• Save
Is3314841490
Upcoming SlideShare
Loading in...5
×
 

Is3314841490

on

  • 230 views

International Journal of Engineering Research and Applications (IJERA) is an open access online peer reviewed international journal that publishes research and review articles in the fields of ...

International Journal of Engineering Research and Applications (IJERA) is an open access online peer reviewed international journal that publishes research and review articles in the fields of Computer Science, Neural Networks, Electrical Engineering, Software Engineering, Information Technology, Mechanical Engineering, Chemical Engineering, Plastic Engineering, Food Technology, Textile Engineering, Nano Technology & science, Power Electronics, Electronics & Communication Engineering, Computational mathematics, Image processing, Civil Engineering, Structural Engineering, Environmental Engineering, VLSI Testing & Low Power VLSI Design etc.

Statistics

Views

Total Views
230
Views on SlideShare
230
Embed Views
0

Actions

Likes
0
Downloads
0
Comments
0

0 Embeds 0

No embeds

Accessibility

Categories

Upload Details

Uploaded via as Adobe PDF

Usage Rights

© All Rights Reserved

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Processing…
Post Comment
Edit your comment

Is3314841490 Is3314841490 Document Transcript

  • M.Lakshmi Prasanna, Mr. Sk.Mahaboob Basha / International Journal of EngineeringResearch and Applications (IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.1484-14901484 | P a g eExtended Adaptive Pixel Pair MatchingM.Lakshmi Prasanna1,Mr. Sk.Mahaboob Basha2Department of Electronics and Communication Engineering, SreeVidyanikethan Engineering College,TIRUPATIAbstract:Steganography is a method of hidingsecret messages into a cover-media such that anunintended observer will not be aware of theexistence of the hidden messages. The carrier ofsteganography can be various kinds of digitalmedia such as image, audio, and video, etc. Digitalimages are widely transmitted over the Internet;therefore, they often serve as a carrier for covertcommunication. A good data-hiding methodshould be capable of evading visual and statisticaldetection while providing an adjustable payload.In [1], authors proposed Adaptive Pixel PairMatching (APPM). The basic idea of Pixel PairMatching (PPM) is to use the values of pixel pairas a reference coordinate, and search a coordinatein the neighborhood set of this pixel pairaccording to a given message digit. The pixel pairis then replaced by the searched coordinate toconceal the digit. APPM allows users to selectdigits in any notational system for dataembedding, and thus achieves a better imagequality. APPM not only resolves the low-payloadproblem in EMD, but also offers smaller MSEcompared with OPAP and DE. Moreover, becauseAPPM produces no artifacts in stego images andthe steganalysis results are similar to those of thecover images, it offers a secure communicationunder adjustable embedding capacity. This paperproposes an extension to APPM, for highdefinition color images. EAPPM provides highdata capacity without compromising onperformance and security of APPM.Keywords: Steganography, Pixel Pair Matching(PPM), Diamond Encoding (DE), Least SignificantBit (LSB)I. IntroductionThe purpose of steganography is to sendsecret messages after embedding them into publicdigital multimedia. It is desired to embed as manymessages as possible per change of the cover-object.In general, for given messages and covers, thesteganography that introduces fewer embeddingchanges will be less detectable, i.e., more secure.Two fields of research have been proposed toenhance the communication security: cryptographyand information hiding. Although they are bothapplied to the protection of secret message, the majordifference is the appearance of the transmitted data.The cryptography methods, such as DES and RSA,referred almost exclusively to encryption which is theprocess of converting ordinary information(plaintext) into unintelligible gibberish (cipher-text).After data encryption, the secret data appears to be atotal chaos of seemingly meaningless bits. However,the existence of the transmitted secret message can bedetected. Because it does not conceal the fact thatthere is an important message, the encrypted messagecould motivate an unauthorized user to decrypt ordestroy it.Many approaches