Novel Approach for Image Stenography Based on Integer Wavelet Transform First Author#, Second Author*, Third Author#

First-Third Department, First-Third University Address Second Company Address Including Country Name
*

Abstracts- The proposed system presents a novel approach of building a secure data hiding technique of Steganography using integer wavelet transform along with Genetic algorithm. The novel scheme embeds data in integer wavelet transform coefficients by using a mapping function based on Genetic Algorithm in an 8x8 block on the cover image. The optimal pixel adjustment process is applied after embedding the message. We employed frequency domain to increase the robustness of our steganography method. Integer wavelet transform avoid the floating point precision problems of the wavelet filter. We use GA and Optimal Pixel Adjustment Process to obtain an optimal mapping function to reduce the difference error between the cover and the stego-image and to increase the hiding capacity with low distortions respectively. Simulation results show that the novel scheme outperforms adaptive steganography technique based on integer wavelet transform in terms of peak signal to noise ratio and capacity. Keywords---Steganography, Integer Transform, Peak Signal to Noise Ratio. Wavelet

I.INTRODUCTION Steganography is the art of hiding secret information in the form of cover which can be image[1], complex audio[2], video or any sophisticated biometrics formats [3]. Clearly, the goal of cryptography is to protect the content of messages [4], steganography is to hide the existence of messages. An advantage of steganography is that it can be employed to secretly transmit messages without the fact of the transmission being discovered. Generically, the steganography process is classified into two phases in majority of the prior research work e.g. message embedding and extraction. In the embedding operation, a secret message is transformed into a bit stream of bits; this is embedded into the least significant bits (LSBs) [5] of the image pixels. The embedding overwrites the pixel LSB with the message bit if the pixel LSB and message bit do not match. Otherwise, no changes are necessary. For the extraction operation, message bits are retrieved from pixel LSBs and combined to form the secret message. There are two main selection algorithms that can be employed to embed secret message bits: sequential and random. For sequential selection, the locations of pixels used for embedding are selected sequentially one after another. For instance, pixels

are selected from left to right and top to bottom until all message bits are embedded. With random selection, the locations of the pixels used for embedding are permuted and distributed over the whole image. The distribution of the message bits is controlled by a pseudorandom number generator whose seed is a secret shared by the sender and the receiver. This seed is also called the stego-key. The latter selection method provides better security than the former because random selection scatters the image distortion over the whole image, which makes it less perceptible. In addition, the complexity of tracing the selection path for an adversary is increased when random selection is applied. Apart from this, steganographic security can be enhanced by encrypting the secret message before embedding it. Whereas the proposed work has considered exclusive gray image from standard image datasets of Lena, Boat, section 1 we give an overview of related work which identifies all the major research work being done in this area. Section 2 highlights about the proposed system along with system architecture and algorithm description. Implementation and result analysis is discussed Section 3 followed by conclusion in Section -4. II. PROPOSED METHOD Proposed method introduces a new method of embedding secret data within skin as it is not that much sensitive to HVS (Human Visual System). Instead of embedding data anywhere in image, data will be embedded in selected regions. Overview of method is briefly introduced as follows. cover image is transformed in frequency domain. This is performed by applying Haar-DWT, the simplest DWT on image leading to four subbands. Then payload (number of bits in which we can hide data) is calculated. Finally secret data embedding is performed in one of the high frequency sub-band before performing all steps cropping on input image is performed and then in only cropped region embedding is done, not in whole image. Cropping results into more security than without cropping. Since cropped region works as a key at decoding side. Here embedding process affects only certain Regions of Interest (ROI) rather than the entire image. So utilizing objects within images can be more advantageous. This is also called as Object Oriented steganography. Next sub-

sections briefly introduce skin tone detection and DWT. III. INTEGER WAVELET TRANSFORM The proposed algorithm employs the wavelet transform coefficients to embed messages into four subbands of two dimensional wavelet transform. To avoid problems with floating point precision of the wavelet filters, we used Integer Wavelet Transform. The LL subbands in the case of IWT appears to be a close copy with smaller scale of the original image while in the case of DWT the resulting LL subbands is distorted [9] as shown in "Fig.1 Lifting Scheme is one of the techniques on integer wavelet transform. The decomposing filter in integer wavelet transform can be calculated according to [13]: S1, k= (S0, 2k+S0, 2k+1)/2 d1, k = S0, 2k+1-S0, 2k (1)

Di, j=I2i+1, 2j+1-I2i, 2j

(3)

