Sunil Karforma Associate Professor Dept. of Computer Science The University of Burdwan Burdwan, West Bengal, India

Transcription

1 Volume 4, Issue 8, August 2014 ISSN: X International Journal of Advanced Research in Computer Science and Software Engineering Research Paper Available online at: Analysis of Least-Significant-Bit and Pixel-Value-Difference Steganography, an E-Governance Data-Security Issue Tanmoy Halder Asst.Professor Dept. of Computer Application Dr.B.C.Roy Engg. Coll. Durgapur West Bengal, India Sunil Karforma Associate Professor Dept. of Computer Science The University of Burdwan Burdwan, West Bengal, India Rupali Mandal Asst.Professor Dept. of Computer Application Bengal College of Engg.& Tech. Durgapur, West Bengal, India Abstract steganography, an efficient way of digital data hiding have travelled a long path by hiding information within image, audio, text and video. E-governance, one of the most popular ways of interaction between mass and government has been benefited by steganography algorithms. In this paper we have discussed, compared and analyzed two most popular steganographic algorithms LSB (least significant bit)-replacement and PVD (Pixel value differencing) from spatial domain, which are widely used to hide secret E-governance information. In this paper we have discussed, compared and analyzed those two algorithms. During comparison we have reviewed capacity, PSNR and Histogram for each of the algorithms. Keywords: Steganography, LSB, Pixel-value-differencing, Data-hiding. I. INTRODUCTION In the modern age of ICT(Information and communication technology) driven digital communication, security has become an important part to protect the contents against attacks caused by hackers using loopholes of internet. Cryptography, an already established security process which guard against data-stealing has become obsolete due to a disadvantage. Presence of secret information within a cover image through cryptography is detectable. On the other hand steganography hide data within the cover image in such a way that nobody could detect the presence of hidden message. E-governance process, which communicates between government-sector and other sectors, also suffers from insecurity of important data. Steganography could be a solution against these kinds of threats. A. Introduction to Steganography Steganography works with image, audio, video and text. Here we have discussed only about imagesteganography. In Steganography the image within which the secret data is hidden is known as cover image, and the encrypted image with secret data is known as stego-image. While embedding secret information Image steganography read the cover image data (in binary or decimal) and embed the secret data within those data in such a way that it could not be detected by human vision, to check the visual distortion between cover image and stego image histogram analysis is an essential tool. Apart from visual distortion statistical distortion is another important factor which could be detected by some statistical tools (Chi-square test, R-S steganolysis). To measure the distortion amount between cover image and stego-image MSE (Mean Square Error) and PSNR (Peak Signal to Noise Ratio) is used. While comparing we have used Histogram to compare visual distortion, payload to check capacity and PSNR to check statistical distortion. Fig [1] shows a basic steganographic process using diagram. Fig 1.Embedding procedure of steganography 2014, IJARCSSE All Rights Reserved Page 1150

2 Rest of this paper is organized as follows: section II discuss about LSB steganography, sec III PVD steganography sec IV results and discussion and sec V is about conclusion. II. OVERVIEW LSB-REPLACE/MATCH STEGANOGRAPHY This is the simplest steganographic method based on the use of LSB, The process is as follows: Step1: extract pixel value from secret message and convert it in binary. Step 2: Hide bits from secret message within the LSB bits of the cover image. Step 3: Convert the new stream of bits in decimal. Step 4:The new decimal value is the new pixel for stego image. Step 4: Continue the process until end of secret message. As we know LSB side of a pixel value contains less important bits altering those bits will not affect the actual pixel much more. Altering LSB most bit will only increment or decrement the number by one. For increasing capacity we may alter more bits from LSB side of a pixel value. The major disadvantage with LSB steganography is as this is a very simple method therefore the most vulnerable. The embedding process consists of the sequential substitution of each least significant bit of the image pixel for each bit of the message. For its simplicity, this method can camouflage a great volume of information. A. Literature survey of LSB steganography Sharp.T in 2001 proposed simple Least Significant Bit (LSB) steganography, long-known to steganographers, in which the hidden message is converted to a stream of bits which replace the LSBs of pixel values in the cover image. [1] NEIL F. Johnson and Sushil Jajodia, Discuss three popular methods for message concealment in digital images. These methods are LSB insertion, masking and filtering and algorithmic transformations [2].