of information hiding havebeen proposed for different applications, such ascopyright protection, secret transmission, tamperingdetection, and image authentication. The most well-known data hiding scheme is the least significant bits(LSBs) substitution method. This method embedsfixed-length secret bits into the least significant bitsof pixels by directly replacing the LSBs of coverimage with the secret message bits.Although this method is simple, it generallyeffects noticeable distortion when the number ofembedded bits for each pixel exceeds three. Severalmethods have been proposed to reduce the distortioninduced by LSBs substitution. Another way ofimproving LSBs scheme is to reduce the amount ofalterations necessary to be introduced into the coverimage for data hiding when the number of secret bitsis significantly less than that of available coverpixels. The method proposed by Tseng et al. [15] canconceal as many as log2(mn+1) bits of data in abinary image block sized m X n by changing, at most,two bits in the block. Matrix encoding, on the otherhand, uses less than one change of the leastsignificant bit in average to embed w bits into2w-1cover pixels.Diamond Encoding(DE)method is the extensionof the exploiting modification direction (EMD)embedding scheme [2]. The main idea of the EMDembedding scheme is that each (2n +1)-ary notationalsecret digit is carried by n cover pixels, and only onepixel value increases or decreases by 1 at most. Foreach block of n cover pixels, there are 2n possiblestates of only one pixel value plus 1 or minus 1. The2n states of alteration plus the case in which no pixelis modified form (2n + 1) different cases. Therefore,the (2n + 1)-ary notational secret digit is embeddedinto the cover pixels by changing the state. Before thedata embedding procedure, the pre-process canconvert the secret data into sequences of digits with(2n + 1)-ary notational representation.
  • M.Lakshmi Prasanna, Mr. Sk.Mahaboob Basha / International Journal of EngineeringResearch and Applications (IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.1484-14901485 | P a g eThe basic idea of the PPM-based data-hidingmethod is to use pixel pair (x, y) as the coordinate,and searching a coordinate (x’, y’) within apredefined neighborhood set Ø(x, y) such that f(x’,y’) = SB , where f is the extraction function and SB isthe message digit in a -ary notational system to beconcealed. Data embedding is done by replacing (x,y) with (x’, y’).The rest of the paper is organized as follows: Section2 gives the Literature review on the present work,Section 3 explains the proposed EAPPM with Section4 presenting the Performance and Security analysis ofproposed methods. Section 5 concludes the paper.II. Literature ReviewOPAP effectively reduces the imagedistortion compared with the traditional LSB method.DE enhances the payload of EMD by embeddingdigits in a B-ary notational system. These twomethods offer a high payload while preserving anacceptable stego image quality. In this section, OPAPand DE will be briefly reviewed. The OPAP methodproposed by Chan et al. in 2004 greatly improved theimage distortion problem resulting from LSBreplacement.2.1 LSB substitutionLet C be the original 8-bit grayscale cover-image ofMc×Nc pixels represented asM be the n-bit secret message represented asSuppose that the n-bit secret message M is to beembedded into the k-rightmost LSBs of the cover-image C. Firstly, the secret message M is rearrangedto form a conceptually k-bit virtual image M’represented aswhere n’ <Mc×Nc. The mapping between the n-bitsecret message M ={mi} and the embedded messageM’={mi’}can be defined as follows:Secondly, a subset of n’ pixels {xl1, xl2,..,xln’}is chosenfrom the cover-image C in a predefined sequence.The embedding process is completed by replacing thek LSBs of xli by mi’. Mathematically, the pixel valuexli of the chosen pixel for storing the k-bit messagemi’ is modified to form the stego-pixel xli’ as follows:In the extraction process, given the stego-image S,the embedded messages can be readily extractedwithout referring to the original cover-image[3].Using the same sequence as in the embeddingprocess, the set of pixels{xl1, xl2,..,xln’} storing thesecret message bits are selected from the stego-image. The k LSBs of the selected pixels areextracted and lined up to reconstruct the secretmessage