Where I is the original image, A, H, V and D are the low pass, horizontal, vertical and diagonal coefficients. The inverse is given in (4). I2i, 2j=Ai, j - Hi, j/2 I2i, 2j+1=Ai, j+ (Hi, j+1/2 I2i+1, 2j=I2i, 2j+Vi, j-Hi, j I2i+1, 2j+1=I2i+1, 2j+Di, j+Vi, j Splits component into numerous frequency bands called sub bands known as, LL Horizontally and vertically low pass LH Horizontally low pass and vertically high pass HL - Horizontally high pass and vertically low pass HH - Horizontally and vertically high pass Since Human eyes are much more sensitive to the low frequency part (LL subband) we can hide secret message in other three parts without making any alteration in LL subband [12]. As other three sub-bands are high frequency subband they contain insignificant data. Hiding secret data in these sub-bands doesnt degrade image quality that much. IWT used in this work is Haar IWT, the simplest IWT. IV.EMBEDDING PROCESS Suppose C is original 24-bit color cover image of MN Size. It is denoted as: C = {xij, yij, zij |1 i M, 1 j N, xij, yij, zij {0, 1... 255}} Let size of cropped image is McNc where McM and NcN and Mc=Nc. i.e. Cropped region must be exact square as we have to apply DWT later on this region. Let S is secret data. Here secret data considered is binary image of size ab. Fig. 1 represents flowchart of embedding process. Different steps of flowchart are given in detail below. size ab. Fig. 1 represents flowchart of embedding process. Different steps of flowchart are given in detail below. Steps of our Projects: Step 1 : load the Image (M x N). Step 2 : Cropped Image Creation from Input Image (Mc x Nc),

Fig - 1 L (a) Original image Lena and how to analyze in wavelet domain. (b) One level 2DDWT in subbands LL (c) One level 2DIWT in subbands LL. Then the inverse transform can be calculated by: S0, 21=S1,1 d1,1/2 S0, 21+1 = S1, 1+d1, 1+1)/2 (2)

These equations should be in 2D in order to be applied on images. Simple 2D transform has employed in this paper and it can be computed for an image according to [14]: Ai, j= (I2i, 2j+I2i+1, 2j)/2 Hi, j=I2i, 2j+1-I2i, 2j Vi, j=I2i+1, 2j-I2i, 2j

Step 3: Apply DWT to only cropped area (McNc) not whole image (MN). This yields 4 sub-bands denoted as HLL, HHL, HLH, and HHH. (All 4 subband are of same size of Mc/2, Nc/2). Payload of image to hold secret data is determined based on no. of skin pixels present in one of high frequency sub-band in which data will be hidden.

Load Image

pixels of cropped area in original image so that r calculated. V. EXTRACTION PROCESS Secret data extraction is explained as follows: 24 bit color stego image of size MN is input to extraction process. We must need value of cropped area to retrieve data. Suppose cropped area value is stored in rect variable that is same as in encoder. So this rect will act as a key at decoder side. All steps of Decoder are opposite to Encoder. Care must be taken to crop same size of square as per Encoder. VI. SIMULATION RESULTS In this section we demonstrate simulation result for proposed scheme. This has been implemented using MATLAB 7.0. We use Peak signal to noise ratio (PSNR) to evaluate quality of stego image after embedding the secret message. The performance in terms of capacity and PSNR (in dB) is demonstrated for the method in the following subsections. Stego Image

Cropped Image

Integer Wavelet Transform

Embedding

Inverse IWT

Key
Cropped Image Stego Image Perform IWT Fig.2 Flowchart of Embedding Process Step 4 : Perform embedding of secret data in one of sub-band. Low frequency sub-band any high frequency sub-band can be selected for embedding as LL sub-band contains significant information. Embedding in LL sub-band affects image quality greatly. We have chosen high frequency HH subband. While embedding, secret data will not be embedded in all pixels of IWT subband but to only cropped pixels. Embedding is performed in Gplane and B-plane but strictly not in R-plane as contribution of R plane in skin color is more than G or B plane. So if we are modifying R plane pixel values, decoder side doesnt retrieve data. Step 5 : Perform IDWT to combine 4 sub-bands. Step 6 : A cropped stego image of size McN c is obtained in above step (step 5). This should be similar to original image after visual inspection but at this stage it is of size Mc Nc, So we need to merge the cropped stego image with original image to get the stego image of size MN. To perform merging we require coefficients of first and last Retrieval Fig 3 Block Diagram of Extraction Process PSNR is defined as per the below Eq, PSNR = 10 log10 (2552 / MSE) Where, MSE = (1/ (M N)) (xij - yij) ^2 xij and yij represents pixel values of original cover image and stego image respectively. The calculated PSNR usually adopts dB value for quality judgement, the larger PSNR is, higher the image quality (which means there is a little difference between cover image and stego image).On the contrary smaller dB value means there is a more distortion. PSNR values falling below 20 dB indicate fairly a low quality.

