COMPARATIVE ANALYSIS OF IMAGE STEGANOGRAPHY USING LSB,DCT AND DWT TECHNIQUES

COMPARATIVE ANALYSIS OF IMAGE STEGANOGRAPHY USINGLSB,DCT AND DWT TECHNIQUES

Chinmay A V1, Lekshmi D2, Neethu P V3 and RemyaAjai A S4

Department of Electronics and Communication Engineering, Amrita School of Engineering, Amritapuri

[email protected], [email protected],[email protected] and [email protected]

Abstract. Steganography is the technique of hiding secret message within another message such that no one except the sender and the intended receiver suspects the existence of a secret message [1]. This hidden message can be plain text or even images. Image steganography is the art of inconspicuously hiding image within an image. In today’s world, there are a number of legitimate uses of steganography.This paper aims to provide a comparative analysis between various existing techniques such as LSB ( Least Significant Bit ) ,DCT ( Discrete Cosine Transform ) and DWT ( Discrete Wavelet Transform ) on four different aspects namely, undetectability, robustness, capacity and PSNR value. Our analysis shows that DWT is comparatively better. We also performed the noise analysis for the above methods. We used MATLAB version R2010 for our entire analysis.

Keywords: Image, Steganography, LSB, DCT, DWT, Undetectability, Robustness, Capacity & PSNR.

1. INTRODUCTION

Increasing long distance communications advancements in technologies andincreasing awareness of importance of interception issues has dragged inthe issue of secure communication. But even when faced with the mostubiquitous of modern surveillance, there are ways to keep yourcommunications away from prying eyes. Thus steps in the art of invisiblecommunication leads to Steganography. The secrets can be hidden inside allsorts of cover information: text, image, audio, video, and so on. Moststeganographic utilities hide information inside images, as it isrelatively easy to implement. The idea is to hide a secret image withinanother giving us a stego image. The process of hiding is such that thepixels of the secret image blend with the pixels of the cover image andescape detection. Unlike cryptography, steganography hides the veryexistence of communication that is, unless one can prove the data is there(which is usually not easy), it is deniable that the file contains any.

2. IMAGE STEGANOGRAPHY TECHNIQUESBased on the different techniques available, there are mainly 2classifications of image steganography namely spatial domain technique andfrequency domain technique [2].

2.1. Spatial Domain Based Steganography

In this technique, more importance is given to pixel variation and allprocedures are directly performed on these pixel values. This mainly usesLSB based technique where LSB’s of each byte are changed according to thesecret image. The main disadvantage of this technique is that it is proneto attacks.

2.2. Transform Domain Based Steganography

In this method, all embedding is done in transform domain. When data isembedded in transform domain, hidden data is evenly spread across allarea. This it provides better resistance to statistical attacks. Two maintechniques under transform domain namely Discrete Cosine Transform andDiscrete Wavelet Transform

3. LSB (LEAST SIGNIFICANT BIT) TECHNIQUE

LSB technique is one of the simplest techniques available under spatialdomain techniques. In this technique, cover image is the main image onwhich the hidden data is embedded forming the stego-image. Here a secretkey (sohel) is used for embedding.

3.1. Method for Image Hiding Using LSB[3]

In the embedding process, first the cover image is broken down into 3planes. They are red, blue and green. The secret key is converted into 1Darray. Each bit of secret key is xored with the LSB of red matrix and theresult determines whether replacement is to be done on the green matrixLSB or blue matrix LSB. This is continued till the last bit of the hiddeninformation is taken.The flowchart to hide the information is Figure 1.

3.2 Method for Recovery ofMessage Using LSB

For the retrieval process, stego-image is divided into 3 matrixes, namelyred, blue and green. The secret key is converted into 1 D array and eachLSB of the red matrix of stego-image is xored with the each bit of secretkey whose result determines whether to hide in the blue or green matrix.This recovery technique is continued till the last bit of the hidden

information is encountered. The flowchart for the recovery of the hiddeninformation is shown in Figure 2.