bits. Mathematically, the embedded messagebits mi’ can be recovered by2.2 Diamond Encoding (DE)The EMD scheme embeds(2n + 1)-ary digit into ncover pixels, but the diamond encoding scheme canconceal (2k2 + 2k + 1)-ary digit into a cover pixelpair where k is the embedding parameter. The detailof this scheme is described as follows.Assume that a, b, p, and q are pixel values, and k is apositive integer. The neighborhood set Sk (p, q)represents the set that contains all the vectors (a, b)with the distance to vector (p, q) smaller than k, andSk(p, q) is defined as the following form:Let the absolute value |Sk| denote the number ofelements of the set Sk, and each member in Skiscalled neighboring vector of (p, q). We calculate thevalue of |Sk| to obtain the embedding base andembedded base with a parameter k. Diamondencoding method uses a diamond function f tocompute the diamond characteristic value (DCV) inembedding and extraction procedures. The DCV oftwo pixel values p and q can be defined as follows:where l is the absolute value of Sk. The DCV havetwo important properties: the DCV of the vector (p,q) is the member of Sk belongs to {0, 1, 2, . . ,l-1}andany two DCVs of vectors in Sk(p, q) are distinct.Assume that Ek represents the embedded digit and Ekbelongs to{0, 1, 2, . . . ,l − 1}. For secret dataembedding, we replace the DCV of the vector (p, q)with the embedded secret digit. Therefore, themodulus distance between f (p, q) and Sk is dk= f (p,q)−Ek mod l. For each k, we can design a distancepattern Dk to search which neighboring pixel ownsthe modulus distance dk. Then, the vector (p, q) isreplaced with the neighboring vector (p’, q’) by dk.The vector (p’, q’) is the member of Sk(p, q) and theDCV of (p’, q’) equals to the embedded secret digitEk. The vector (p’, q’) can extract the correct secretdigit byThe diamond encoding scheme promises that thedistortion of vector (p, q) is no more than k afterembedding a secret digit Ek. Therefore, this minimaldistortion scheme can be employed to embed largeamount of data.2.3 Adaptive Pixel Pair MatchingThe basic idea of the PPM-based data-hiding methodis to use pixel pair (x, y) as the coordinate, and
  • M.Lakshmi Prasanna, Mr. Sk.Mahaboob Basha / International Journal of EngineeringResearch and Applications (IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.1484-14901486 | P a g esearching a coordinate (x’,y’) within a predefinedneighborhood set Ø(x, y) such that f(x’, y’) = SB,where f is the extraction function and SB is themessage digit in a B-ary notational system to beconcealed. Data embedding is done by replacing (x,y) with (x’, y’) [1].Suppose the cover image is of size M×M, Sis themessage bits to be concealed and the size of S is |S|.First we calculate the minimum B such that all themessage bits can be embedded. Then, message digitsare sequentially concealed into pairs of pixels. Firstminimum B satisfying |M×M / 2| ≥|SB|, and convert Sinto a list of digits with a B-ary notational system SB.The discrete optimization problem is solved to find cBand ØB(x,y). In the region defined by ØB(x, y), recordthe coordinate (x’, y’) such that f(x’, y’) = i, 0 ≤ i≤ B-1.Construct a nonrepeating random embeddingsequence Q using a key Kr. To embed a message digitsB, two pixels (x, y) in the cover image are selectedaccording to the embedding sequence Q, andcalculate the modulus distancebetween sB and f(x, y), then replace (x, y) with (x +x’, y+y’).III. Proposed MethodologyAs APPM is proved to offer better securityagainst detection and lower distortion, we takeforward APPM for colored images. We propose toexplore a better mechanism and provide bettersecurity and lower distortion for embedding data incolored images. Also, performance in terms ofpayload can be improved. In colored images, whichconsists three different colored layers, in each layerone can embed message bits so that the capacity ofthe Adaptive Pixel Pair Matching can be improvedwithout any distortion in the original colored image.The R, G, B layers are separated and the each isconsidered as a gray image, referred as ChannelImage. For a PPM-based method, suppose a digit sBis to be concealed. The range of sB is between 0 andB-1, and a coordinate (x’, y’) in Ø(x, y) has to befound such that f(x’,y’) = sB. Therefore, the range of(x ,y) must be integers between 0 and B-1, and eachinteger must occur at