(Chang et al., 2002)(Thein et al., 2003) had insinuate, LSB substitution is basically a method to directly switching the LSBs of pixel in the cover image with secret bits to get the stego-image. This method replaces the LSB of each pixel with the encrypted message bit stream to hide message in a host image. Authorized receivers can extract the message by decoding the host image with the help of a pre-shared key. Capacity of algorithm is 1 bit per pixel. [3]-[5](Wang et al., 2000 and 2001) examined on to improve the quality of stego-image and employed a genetic algorithm to generate a substitution table. The substitution table consists the value of the secret data to be embedded into each host pixel is transformed to another value in advance which is closer to the original value of the host pixel. But substitution table may not be the optimal solution. [6] To find out the optimal solution (Chang et al., 2003 and 2006) proposed dynamic programming [4, 7] strategy. But the optimal substitution process may require huge computational cost because of using genetic algorithm and dynamic programming strategy. Rather to use genetic algorithm, an Optimal Pixel Adjustment Process (OPAP) is used to enhance the visual quality of the stego-image by LSB substitution. (Ker et al., 2004) was established the LSB matching scheme. It is a different technique than LSB Substitution; it modifies the LSBs of the cover image. If one secret bit does not match the LSB of the cover image, then another one will be randomly added or subtracted from the cover pixel value. [8, 9](Mielikainen, 2006) proposed a modified version of LSB matching, which progress it by lesser the expected number of modifications per pixel (ENMPP), from 0.5 to 0.375, so the histogram is less significant. [10] (LI, 2009) presented the simplification of LBS matching, in which sum and difference covering set of finite cyclic group were used to more decrease ENMPP and providing improved protection. [11].These techniques had some problems, mainly the artificial noises in the smooth regions of the image, which damage the visual quality of the stego-image. Not all pixels in the image can bear same length of changes without clear alteration, and as a result the stego-image has low quality. After research done, scheme establish that, if an image is processed with simple LSB substitution the histogram of the image will be showed in a pair-wise manner which known as Pairs of Values (PoV) which can be identified by Chi-square Test given by (Stanley,2005). [12] All LSB matching techniques were successfully attacked by best-known detector for LSB matching which is based on the centre of mass (COM) of the histogram characteristic function (HCF) discussed by (Ker, 2005). [13] T.Halder and S.Karforma modified the basic LSB-replacement technique with indexing; this method doesn t directly hide the message in LSB but apply indexing and match the bits.chi-square test can detect the presence of hidden bits but could not retrieve the bits as they are not directly embedded. [14] B. Experimented result using LSB steganography We have Experimented LSB steganography over some standard image and observed visual distortion within cover image and stego image by histogram analysis. Here is the result in fig , IJARCSSE All Rights Reserved Page 1151

3 Fig 2. shows image and its histogram. (a) image lena(512*512) with histogram, (b) histogram of stego image obtained using lsb1, (c)histogram of stego image using lsb2. (d) image baboon(512*512)with histogram, (e) histogram of stego image obtained using lsb1, (f) histogram of stego image using lsb2. III. PIXEL VALUE DIFFERENCE (PVD) STEGANOGRAPHY In section II-A we discussed disadvantages of LSB Steganography. Keeping in mind those factors of LSB methods Wu and Tsai [15] proposed a method known as Pixel value differencing, In PVD, a non-overlapping block of two pixels are chosen from the cover image. Difference between the pixels point out smooth and edge areas. More number of bits could be hidden in edge areas rather than smooth areas. Applying this methodology PVD method can hide more amount of data than LSB method and more susceptible