3.3 Limitations of LSB Based Technique

This method made use of the advantage that the LSB of image pixels containhigh frequency information. Comparing to other similar techniques, therobustness of the stego image is quite low for this method. Also, thismethod is quite vulnerable to accidental attacks.

Fig 1. Flow chart to hide information to cover image

Fig 2. Flow chart to retrieve information from cover image

4. DCT (DISCRETE COSINE TRANSFORM) TECHNIQUE

A discrete cosine transform (DCT) expresses a sequence of finitely manydata points in terms of a sum of cosine functions oscillating at differentfrequencies. To calculate DCT coefficients,

D=CXCT(1)where, C is the coefficient matrix and X is the input signal. In pure DCT-based steganography, the DCT coefficients are modified to embed the hiddenimage.

4.1. Method for Data Hiding Using DCT [4]

In this method, the cover image is read first. The secret image is then read and converted into binary stream. The cover image is broken into 8×8blocks of pixels. DCT is applied to each block. Now, replace each bit of secret message in middle and high frequency coefficients of DCT. Stego image is then written.

4.2. Method for Message Recovery Using DCT

For retrieving, the stego image is read first. It is then broken into 8×8 block of pixels. DCT is applied to each block. Each LSB bit is retrieved

from middle and higher frequency. After that all the bits are grouped and the image is thus retrieved.

5. DWT (DISCRETE WAVELET TRANSFORM) TECHNIQUE

Wavelet transform provide time frequency localisation. Hence it hasadvantages over fourier transform and short time Fourier transform. DWTinvolves filtering and downsampling. Based on different wavelets, DWT canbe applied for various image processing techniques. In our analysis, wetook Haar wavelet.

5.1.Method for Data Hiding Using DWT [5]

In this method, the original image is read and then the hidden image and thescaling factor are identified. Obtain 2 level wavelet transform for originalimage and 1 level wavelet transform for hidden image. The value of originalimage coefficient, lr is obtained. Similarly the value of hidden imagecoefficient, Ww is also calculated. To the wavelet coefficients of originalimage, the hidden coefficients are embedded using the equation,

lw=lr (1+αWw )(2)

where, α is the scaling factor to maintain the invisibility of the hiddenpixels in the smooth part of the cover image. The value may range between0.2 and 0.9. Inverse DWT is applied t the image whose coefficients arechanged using the above equations. Then the hidden image is displayed. Theblock diagram of this method is shown in Figure 3.

Fig 3. Block diagram of steganography using DWT

5.2. Method for Retrieval of Image

For retrieving, the watermarked image is read. Then the wavelet transformof both images are obtained at 2 levels. From the highest value waveletcoefficients of the low resolution representation of the image, theapproximation coefficients of the watermark is found using the equation,

Ww=(1α )∗IwIr

−1(3)

DWT is then applied to the watermarked image to obtain the original image.The block diagram for image retrievel is shown in figure 4

Fig 4. Block diagram for retrieving image from the cover image

6. PARAMETERS USED FOR ANALYSING THE DIFFERENT TECHNIQUES

Comparative analysis of LSB, DCT and DWT is being done on the basis of 4major factors namely undetectability, robustness, payload capacity andPSNR.

6.1. Undetectability:

It represents the ability to avoid detection of hidden image. Themathematical measure used to represent this parameter is “K L distance”.It is a parameter form information theory and coding. It is given by theformula,

D(Px||Py)=Ʃ[Px *log(Px/Py)](4)

where, Px is the probability density of cover image and Py is theProbability density of stego image.