Secret Data

Fig - 5 Cover Image

Fig - 6 Image to hide

own advantages and disadvantages. But if method is implemented with cropping then it will ensure more security than without cropping case. As with cropping case we need cropped region at the decoder side then only secret data extraction is possible. So cropped region works as a key at decoder side. For without cropping method intruder may try to perform DWT randomly and can hack secret data from sub-band with trial and error method. From the table 1 it is obvious that PSNR of without cropping case is more than with cropping case. So, this is trade off that occurs if we need more security. TABLE 1. CAPACITY AND PSNR OF 4 FINAL STEGO IMAGES IN PROPOSED METHOD

Fig - 7 Cropped Stego Image After step 5 of Embedding process

Fig - 8 Final Stego Image after step 6 of embedding process

Cover Image (356 x 356) Image 1 Image 2 Image 3 Image 4

Capacity of Cover Image Case A Case B 7173 1067 1452 4850 5294 1056 1354 2752

PSNR Size of Logo Case A Case B 53.0 51.9 51.2 46.4 50.7 50.5 49.7 49.2 45.4 48.7 64 x 70 32 x 32 32 x 32 32 x 32

Fig - 9 Retrieved image

A. Performance of the proposed

Method: After inserting secret data in cropped image, resulted cropped stego image is shown in Figure 7. (Result of step 5 of embedding process). As this doesnt look like cover image merging is performed to obtain final stego image that is shown in Figure 8. (Result of step 6 of embedding process). For merging co-ordinates of first and last pixels of cropped Image in original image are calculated. After performing extraction process retrieved image is shown in figure 9. Above method uses cropping. Same proposed method is implemented for without cropping case. In without cropping case secret data is hidden in one of the sub-band which is obtained by performing the DWT on whole image and not only to cropped region. PSNR is calculated for four different final stego images resulted from a considered image and three more sample images. This PSNR for different cases is shown in table 1. Average PSNR of proposed method is calculated based on the obtained PSNR. Performing biometric steganography with cropping or without cropping, both are having its

Average PSNR Case A- Without Cropping Case B- With Cropping

V. CONCLUSIONS Digital Steganography is a interesting scientific area which falls under the umbrella of security systems. In this paper biometric steganography is presented that uses skin region of images in DWT domain for stego the secret data. By embedding data in only convinced region (here skin region) and not in entire image security is enhanced. Also image cropping concept introduced, maintains security at reputable level since no one can take out message without having value of cropped region. Features obtained from DWT coefficients are utilized for secret data embedding. This also improves the quality of stego because secret messages are embedded in high frequency sub-bands which human eyes are less sensitive to. According to simulation results, future approach provides well image quality.

ACKNOWLEDGMENT The authors would like to thank to the earlier work regarding steganography whose guidance significantly contributed to the work made in this paper. All work done, images shown in this paper are for educational purpose and not for commercial purpose. REFERENCES [1] A. Cheddad, J. Condell, K. Curran and P. McKevitt, Biometric inspired digital image Steganography, in: Proceedings of the 15th Annual IEEE International Conference and Workshops on the Engineering of ComputerBased Systems (ECBS08), Belfast, 2008 pp. 159168. [2] Petitcolas, F.A.P.: Introduction to Information Hiding. In: Katzenbeisser, S and Petitcolas, F.A.P (ed.) (2000) Information hiding Techniques for Steganography and Digital Watermarking. Norwood: Artech House, INC. [3] Lin, E. T. and Delp, E. J.:A Review of Data Hiding in Digital Images. Retrieved on 1.Dec.2006 from Computer Forensics, Cyber crime and Steganography Resources, Digital Watermarking Links and Whitepapers, Apr 1999 [4] Johnson, N. F. and Jajodia, S.: Exploring Steganography: Seeing the Unseen. IEEE Computer, 31 (2): 26-34, Feb 1998. [5] Fridrich, J., Goljan, M. and Du, R., (2001). Reliable Detection of LSB Steganography in Greyscale and Color Images. Proceedings of ACM, Special Session on Multimedia Securit and Watermarking,Ottawa, Canada, October 5, 2001, pp. 27- 30. [6] Po-Yueh Chen and Hung-Ju Lin A DWTBased Approach for Image Steganography, International Journal of Applied Science and Engineering, 2006. 4, 3: 275-290 [7] Chang, C. C., Chen, T.S and Chung, L. Z., A steganographic method based upon JPEG and quantization table modification, Information Sciences, vol.[4], pp. 123-138(2002).

[8] Provos,N. and Honeyman, P: Hide and Seek: An introduction to steganography. IEEE Security and privacy, 01 (3): 32-44, May-June 2003 [9] Abbas Chedda, Joan Condell, Kevin Curran and Paul Mc Kevitt A Skin Tone Detection Algorithm for an Adaptive Approach to Steganography, School of Computing and Intelligent Systems, Faculty of Computing and Engineering, University of Ulster, BT48 7JL, Londonderry, Northern Ireland, UK, 2008 [10] Ahmed E., Crystal M. and Dunxu H.: SkinDetection-a short Tutorial, Encyclopedia of Biometrics by Springer-Verlag Berlin Heidelberg 2009 [11]Sobottka, K. and Pitas, I.:Extraction of facial regionsand features using color and shape information. Proc. IEEE Internationa Conference on Image Processing, pp. 483- 486.(1996) [12]Chen,P. Y.and Liao,E.C., :A new Algorithmfor Haar Wavelet Transform, 2002 IEEInternational Symposium on Intelligent Signal Processing and Communication System, Pp.453457(2002). 43