A Novel Steganographic Technique Based on LSB-DCT Approach

8/2/2019 A Novel Steganographic Technique Based on LSB-DCT Approach

1/5

ISBN: 9788192211947 1

Proceedings of the National Conference on Emerging Trends in Information and ComputingTechnologies(NCETICT-2012), 30

thMarch, 2012.

A Novel Steganographic Technique Based on

LSB-DCT Approach

Mohit Kumar GoelDepartment of Electronics & Electrical Comm. Engineering

PEC University of Technology

Chandigarh, [email protected]

Neelu JainDepartment of Electronics & Electrical Comm. Engineering

PEC University of Technology

Chandigarh, [email protected]

AbstractSteganography is the art of concealing data withinother digital media like an image or an audio signal for providing

higher security. The security includes both imperceptibility andundetectability. But most of the steganographic methods dont

pay enough attention to the undetectability. In this paper, wepropose a novel DCT-based steganographic method for hidingthe data. Each bit of data is embedded by altering the leastsignificant bit of low frequency DCT coefficients of cover image

blocks. The experimental results show that this algorithm hasbetter PSNR value and high capacity in comparison to othertechniques such as LSB, modulus arithmetic, SSB4-DCT. It also

maintains satisfactory security as secret message cannot beextracted without knowing the decoding algorithm.

Keywords: Steganography; Discrete Cosine Transform; Zigzagscanning; Data hiding.

I. INTRODUCTIONWith the development of science, digital media can be

transmitted conveniently over the internet. The security of the

data is essential issue for the internet. Encryption is introduced

for the data security. Encryption scrambles the secret messageso that it cannot be understood. But, it makes the messagesuspicious enough to attract eavesdroppers attention. The

commonly used encryption schemes include DES (DataEncryption Standard) [1], AES (Advanced Encryption

Standard) [2] and RSA [3]. After encryption watermarking isintroduced. In watermarking technique the owner property isprotected by some hidden watermarks. A new scheme, calledsteganography [4], is proposed to conceal the secret

messages within some image, music or audio file so that it isnot visible to others. Steganography differs from cryptographyin the sense that where cryptography focuses on concealing

the contents of a message, steganography focuses on

concealing the existence of a message [5]. The wordsteganography in Greek means covered writing (Greekwords stegos meaning cover and grafia meaningwriting). Steganography is the art and science of hidinginformation in a cover document such as digital images in away that conceals the existence of hidden data. The main

objective of steganography is to communicate securely in sucha way that the true message is not visible to the intruder.Image steganography schemes can be divided into twocategories: Spatial Domain and Frequency Domain.

A. Spatial domain steganography

Spatial domain techniques embed messages in the intensityof the pixels directly [6][7][8]. Least Significant Bit (LSB) isthe first most widely used spatial domain steganography

technique. It embeds the bits of a message in the LSB of theimage pixels [9][10]. But the problem with this technique isthat if the image is compressed then the embedded data maybe lost. Thus, there is a fear for loss of data that may have

sensitive information [11]. LSB has been improved by using aPseudo Random Number Generator (PRNG) and a secret keyin order to have private access to the embedded information[12]. The embedding process starts with deriving a seed for aPRNG from the user password and generating a random walkthrough the cover image that makes the steganalysis hard.

Another recent improvement based on random distribution ofthe message was introduced by M. Bani Younes and A. Jantan[13]. In this method they utilize an encryption key to hideinformation about horizontal and vertical blocks where the

secret message bits are randomly concealed. SSB-4steganography approach introduced by Rodrigues, Rios andPuech is about changing the 4 thbit of a pixel in the original

image according to the bit message. Then modify the otherbits (1st, 2nd, 3rdand/or 5th) to minimize the difference betweenthe changed pixel value and the original one [14]. The 4 thdigitis a significant bit and if the image is compressed the

embedded information is not destroyed [15]. Tu C. and Tran TD. argued that the difference must be equal or less than four(i.e., 4) [16]. The 4thbit was chosen because it satisfies that

changing of 4 units in the channel color value isimperceptible to human eyes, and it is the most significant bitwhich provides the minimum change in the pixel values.

Modulus arithmetic steganography proposed by Sayuthi Jaafarand Azizah A Manaf has calculated last four bits of each pixelby mod-16 operation. Then these bits are replaced with databits [8]. In this the amount of the data that can be embedded is

more but stego image has less PSNR value than LSB andSSB-4 techniques.

