Kongu Nagaraj Final M

INTERNATIONAL CONFERENCE ON “CONTROL, AUTOMATION, COMMUNICATION AND ENERGY CONSERVATION -2009, 4th-6th June 2009 1

A Modulo Based LSB Steganography Method Dr. V. Vijayalakshmi, Dr. G. Zayaraz, and V. Nagaraj Abstract— Steganography is the art of hiding the very presence of communication by embedding secret messages into innocuous looking cover images. The Least Significant Bit (LSB) steganography that replaces the least significant bits of the host medium is a widely used technique with low computational complexity and high insertion capacity. Although it has good perceptual transparency, it is vulnerable to steganalysis which is based on histogram analysis. In all the existing schemes detection of a secret message in a cover image can be easily detected from the histogram analysis and statistical analysis. Therefore developing new LSB steganography algorithms against statistical and histogram analysis is the prime requirement. Index Terms— Least Significant Bit (LSB), Embedding, Stego, Histogram, Statistical Analysis.

——————————  ——————————

1 INTRODUCTION

T

he LSB insertion method is the most common and easiest method for embedding messages in an image with high capacity. But the limitation of this is the secret message is easily detectable from the histogram analysis method. So this paper has proposed a modulo based LSB image[1] steganography algorithm that can effectively resist image steganalysis based on histogram analysis and also statistical analysis. The proposed method combines samples of LSB bits by using addition modulo to form the value which is compared to the part of the secret message. If these two values are equal, no change is made. Otherwise, add the difference of these two values to the sample. Thus, this proposed method embeds the part of the secret message effectively. Statistical analysis is performed on stego image created using the steganography technique. The proposed method shows that it can effectively resist image steganalysis based on statistical analysis and histogram analysis. The following section deals with the existing steganography method and its limitation. Section three deals with proposed method embedding and extraction processes. Then section four discusses the experimental results. Finally section five deals with the conclusion of the work.

2 THE CLASSIC LSB STEGANOGRAPHY AND ITS LIMITATION The existing LSB steganography embeds message into cover image by using message bit stream to replace the cover image’s LSB directly. For increasing the [2]embedding capacity, two or more bits in each sample value can be used to embed messages without detectably degrading the cover image. Although there are several types of LSB embedding methods, for example, the sample value to embed is selected by a pseudo random number in order to resist [3] the visual attacks or steganalysis on sequential steganography, these existing LSB steganography methods have certain limitation. The limitation is that, the hidden information is converted to another format which causes the hidden data to be lost. Histogram [4] and statistical analysis methods [5] can be used to possibly identify an image with hidden information.

Fig. 1

Steganography model

The general steganography model is shown in the Fig.1 .The stego function is used for the encoding of secret data [6] on to the cover image with stego key word. Similarly stego inverse function is used for decoding of secret data from the stego image.

3 PROPOSED STEGANOGRAPHY METHOD The proposed steganography method consists of two main processes. The two processes are embedding and extraction of secret data from the image. 3.1 Embedding Process The Fig. 2 shows the flow chart for the embedding and extraction process. Cover image, secret message and steg key are input for modulo based embedding process.

Dr. G. Zayaraz is Assistant Professor of Pondicherry Engineering College, Puducherry, India. E-mail: [email protected] Dr. V. Vijayalakshmi is Senior Lecturer, Department of Electronics and Communication Engineering in Pondicherry Engineering College, Puducherry, India. E-mail: [email protected] V. Nagaraj is an M.Tech (Wireless Communication) student in Pondicherry Engineering College, Puducherry, India. E-mail: [email protected]

2

IEEE TRANSACTIONS ON XXXXXXXXXXXXXXXXXXXX, VOL. #, NO. #, MMMMMMMM 1996

image by the proposed method, the image does not reveal any remarkable changes in appearance. The changes in the image are also not revealed in the histogram obtained from the stego image. Fig. 4 shows the comparison result of existing LSB method with the histogram results of the proposed method. From Fig. 5 shows the comparison of the histogram of the cover image and the stego image of the proposed method. The histogram analysis of the stego image looks alike to