least once. In addition, toreduce the distortion, the number of coordinates inØ(x, y) should be as small as possible. The best PPMmethod shall satisfy the following threerequirements:1) There are exactly B coordinates in Ø(x, y).2) The values of extraction function in thesecoordinates are mutually exclusive.3) The design of Ø(x, y)and f(x, y) should be capableof embedding digits in any notational system so thatthe best can be selected to achieve lower embeddingdistortion.The definitions of Ø(x, y) and f(x, y) significantlyaffect the stego image quality. The designs of Ø(x,y)and f(x, y) have to fulfill the requirements: allvalues of in have to be mutually exclusive and thesummation of the squared distances between allcoordinates in Ø(x, y) and f(x, y) has to be thesmallest. This is because, during embedding, (x, y) isreplaced by one of the coordinates in Ø(x, y).Suppose there are B coordinates in Ø(x, y), i.e., digitsin a B-ary notational system are to be concealed, andthe probability of replacing (x, y) by one of thecoordinates in Ø(x, y) is equivalent. The averagedMSE can be obtained by averaging the summation ofthe squared distance between and other coordinates inØ(x, y). Thus, given a Ø(x, y), the expected MSEafter embedding can be calculated byThe solution of Ø(x, y) and f(x, y) is indeed a discreteoptimization problemfor 0 ≤ i, j ≤ B-1.Embedding Procedure:Consider the cover image is of size M×M, then eachof R, G, B channels will be of size M×M. S is themessage bits to be concealed for each channel imageand the size of S is |S|. First we calculate theminimum B such that all the message bits can beembedded. Then, message digits are sequentiallyconcealed into pairs of pixels.1. First minimum B satisfying |M×M / 2| ≥|SB|,and convert S into a list of digits with a B-ary notational system SB.2. The discrete optimization problem is solvedto find cB and ØB(x, y).3. In the region defined by ØB(x, y), record thecoordinate (x’, y’) such that f(x’, y’) = i, 0 ≤i ≤ B-1.4. Construct a nonrepeating randomembedding sequence Q using a key Kr.5. To embed a message digit sB, two pixels (x,y) in the cover image are selected accordingto the embedding sequence Q, and calculatethe modulus distancebetween sB and f(x, y), then replace (x, y) with (x +x’, y + y’).6. Repeat step 5, until all the message bits areembedded.To avoid any distortion because of replacing pixelsright under each other in different layers, the regionsØ(x, y) for each layer are taken as distinct subsets.
  • M.Lakshmi Prasanna, Mr. Sk.Mahaboob Basha / International Journal of EngineeringResearch and Applications (IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.1484-14901487 | P a g eSay the regions for red channel image be ØBr, greenchannel image be ØBgand blue channel image be ØBbØBr, ØBgand ØBbare selected such thatØBr⊄ ØBg⊄ ØBbAs the three are taken as distinct sets, the clash ofgetting pixel pair at the same positions is avoided.The three layers are merged again to retain originalimage.Figure 1. Embedding respective watermarks fordifferent Channel ImagesExtraction Procedure:To extract the embedded message digits,pixel pairs are scanned in the same order as in theembedding procedure. The embedded message digitsare the values of extraction function of the scannedpixel pairs.1. The watermarked image is split intorespective R, G, B layers and each isconsidered as a Channel Image.2. Construct the embedding sequence Qusing a key Kr.3. Select two pixels (x’, y’) according tothe embedding sequence Q.4. Calculate f(x’, y’), the result is theembedded digit.5. Repeat Steps 2 and 3 until all themessage digits are extracted.6. Finally, the message bits can beobtained by converting the extractedmessage digits into a binary bit stream.Theoretical Analysis:Image distortion occurs when data areembedded because pixel values are modified. MSE isused to measure the image quality.where M×M denotes the image size, pi,j and p’i,jdenote the pixel values of the original image and thestego image, respectively.MSE represents the meansquare error between the cover image and stegoimage. A smaller MSE indicates that the stego imagehas better image quality.When data are embedded using LSBs of each pixel,each bit valued 0 or 1 has equal probability. Thesquared error caused by embedding a bit in the ithLSB is (1/2)(2i-1)2; therefore, the averaged MSE ofembedding LSBs is given byLet the original pixel value be v and the stego pixelvalue be v’. The probability of |v – v’| = 0 or |v – v’| =2r-1is 1/2r; the probability of |v – v’| to be within therange [1, 2r-1-1]is 1/2r. Therefore, the averaged MSEcaused by embedding r bits isFor the DE method, assume that the probability ofselecting a coordinate (xi, yi) in the diamond shapeØ(x, y) to replace a pixel pair (x, y) is the same.Therefore, the averaged MSE caused by embeddingdigits in a B-ary notational system isWhere k is the embedding parameters of DE.