B. Frequency domain steganography

In frequency domain, images are first transformed and thenthe message is embedded in the image [17][18][19]. When thedata is embedded in frequency domain, the hidden data resides


2/5

2


thMarch, 2012.

in more robust areas, spread across the entire image, andprovides better resistance against statistical attacks. There aremany techniques used to transform imagefrom spatial domainto frequency domain. The most common frequency domainmethod usually used in image processing is the 2D discretecosine transform [20][21]. In this technique the image isdivided into 88 blocks and DCT transformation on each

block is performed. DCT arranged the pixel of imageaccording to their frequency value. The data bits areembedded in the low frequency coefficients of DCT. SSB-4 &

DCT steganography proposed by Nedal M. S. Kafri and HaniY Suleiman uses DCT approach with SSB-4 technique [21].But in this stego image PSNR value is not so high. To improveit, a novel LSB & DCT based steganographic method for is

proposed in this paper, which can not only preserve goodimage quality, but also resist some typical statistical attacks.

II. PROPOSED STEGANOGRAPHY METHODThe challenge in this work was to find a way to

camouflage a secret message in an image without perceptible

degrading the image quality and to provide better resistanceagainst steganalysis process. Therefore, a combination offrequency domain by means of DCT and LSB technique ofspatial domain steganography has been used to hide data. Twodimensional DCT converts the image block from spatial

domain to frequency domain and then data bits are embeddedby altering LSB of DCT coefficients.

Figure. 1 Block diagram of LSB-DCT steganography

A. Discrete Cosine Transform

The image of size MN is divided into 88 blocks and twodimensional (2-D) DCT is performed on each block. The DCT

is calculated as follow:

)1(16

)12(cos

16

)12(cos),()()(

4

1),(

7

0

7

0

yuxyxfvCuCvuF

x y

for x=0,..., 7 and y=0,..,7

otherwise

kforkCwhere

1

02/1)(

In DCT block lower frequency cofficents are at upper left

positions and high frequency coefficients are lower rightpositions. Low frequency coefficients are of larger value thanhigh frequency coeffcients. An example of a 88 block of

DCT cofficient is shown in fig. 2.

05026711

27186224

28220210

6164151731

31532258338

7363143126094

28527321010830

11230722040162

F

Figure. 2 DCT coefficient

B. Quantization

The 8 x 8 block of DCT coefficients is compressed byquantization. A useful feature in this process is the image

compression and quality is obtainable through selection of

specific quantization table. The standard quantization matrix[27] is shown in fig. 3.

9910310011298959272

10112012110387786449

921131048164553524

771031096856372218

6280875129221714

5669574024161314

5560582619141212

6151402416101116

Q

Figure. 3 Quantization Matrix

Quantization is achieved by dividing each element in theDCT coefficient block by the corresponding value in thequantization matrix, and the result is rounded to the nearestinteger. As eye is not able to discern the change in high

frequency components so these can be compressed to largerextent. Lower right side components of quantization matrixare of high value so that after quantization high frequencycomponents become zero. The quantized DCT coefficients

matrix P is computed by (2) and shown in fig. 4.


3/5

3


thMarch, 2012.

)2(),(

),(),(

vuQ

vuFvuP

00000000

00000000

00000000

00000012

00001053

00012157

00012193

000152410

P

Figure. 4 Quantized DCT Block

C. Zigzag Scanning

Although the DCT coefficients have been decorrelated byDCT transform to some extent, DCT coefficients in the sameblock are still not independent, which is called as intra-block

correlation [16]. While neglecting the impact of block edge,the general trend in magnitude of the block coefficients ineach block is non-increasing along zigzag scan order. After

block DCT coefficients are arranged by zigzag scan pattern,dependencies among neighboring coefficients in bothhorizontal and vertical directions can be convenientlyinvestigated [23]. For example, the sequence pairs (5,6) and

(14,15) describe the correlations along horizontal direction,while the sequence pairs (2,3) and (10,20) describecorrelations along vertical direction. Zigzag scan converts 88block into 64 elements one dimensional array.

20

1910

18119

171283

1613742

15146510

Figure. 5 Zigzag scan order

D. Data embedding

Data bits are concealed by altering the LSB of elements of

zigzag array.

a)If data bit is 0, then make the DCT coefficient even or,b)If the data bit is 1, then make the DCT coefficient odd.

E. Dequantization and inverse DCT

After embedding data zigzag array is again converted into88 block. These blocks are dequantized and inverse DCT is

performed. The entire 88 blocks are combined to form thestego image which is then sent to receiver. Completeembedding algorithm is given as follow:

Input: An MN size cover image and data to be concealed.

Output: Stego image.

Step 1: Divide the cover image into 88 blocks.Step 2: Perform 2-D DCT on each block.

Step 3: Perform quantization on each block.Step 4: Perform zigzag scan to convert 88 block into onedimensional array.Step 5: Replace the LSB of DCT coefficients with data bits.Step 6: Convert 1-D zigzag array back to 88 block.Step 7: Perform Inverse DCT on each block.Step 8: Combine all the blocks to form stego image.