Fig.2 Flowchart for embedding and extraction process Let G be the set of pixels which is selected by a pseudo random number from the cover image. A stego key is used as the seed of the pseudo random number generator. x is the gray value of cover image pixel. n is determined by the size of embedded message and how many LSB bits in each pixel can be used to embed messages. It can be calculated by function. Fig.5 Histogram comparison of cover and stego image n=l/m

(1)

where l is the length of bit stream of embedded message, and m is the number of bits used to embed messages in each pixel. The bit stream of embedded message is divided into bit segment of m bit length and denoted with E = <e1, e2,….en> ,e{0,1,….2m-1}

(2)

Let LSB(x) be the function to get the m bit LSB value from x, then combining every two samples of LSB bits using addition mod 2 to form the value which is compared to the part of the secret message. The embedded message is given by xi = xi+ ei – (LSB(xi-1)+LSB(xi))Mod 2

(3)

The proposed method gets new G values and uses these new values to replace the corresponding pixel values according to the same pseudo random number. Thus, the message has been embedded into the cover image, and this embedded cover image is called as the stego image.

Fig.6 Comparison of cover image with stego image that of the original image. A comparison of cover image with stego image is shown in Fig.6. These images also look alike. Table.1 shows the statistical analysis results for black-nwhite Lena cover image on comparison with existing LSB method. From Table.1 it is inferred that the statistical parameter values of cover image like mean, variance and RMS values change only in their decimal positions thereby proving that the proposed method can effectively resist steganalysis. Table.1 Statistical analysis results

3.2 Extraction Process The same stego key is used for decoding of secret message from the stego image. The stego key is used to generate the same pseudo random number with which selection of the pixels is done. According to the pseudo random number to construct secret message from the G = <x0, x1, x2 ….xn> pixels, the message can be extracted as given below. ei = (LSB(xi-1)+LSB(xi))Mod 2

(4)

where the proposed method gets E = <e 1, e2,….en> values from the stego image and then it can reconstruct the original secret message.

4

EXPERIMENTAL RESULTS

The proposed method uses Lena image which is the black-n-white image as cover image. Fig. 3 shows the gray scale Lena image used as the cover image, and also the histogram of the cover image which is noted before adding the secret information. After adding secret information within an

Original coverEncoded imageEncoded image (Lena) image with existingwith our LSB method proposed method Mean

125.160488

125.079448

125.152133

Variance 45.205457

45.518826

45.205357

RMS

133.104590

133.066077

133.073968

AUTHOR: TITLE

3

Similarly encoding of the secret message in colour image was also done and the results were compared and tabulated. For the colour image encoding nature6 image is used as cover image. Before encoding secret message within colour

Fig.4 Histogram comparison of existing LSB method and proposed method

Fig.7 colour cover (nature6) image image, histogram analysis of colour cover image was noted. Fig.7 is the colour cover image used for the proposed steganography method. Before encoding of secret data histogram of cover image was noted. Fig.8 shows the histogram analysis of the colour cover image. After encoding the secret message in the selected pixel of the colour cover image, the embedded image has still good transparency to eyes. It is not possible to identify hidden

the stego image is nearly equal to the histogram of the cover image. Thus, it is difficult for a steganalyst to find a secret data hidden in the colour stego image. Fig.9 shows that the histogram comparison of cover colour image and proposed stego image. From this comparison it is very difficult to find the differences. Fig.10 shows there is no difference in the colour cover image and stego image.