against visual attack because rather than hiding secret bits throughout the image PVD finds out edge areas to hide large amount of data, that is more sharp edge contains more amount of secret data. A. Overview of PVD method The PVD method proposed by Wu and Tsai takes two consecutive pixels(p and Q) from cover image, the difference(d) between the pixels are calculated as D= P-Q (1) Value of D will be within the range from 0 to 255. In a smooth region value of D is less and in a sharp-edge region value of D is high, depending on the value of D number of bits could be chosen from the secret message, The range table Ri (i= 1, 2,..n) calculates the range of D, main purpose of the range table is to calculate the capacity of that pair. Each range has a lower and upper limit says Li and Ui, the width (Wi) of the range is a power of two and is calculated as Wi = (Ui-Li+1) (2) This restriction of width facilitates the embedding of binary data. If D falls in smooth area, less secret data will be hidden in the block. On the other hand, if D falls in sharp area, then the block has higher tolerance and thus more secret data can be embedded inside it. In this way hiding capacity (Ti) is calculated as Ti = [lg(wi)] (3) Here Ti is the number of bits that could be fetched from secret message. Now convert the secret bits in corresponding decimal number (Ti ). The new difference between the pixels P and Q is calculated as D = Li + Ti (4) The secret data could be hidden by adjusting P and Q as P and Q, the adjustment is as follows (5) B. Literature survey of PVD method Modified version of PVD method is Tri-way pixel value differencing (TPVD) method [16], In the TPVD method three direction pixel selections is made for increasing the payload by embedding secret bits in three different directions on cover image. PVD method only uses one direction for data embedding. Whereas TPVD use horizontal, vertical and diagonal edges of image for embeds secret data. Pre-processing phase of cover image into 2x2 pixel blocks is required in TPVD method. Another modified version of PVD is Adaptive pixel value differencing (APVD) method. APVD method is only applied for grey scale images. In grey scale digital image having pixel value ranges from 0 to 255. The stego image of exceeding the range of grey scale will be problem [17]. So the APVD method will solve these problems with modulus function and some conditions. So the pixel values of stego image will not exceed the grey scale range. A result and discussion of PVD method: Here in Figure 3 histogram shows changes in image using PVD method. Wu et al. (2005) proposed another method by combining PVD and LSB replacement scheme which shows better capacity than PVD method. [18]. Later this paper modified and analysed by Yang et al. (2007) using R-S diagram [19]. 2014, IJARCSSE All Rights Reserved Page 1152

4 Fig 3. (a)original image lena(512*512), (b) stego image by PVD method, (c)original image mandrill, (d) stego image mandrill IV. COMPARISON of LSB & PVD To compare these two methods we have implemented them using C language. After embedding same secret data within same cover image by different approaches(lsb-1, LSB-2 and PVD methods) we observed their statistical distortion between cover image and stego image by calculating PSNR(peak signal to noise ratio) using equation 6 and 7.Table 1 shows the result. Let x and y arrays of size NxM, respectively representing the Y-channel frame of reference (i.e. the original copy) and Y- channel frame of the encoded/impaired copy. The MSE between the two signals is defined in equation (6) and PSNR is obtained using equation (7). (6) (7) We have experimented LSB and PVD method over several images and tabulated their result in table 1. Lena Baboon Peppers Sailboat Barbara Tiffany f-18 f-14 Wu & Tsai s PVD method LSB-1 method LSB-2 method PSNR Payload PSNR Payload PSNR Payload A. Discussion From the result in table 1 implementation we have observed that LSB-1 & LSB-2 steganography are very flexible regarding capacity, depending on the size of secret message we may choose either LSB-1 or LSB-2, for a small size secret message LSB-1 is appropriate with a good range of PSNR, & for bigger size message we may select more bits from LSB side which will almost double the capacity with a degradation of PSNR. But the main disadvantage of these method is that hidden bits are easily detectable because research found out that, if an image is processed with simple LSB substitution the histogram of the image will be showed in a pair-wise manner. These pair-wise blocks are known as Pairs Of Values (PoV) which can be identified by Chi-square Test given by (Westfled et al., 1999), (Provos et al., 2002), (Stanley, 2005) test as bits are scattered throughout the whole image. 2014, IJARCSSE All Rights Reserved Page 1153