F. Extraction of secret message

The stego-image is received in spatial domain. Now stegoimage is divided into 88 blocks and DCT is performed oneach block. Then scan the DCT block in zigzag way and

extract the embedded data. Extraction algorithm is given asfollows:

Input: Stego image.

Output: Secret message.

Step 1: Divide the stego image into 88 blocks.Step 2: Perform 2-D DCT on each block.Step 3: Perform quantization on each block.Step 4: Perform zigzag scan to convert 88 block into one

dimensional array.Step 5: Check the DCT coefficient.

a) If DCT coefficient is even then data bit is 0 or,b) If DCT coefficient is odd then data bit is 1.

Step 6: Concatenate the bits to obtain secret message anddisplay it on screen.

III. EXPERIMENTAL RESULTSSince the visual detection of stego images is depending on

the nature of the image [24] so, varieties of image categoriesare utilized in the experiments. The experimental image dataset consists of 75 JPEG images, which were taken by digital

camera. We focused on short messages with length of 1400bits because they are the most challenging to detect [24].Comparative analysis of LSB, Modulus arithmetic (mod-16),

SSB4-DCT and LSB-DCT method has been done on basis ofPeak signal to noise ratio (PSNR). To calculate PSNR, firstMSE is calculated as follow:

)3(),(),(1 2

1

0

1

0

m

i

n

j

jiKjiImn

MSE

Where MSE is the Mean Squared Error of Original image(I) and stego image (K). Thereafter PSNR value is calculated

as follow:

)4(log.20log.10 10

2

10

MSE

MAX

MSE

MAX

PSNRii

Where, MAXi is the maximum pixel value of the image. In

other words MAXi = 2b 1, where b is the bit depth of the

original image (e.g., MAXi = 255 in the case of 8 bits depthgrayscale images). PSNR computes the peak signal to noise

ratio, in decibels, between two images. This ratio is used as aquality measurement between two images. The comparativeanalysis of PSNR value of different steganography technique,is given in table 1, shows that LSB-DCT steganography has

better image quality of stego image than other techniques.


4/5

4


thMarch, 2012.

(a) Original Tree.jpg (b) Stego Tree.jpg

(c) Original Human.jpg (d) Stego Human.jpg

(e) Original Temple.jpg (f) Stego Temple.jpg

(g) Original Flower.jpg (h) Stego Flower.jpg

Figure. 6 Original Images and Stego Images using LSB-DCT steganography


5/5

5


thMarch, 2012.

Table 1. Comparative analysis of PSNR values of differentsteganography techniques

Image

PSNR Value

LSBModulus(mod-16)

BIT-4&DCT

LSB -DCT

Tree.jpg 50.10 48.23 52.37 54.43

Human.jpg 52.54 50.53 54.37 56.46

Temple.jpg 50.43 48.77 51.56 54.68

Flower.jpg 53.46 50.46 54.57 57.51

IV. CONCLUSIONIn this paper a mixed approach that applies the spatial

domain with the frequency domain steganography techniques

has been proposed. The average PSNR value of 75 imagesusing LSB-DCT is above 55. The obtained experimentalresults indicate that, the proposed method is a good andacceptable steganogaphy scheme. Furthermore, by embedding

information in the least significant bits of the DCT domain,

the hidden message resides in more robust areas, spread acrossthe entire stego image, and provides better resistance against

statistical attacks than other techniques. The future work mayfocus on the improvement and further development in thistechnique.

V. REFERENCES[1] DES Encryption Standard (DES), National Bureau of Standard (U.S.).

Federal Information Processing Standards Publication 46, NationalTechnical Information Service, Springfield, VA, 1997.

[2] Daemen J., and Rijmen, V. Rijndael: The Advanced EncryptionStandard, Dr. Dobbs Journal, March 2001.

[3] R. Rivest, A. Shamir, and L. Adleman, A method for obtaining digitalsignatures and public-key cryptosystems. Communication of the ACM,

pp. 120-126, 1978.[4] Pfitzmann, B. Information hiding terminology, Proc. First Workshop

of Information Hiding Proceedings, Cambridge, U.K., Lecture Notes inComputer Science, Vol.1174, pp. 347-350, 1996.

[5] H & Wang, S, Cyber warfare: Steganography vs. Steganalysis,Communications of the ACM, 47, pp.10-12, October 2004

[6] Chan, C.K. and Cheng. L.M. Hiding data in image by simple LSBsubstitution. Pattern Recognition, 37, pp. 469 474, 2004.