Statistical parameters like mean, variance and RMS values are also calculated from the colour image. The values are noted before and after encoding of secret message. The obtained values were tabulated and compared with existing LSB steganography. Table. 2 Statistical analysis results for colour image Original Encoded imageEncoded (nature6) image with existingwith LSB method method

Fig.8 Histogram of the cover (naturer6) image message in the cover object. The proposed method was simulated by encoding the secret message within the colour cover image, and then histogram of colour stego image was noted. From this histogram it will inferred that histogram of

image proposed

Mean

192.5268259

192.6870127

192.5261886

Variance

70.92776542

71.04948301

70.92728542

RMS

205.1763305

205.3687267

205.1755666

From the Table.2 it is inferred that the proposed stego image values are nearly equal to the cover image values. These values are also compared with the existing LSB steganography method. Hence, statistical analysis proves that it not possible to identify the proposed stego image. It also implies that it is not possible to recover the secret message from the stego image. Table 3 Colour differences of colour image Fig.9 Comparison of histogram with (a) cover image and (b) stego image

Fig.10 Comparison of (a) cover image and (b) stego image

Fig.3 Lena cover image and Histogram of Lena cover image

Red

Green

Blue

Cover(nature6)image

3.0777

0.7292

1.7444

Stego image

3.0779

0.7294

1.7446

The RGB colour percentage present in the cover image before and after encoding of the secret message was noted. Then the differences between the RGB values in the colour image before and after encoding were tabulated. Table 3 shows the results of colour differences. It was inferred that the differences in the cover image and stego image was minimum.

4

IEEE TRANSACTIONS ON XXXXXXXXXXXXXXXXXXXX, VOL. #, NO. #, MMMMMMMM 1996

CONCLUSION In this work, a modulo based image steganography algorithm was proposed for both colour and black-n-white images. The proposed algorithm was stimulated with secret data using Lena image as the cover image. The resultant stego image obtained after embedding of the secret message does not show any change when compared to original cover image. Histogram and statistical analysis were performed on the stego image and proved that the proposed method can effectively resist image steganalysis. Comparision of the statistical values like mean, variance and RMS for proposed method results with existing LSB steganography method was also done. From these statistical analyses, it was also inferred that changes in values were obtained only in the fourth or fifth decimal places, thereby not affecting the appearance of the image. Thus, the proposed method provides greater security for the hidden data.

REFERENCES [1]

H. Zhang and H. Tang, “A Novel image steganography algo rithm against statistical analysis,” proceeding of the IEEE, vol. 19, no. 22, pp. 3884-3888, Aug. 2007.

[2]

J. Fridrich, M. Goljan, and R. Du, “Detecting LSB Steganography in Color and Gray-Scale Images,” Magazine of IEEE Multimedia Special Issue on Security, pp. 22-28, Nov. 2001.

[3]

F. Petioles, J. Anderson, and G. Kuhn, “Information hiding-A survey,” proceeding of the IEEE, vol. 87, no. 7, pp.1062-1078, June 1999.

[4]

Gougelet Pierre-emmanuel.: http:// www.xnview.com.

[5]

J. Fridrich and M. Goljan, “Practical Steganalysis of Digital Images – State of the Art”, Proceeding of SPIE, Photonics West, Vol. 4675, Electronic Imaging 2002, Security and Watermarking of Multimedia Contents, San Jose, California, pp. 1-13, Jan. 2002.

[6]

H. Farid, “Detecting hidden messages using higher-order statistical models”, Proceeding of the IEEE, vol. 15, no.6, pp.68-72, Oct.2002.

XnView,

Software available at

Dr. V. Vijayalakshmi is working as Senior Lecturer in the Department of Electronics and Communication Engineering at Pondicherry Engineering College, Puducherry, India. She received her PhD in Information Security. She teaches courses on Information Security for both under graduate and post graduate engineering students. She has authored 6 international journals and 15 international and national conferences. Her research areas of interest include Cryptography and Network Security, VLSI and ASIC Design. Dr. G. Zayaraz is working as Assistant Professor in the Department of Computer science and Engineering at Pondicherry Engineering College, Puducherry, India. He received his PhD in Software Architecture. He has authored more than 13 research papers both in international conferences and reputed journals. His research areas of interest include Information Security and Software Architecture. V. Nagaraj pursued his B.E in Electronics and Communication Engineering degree in I.F.E.T college of Engineering ,Villupuram ,Anna University, and at present he is pursuing his M.Tech in Wireless Communication at Pondicherry Engineering College, Puducherry.