6.2. Robustness

It measures the ability of steganography system to survive attempts ofremoval of hidden information. It is mathematically represented by theparameter “CORRELATION FACTOR”, given by

ρ (w,w )=∑i=1

Nwiwi

√∑i=1Nwi

2√∑i=1

Nwi

2

where,wi=hiddenimagewi=extractedhiddenimage

(5)

6.3. Payload capacity

It represents the amount of information that can be hidden. It is given by the formula

Capacity=Totalno:ofbitsembedded∈thecoverimage

Totalno:ofbits∈thecoverimage (6)

6.4. MSE (Mean Square Error)

It is used to find the error between the cover image and the stego-image which is given by,

MSE=1mn∑i=0

m−1

∑j=0

n−1

[I (i,j )−K(i,j)]2(7)

6.5. PSNR:

It represents the quality of the cover image and shows how easily the secret message can be reconstructed.

PSNR=10log10(MAX2

MSE )=20log10( MAX√MSE )(8)7. EXPERIMENTAL RESULTS

Experimental results for the three methods are as follows. We have chosena water fall image as the cover image in which an image displaying themessage “meet me at park” is hidden.

7.1. Result for LSB Technique

Fig 5.Observed result after embedding the secret imageusing LSB technique

Fig 6. Observed image after retrieving the secret image using LSB technique.

7.2. Result for DCT Technique

Fig 7. Observed result after embedding the secret image using DCT technique

Fig 8.Observed image after retrieving the secret image using DCT technique.

7.3. Result for DWT Technique

Fig 9. Observed result after embedding the secret image using DWT technique

8. ANALYTICAL RESULTS

We anlayse the different parameters for the above three techniques and thefollowing results are recorded.

Table 1.Comparison of different techniques

From the table, it is clear that through DWT technique, we can obtain highinvisibility, large capacity, high robustness. But however, the PSNR valueis very low.

9 NOISE ANALYSIS

Fig 10 shows that for LSB and DCT technique the image cannot be retrieved after adding Gaussian noise whereas DWT can recover it. Figure 11reveals that LSB technique can recover the image affected by salt and pepper noiseto a very small extent whereas DWT can recover the entire image perfectly.But for DCT, the effect of salt and pepper noise is still worse.

Fig 10. Observed result after adding Gaussian noise

Fig 11. Observed result after adding salt and pepper noise

10. CONCLUSION

This paper includes detailed comparison between LSB,DCT and DWTsteganography techniques. We also performed noise analysis for threedifferent techniques. Since Gaussian noise and Salt & Pepper noise aremore prone to occur in image communication systems, our noise analysis islimited to these two noises. We found that DWT technique has comparativelybetter values for the parameters undetectability, capacity and robustness.LSB technique shows somewhat better recovery from only salt and peppernoise, but DCT technique is getting affected by both the noise. Theability for recovering the two noises by DWT technique made us to concludethat DWT can be considered as a better method. However, there is a concernwith PSNR value obtained for DWT based technique. Our future work willfocus on identify an algorithm to modify DWT so that acceptable PSNR rangecan be obtained.

References

1.TMorkel, J H P Eloff, M S Olivier, “An Overview of Image Steganography”, Information and Computer Security Security Architecture Research Group, University

of Pretoria, South Africa, In Proceedings of the fifth Annual Information SecuritySouth Africa Conference(ISSA),June/July 2005.

2. ShikhaSharda,SumitBudhiraja,”Image Steganography- A Review”, International Journal of Computational Engineering Research,Volume 3, Issue 1,January 2013.

3. S. M. MasudKarim, Md. SaifurRahman, Md. Ismail Hossain,"A New Approach for LSB Based Image Steganography using Secret Key“, Computer Science and Engineering Discipline,14th International Conference,KhulnaUniversity,Bangladesh,December 2011.

4. GurmeetKaur and AartiKochhar,”A Steganography Implementation based on LSB and DCT”, DAVIET Jalandhar,Vol2,issue 12,December 2013

5. Monika Patel ,PritiSrinivas,Jigar Patel,”Ënhancement of DWT based Watermaking Technique for Images”, International Journal of Advanced Research in computer and communication Engineering Vol2, Issue 12, December 2013.

6.StutiGoel,ArunRana,ManpreetKaur,”Comparison of Image Steganography Techniques”, International Journal of Computers and Distributed Systems, Vol No.3,Issue 1, April-May 2013.