5 On the other hand PVD works in different approach; it hides less number of bits in smooth area and more number of bits in edge area. So hidden bits could not be detected easily by Chi-square detection process. But In terms of capacity it produces an average result which is always greater than Lsb-1 and less than Lsb-2. Another point about PVD is the algorithm is rigid against increased capacity, although lots of advanced PVD algorithm has overcome this serious issue but statistical Distortion is noticeable in those algorithms. V. CONCLUSIONS In this paper we have compared PVD and LSB based steganography with respect to capacity, histogram changes and PSNR. Advantage and disadvantage of both the algorithms are experimented and discussed. We have tested these algorithms over several documents (result of only few is shown here). These novel steganographic approaches could be applied on E-govornance documents to protect against unauthorised acess of those. Although we have used only E- Governance documents but other E-commerce related documents used in E-learning, E-banking could be used to protect them by steganography. REFERENCES [1] Sharp, T.: An implementation of key-based digital signal steganography. In: Proc. Information Hiding Workshop. Volume 2137 of Springer LNCS. (2001) [2] Niel F. Johnson, Zoran Duric, Sushil Jajodia (2000), Information Hiding, and Watermarking - Attacks & Countermeasures, Kluwer Academic Publishers. [3] Chin-Chen Chang, Min-Hui Lin, Yu-Chen Hu (2002), A Fast and Secure Image Hiding Scheme Based on LSB Substitution, International Journal of Pattern Recognition and Artificial Intelligence, vol. 16, no. 4, p [4] Chin-Chen Chang, Ju-Yuan Hsiao, Chi-Shiang Chan (2003), Finding Optimal Least Significant-Bit Substitution in Image Hiding By Dynamic Programming Strategy. Pattern Recognition, Vol. 36, and p [5] Thien, C. C., Lin, J. C. (2003), A Simple and High-Hiding Capacity Method for Hiding Digit-By-Digit Data in Images Based On Modulus Function. Pattern Recognition, vol. 36, p [6] Ran-Zan Wang, Chi-Fang Lin, Ja-Chen Lin, (2001), Image Hiding by Optimal LSB Substitution And Genetic Algorithm. Pattern Recognition, vol. 34, p [7] Chin-Chen Chang, Chi-Shiang Chan, Yi-Hsuan Fan (2006), Image Hiding Scheme with Modulus Function and Dynamic Programming Strategy on Partitioned Pixels. Pattern Recognition, vol. 39, no. 6, p [8] Ker, A. (May 23-25, 2004), Improved Detection of LSB Steganography in Grayscale Images. In Proc. 6 th International Workshop. Toronto (Canada), Springer LNCS, vol. 3200, p [9] A. Ker (June, 2005), Steganolysis of LSB Matching in Grey scale Images, IEEE Signal Process Letter, vol. 12, no.6, pp [10] Jarno Mielikainen (2006), LSB Matching Revisited, IEEE Signal Processing Letters, Vol. 13, no. 5, p [11] Xiaolong Li, Bin Yang, Daofang Cheng, Tieyong Zeng (2009), A Generalization of LSB Matching. IEEE Signal Processing Letters, vol. 16, no. 2, pp [12] Stanley, C.A. (2005), Pairs of Values and the Chi-squared Attack, in CiteSteer. 2005, pp [13] A. Ker (June, 2005), Steganolysis of LSB Matching in Grey scale Images, IEEE Signal Process Letter, vol. 12, no.6, pp [14] T.Halder, S.karforma : A LSB-Indexed steganographic approach to secure E-governance data.paper presented at the SECOND INTERNATIONAL CONFERENCE ON COMPUTING AND SYSTEMS 2013 (ICCS 2013)SEPTEMBER 21 22, 2013, THE UNIVERSITY OF BURDWAN. Burdwan, West Bengal, India. ISBN(13): ISBN(10): P-158. [15] Wu, D.C., Tsai, W.H., A steganographic method for images by pixel-value differencing. Pattern Recognition Letters 24, 9 10, [16] K.C. Chang, C.P Chang, P.S. Huang, and T.M. Tu, A Novel Image Steganographic Method Using Triway Pixel-Value Differencing, Journal of Multimedia, Vol. 3, No. 2, pp.37-44, June [17] J.K. Mandal, Debashis Das Steganography Using Adaptive Pixel Value Differencing (APVD) for Gray Images through Exclusion of Underflow/Overflow, Computer Science & Information Series, ISBN : , pp , [18] Wu, H.C., Wu, N.I., Tsai, C.S., Hwang, M.S., Image steganographic scheme based on pixel-value differencing and LSB replacement methods. IEE Proceedings- Vision Image and Signal Processing 152 (5), [19] Yang, C.H., Wang, S.J., Weng, C.Y., Analyses of pixel-value-differencing schemes with LSB replacement in steganography. In: The Third International Conference on Intelligent Information Hiding and Multimedia Signal Processing, pp [20] T.Halder, S.karforma A Novel E- Governance Data Hiding Approach Combining LSB-Steganography and Cryptography INDIAN SCIENCE CRUISER(ISSN ).VOL 28, NO 4, JULY , IJARCSSE All Rights Reserved Page 1154