[7] Chang,C.C and Tseng, H.W. A Steganographic method for digitalimages using side match. Pattern Recognition Letters, 25, pp. 1431

1437, 2004.[8] Sayuthi Jaafar, Azizah A Manaf, Akram M Zeki, Steganography

Technique using Modulus Arithmetic, 9th International Symposium onSignal Processing and Its Applications, pp. 14, April 2007.

[9] W. Bender, D. Gruhl, N. Morimoto, and A. Lu, Techniques for DataHiding, I.B.M. Systems Journal, 35(3-4): pp. 313-336, 1996.

[10]N. Nikolaidis, and I. Pitas, Robust Image Watermarking in the SpatialDomain, Signal Processing, 66(3), pp. 385-403, 1998

[11] T. Morkel, J. Eloff, and M. Olivier, An overview of imagesteganography, In Proceedings of the Fifth Annual Information

Security South Africa Conference (ISSA2005), (Sandton, South Africa,Jun/Jul. 2005).

[12] J. Fridrich, M. Goljan, Steganalysis of JPEG Images: Breaking the F5Algorithm, Publisher: Springer Berlin, Heidelberg, Lecture Notes inComputer Science, Vol. 2578, pp 310-323, 2003.

[13] M. A. Bani Younes, A. Jantan, A New Steganography Approach forImage Encryption Exchange by Using the Least Significant BitInsertion, IJCSNS, International Journal of Computer Science and

Network Security, Vol. 8 No. 6, June 2008.[14] J. Rodrigues, J. Rios, and W. Puech SSB-4 System of Steganography

using bit 4, In International Workshop on Image Analysis forMultimedia WIAMIS, May, 2005.

[15] J. Fridrich, and M. Goljan, Practical steganalysis: state-of the-art, InProceeding of SPIE Photonics West, Electronic Imaging 2002, volume

4675, pp. 1-13, 2002.[16] Tu C. and Tran T D. Context based entropy co ding of block transform

coefficients for image compression, IEEE Transaction on ImageProcessing, Vol.11, No.11, November, 2002.

[17] Wenqiong Yu, Blind Detection for JPEG Steganography, InternationalConference on Networking and Information Technology, pp. 128-132,

July 2010.[18] Chung, K.L., Shen, C.H. and Chang, A novel SVD - and VQ-based

image hiding scheme.Pattern Recognition Letters, 22, pp. 1051 1058,2001.

[19] Iwata M., Miyake K., and Shiozaki, Digital Steganography UtilizingFeatures of JPEG Images, IEICE Transfusion Fundamentals, E87-A, 4,

pp. 929936, 2004.[20] M. Kharrazi, H. Sencar, and N. Memon, Performance study of common

image steganography and steganalysis techniques, Communications ofthe SPIE and IS&T, 15, No.4, pp. 1017-9909, Oct-Dec., 2006.

[21] Nedal M.S. Kafari, Hani Y. Suleiman, Bit-4 of Frequency DomainDCT Steganography Technique, FIRST NATIONAL CONFERENCE ON

NETWORKED DIGITAL TECHNOLOGIES, PP.286-291,2009.[22] Dr. Ekta Walia, Payal Jain, Navdeep, An Analysis of LSB & DCT

based Steganography, Global Journal of Computer Science andTechnology, Vol. 1, pp. 4-8, April, 2010.

[23]

Zhiping Zhou and Maomao Hui,Steganalysis for Markov Feature of

Difference Array in DCT Domain, Proceedings of Sixth International

Conference on Fuzzy Systems and Knowledge Discovery, pp. 581 -584 , Aug. 2009.

[24] L. Davidson, and P. Goutam, Locating secret message in images, InACM SIGKDD international conference on Knowledge discovery and

data mining, (Seattle, Washington, Aug.22-25. ACM 1-58113-888-1,2004.

[25] Gonzalez, R.C. and Woods, R.E., Digital Image Processing usingMATLAB, Pearson Education, India,2006.

[26] Jayaraman, S., Esakkirajan, S. and Veerakumar, T. Digital ImageProcessing, Tata McGraw Hill Education Private Limited, India, 2009.

[27] www.vision.arc.nasa.gov/dctune/
http://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=4542559http://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=5497982http://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=5497982http://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=5497982http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=Authors:.QT.Zhiping%20Zhou.QT.&newsearch=partialPrefhttp://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=Authors:.QT.Maomao%20Hui.QT.&newsearch=partialPrefhttp://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=Authors:.QT.Maomao%20Hui.QT.&newsearch=partialPrefhttp://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=Authors:.QT.Zhiping%20Zhou.QT.&newsearch=partialPrefhttp://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=5497982http://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=5497982http://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=5497982http://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=5497982http://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=4542559

A Novel Steganographic Technique Based on LSB-DCT Approach

Documents

Transcript of A Novel Steganographic Technique Based on LSB-DCT Approach