Reverse process is followed at time of decoding.

Schematic diagram of the technique is shown in Figure1. Figure.1.(a) shows process of encoding that of Figure.1.(b) depicts the process of decoding.

1.(a): The process to embed the secret data into the source image

(3)

(4)

(5)

Cover image (512 x 512)

2 x 2 block of pixels from cover image

2 x 2 image matrix

ZT on 2 x 2

matrix

Apply GA Watermarked Image

(512 x 512) IZT

108 Amrita Khamrui and J.K. Mandal / Procedia Technology 10 ( 2013 ) 105 – 111

1.(b): Process to extract secret information from the watermarked image Figure 1: Schematic diagram of GASDCT

3. Example

Consider a byte of Jet image (figure 2.(a)) to be inserted into each mask of Pepper image (Figure 2.(c)). Figure 2.(b) shows pixels of Boat image in spatial domain. Two bits of the Jet image are inserted into the Boat image in 2×2 mask. Insertion is done in the second and third position of the image pixel. Resultant image after embedding is shown in figure 2.(d) in frequency domain and Figure 2.(e) in spatial domain. Figure 2.(f) and 2.(g) show New Generated Stego image and Crossover stego image respectively.

000000000110

Figure 2.(a): hidden image bits

21 40

26 113

(Source sub image block)

Figure 2.(b): Source Image Boat

100 -53

-39 34

(Coefficients of transformed sub image block)

Figure 2.(c): Source image Pepper after DCT

100 49

33 32

(Coefficients of embedded sub image block)

Figure 2.(d): Embedded Image Block

25 42

26 107

(Embedded sub image block after IDCT)

Figure 2.(e): Embedded Image Block in spatial domain

25 42

26 107

(stego sub image block after New Generation)

Figure 2.(f): New Generated stego Image Block

26 41

27 106

(stego sub image block after Crossover)

Figure 2.(g): Crossovered stego Image Block

Figure 2: Encoding Process of GASDCT

4. Results, comparison and analysis

Extensive analysis has been made on various images [17] using GASDCT technique. This section represents the results, discussions in terms of visual interpretation and peak signal to noise ratio. Figure 3.(a) shows the host images Mandrill, Pepper, Boat. Figure 3.(b) shows embedded Mandrill, Pepper, Boat on embedding Jet image using

Embedded image matrix (512 x 512)

2 x 2 block embedded image after reverse GA

ZT on 2 x 2

matrix

2 x 2 matrix with integer

valuesAuthenticating

Image

Extract secret data from

embedded data

IZT

109 Amrita Khamrui and J.K. Mandal / Procedia Technology 10 ( 2013 ) 105 – 111

GASDCT. Figure 3.(c) is the authenticating image Jet. Table I shows the PSNR value of embedding against each of the source image. From the table it is seen that the maximum value of the PSNR is 44.92and that of minimum value of the PSNR is 43.68. The proposed technique is compared with Chang Chin et al [1] approach in terms of PSNR and capacity and is shown in Table II. From results it is seen that a high capacity of embedding has been achieved in terms of good PSNR. The equations 5, 6, 7 are used to calculate PSNR, MSE and IF (image fidelity) respectively.

Figure 3: Visual Effect of Embedding in GASDCT

3.a.i. Host Mandrill 3.a.ii. Host Pepper 3.a.iii. Host Boat 3.b.i Embedded Mandrill

3.b.ii Embedded Pepper 3.b.iii Embedded Boat 3.c.i. Hidden Jet

(8)

(6)

(7)

110 Amrita Khamrui and J.K. Mandal / Procedia Technology 10 ( 2013 ) 105 – 111

Table 1. PSNR, MSE, IF values obtained for various images using GASDCT

Host Image PSNR values MSE Values IF

Lenna 44.783592 2.161331 0.999866

Baboon 44.921604 2.093727 0.999888

Peppers 44.829178 2.138763 0.999880

Boat 44.860226 2.123528 0.999888

Cameraman 43.680126 2.786556 0.999846

Elaine 44.789280 2.158501 0.999896

Sailboat 44.799095 2.153629 0.999894

Table 2.Comparison of PSNR(in DB) obtained for various images using GASDCT and Existing [1]

Host Image PSNR values of GASDCT

Capacity of GASDCT

PSNR values of existing[1]

Capacity of existing[1]

Boat 44.860226 80000 29.75 36,710

Lenna 44.783592 80000 30.34 36,850

Mandrill 44.921604 80000 26.46 35,402

Pepper 44.829178 80000 30.65 36,804

Boat 44.860226 80000 29.75 36,710

5. Conclusion

This paper proposed a novel embedding/authentication approach based on Discrete Cosine Transformation for gray scale images. From experimental results it is clear that the proposed technique obtained high PSNR along with good image fidelity for various images which conform that GA based Discrete Cosine Transformed based image steganography can obtain better visibility/quality.

Acknowledgements

The authors express deep sense of gratitude to the Department of CSE, University of Kalyani, India where the computational works have been carried out.

References

1. Chang Chin Chen et al “Reversible hiding in DCT based compressed images”, Science Direct, Information Sciences vol: 177 , issue: 13 pages: 2768–2786 July (2007).

2. Atallah M. Al-Shatnawi “A New Method in Image Steganography with Improved Image Quality”, Applied Mathematical Sciences, Vol. 6, 2012, no. 79, 3907 – 3915. 3. Mandal J K, Khamrui A “A Genetic Algorithm Based Steganography on Color images (GASCI)”, International journal of Signal and

Imaging Systems Engineering (IJSISE), Indersciences, accepted may 2011. 4. Khamrui A et al., “ A Genetic Algorithm based Steganography in Frequency Domain (GASFD)”, IEEE 1st International Conference

on Communications and Industrial Applications (ICCIA 2011),IEEE Conference ID: 19700, 26th -28th December, 2011, Science City, Kolkata, 2011.

111 Amrita Khamrui and J.K. Mandal / Procedia Technology 10 ( 2013 ) 105 – 111

5. Khamrui A et al., “A Data- Hiding Scheme for Digital Images using Pixel Value Differencing (DHPVD)”, International Symposium on Electronic System Design(ISED), 19-21 December, Kochi, India, 2011.

6. Mandal, J. K. et al. “Authentication of images through non convoluted DCT”,IEEE 1st International Conference on Communications and Industrial Applications (ICCIA 2011),IEEE Conference ID: 19700, 26th -28th December, 2011, Science City, Kolkata, 2011. 7. Ghoshal N., Mandal, J. K. “A Bit Level Image Authentication / Secrete Message Transmission Technique (BLIA/SMTT)”, Association for the Advancement of Modelling and Simulation Technique in Enterprises (AMSE), AMSE journal of Signal Processing and Pattern Recognition, Vol. 51, No. 4, pp. 1-13, France, 2008. 8. Nameer N. EL-Emam, “Hiding a large Amount of data with High Security Using Steganography Algorithm,” Journal of Computer Science ISSN 1549-3636, vol. 3, no. 4, pp. 223-232, 2007. 9. C. Rechberger, V. Rijman and N. Sklavos, “The NIST cryptographic Workshop on Hash Functions,” IEEE Security and Privacy, vol. 4, pp. 54-56, Austria, Jan-Feb 2006. 10. S. Pavan, S. Gangadharpalli and V. Sridhar, “Multivariate entropy detector based hybrid image registration algorithm,” IEEE Int. Conf. on Acoustics, Speech and Signal Processing, Philadelphia, Pennsylvania, USA, pp. 18-23, March 2005. 11. Hashad A. I. et al “A Robust Steganography Technique using Discrete Cosine Transform Insertion” Information and communication Technology, Enabling Technologies for the New Knowledge Society: ITI 3rd International Conference.ISBN:0-7803-9270-1, 2005. 12. P. Moulin and J. A. O’Sullivan, “Information-theoretic analysis of information Hiding,” IEEE Trans. On Info. Theory, vol. 49, no. 3, pp. 563-593, March 2003. 13. S. Dumitrescu, W. Xiaolin and Z. Wang, “Detection of LSB steganography via sample pair analysis,” IEEE Trans. on Signal processing, Vol. 51, no. 7, pp. 1995-2007, 2003 14. R. Chandramouli and N. Memon, “Analysis of LSB based image steganography techniques,” Proc. of ICIP, Thissaloniki, pp. 1019- 1022, Greece, 2001. 15. Ran-Zan Wang, Chi- Fang Lib, and Ja- Chen Lin, “Image hiding by optimal LSB substitution and Genetic algorithm,” Pattern Recognition Society. Published by Elsevier Science Ltd, 2001. 16. C.Y. Lin and S. F. Chang, “A robust image authentication method surviving JPEG lossy compression,” Proc. SPIE, vol.3312, San Jose, pp. 296-307, Jan. 1998. 17. Allan G. Weber, The USC-SIPI Image Database: Version 5, Original release: October 1997, Signal and Image Processing Institute, University of Southern California, Department of Electrical Engineering. http://sipi.usc.edu/database/(Last accessed on 20th January, 2011).