  • M.Lakshmi Prasanna, Mr. Sk.Mahaboob Basha / International Journal of EngineeringResearch and Applications (IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.1484-14901488 | P a g eFor embedding digits in a B-ary notational systemusing APPM, assume that the probability of replacing(x, y) with each (x’, y’) in Ø(x, y) is identical. Withthe knowledge of Ø(x, y) , the averaged MSE can beobtained byFor EAPPM,Consider common MSE M, for all the above saidtechniques.The maximum supported payload for LSB can becalculated fromFor DE, the maximum supported payload can becalculated asFor the PPM-based embedding method, a payloadwith r bpp is equivalent to embedding2r bits forevery two pixels, which is equivalent to concealingdigits in a 22r-ary notational system. Hence forAPPM,rappm = 2 × rdeBecause of embedding data in three channels, the rbpp is equivalent to embedding 3 × 2r bits for everythree pairs of pixels. Hence for EAPPM,r = 3 × rappmIV. Performance and Security AnalysisTo evaluate the performance of the proposedscheme, a high definition image is taken. Thesimulation is run using MATLAB. First, LSB, DE,APPM and EAPPM are evaluated for Mean SquareError (MSE) with different payloads. Table 1presents the obtained MSEs. It is observed thatAPPM outperforms APPM, DE and LSB.Table2 presents the maximum payload supported bythe four embedding methods at an MSE of 0.092.EAPPM is able to support 300% more than APPM.As the data is embedded in all the three layers, forEAPPM, the payload support will be more than 3times that of APPM.Payload(bpp)LSB DE APPM EAPPM0.025431 0.1686760.1436490.0949270.076294 0.1686760.1436490.0949270.0914970.114441 0.1643260.1437270.0917710.0956040.15767 0.1687660.1439010.0948070.0914970.203451 0.1689730.1438260.0949710.0914970.257492 0.16924 0.1440030.0953740.0914970.4673 0.0918740.610352 0.0918120.772476 0.091648Table1. MSE ComparisonMax Payload supportedwith MSE at 0.092LSB 0.025431DE 0.025431APPM 0.114441EAPPM 0.772476Table2. Max Payload support ComparisonFigure2. Average vertical and horizontal differencehistograms of LSB
  • M.Lakshmi Prasanna, Mr. Sk.Mahaboob Basha / International Journal of EngineeringResearch and Applications (IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.1484-14901489 | P a g eFigure3. Average vertical and horizontal differencehistograms of DEFigure4. Average vertical and horizontal differencehistograms of APPMFigure5. Average vertical and horizontal differencehistograms of EAPPMFigures 1, 2, 3 and 4 show the average vertical andhorizontal difference histograms of LSB, DE, APPMand EAPPM. It is observed that the difference isabsolutely less to be noticed. This ensures highsecurity of the data against different steganalysis.V. ConclusionThis paper proposed an efficient dataembedding algorithm EAPPM, based on APPM,which can successfully embed data into three layersof an RGB image. EAPPM is able to embed threetimes data more than APPM, without anycompromise on MSE and security. The advantages ofAPPM, like the freedom for user to use anynotational system and better image quality are carriedto EAPPM. EAPPM not only achieves higherpayloads, but also offers smaller MSE compared toDE and APPM. EAPPM offers high security alongwith the adjustable data embedding capacity.References:[1] Wien Hong and Tung-Shou Chen, “A NovelData Embedding Method Using AdaptivePixel Pair Matching”, IEEETRANSACTIONS ON INFORMATIONFORENSICS AND SECURITY, VOL. 7,NO. 1, FEBRUARY 2012.[2] Ruey-Ming Chao,Hsien-ChuWu,Chih-ChiangLee, and Yen-Ping Chu, “A Novel ImageData Hiding Scheme with DiamondEncoding”, EURASIP Journal on InformationSecurity Volume 2009, Article ID 658047.[3] Chi-Kwong Chan, L.M. Cheng, “Hiding datain images by simple LSB substitution”,Pattern Recognition Society. Published byElsevier Ltd.[4] A. Cheddad, J. Condell, K. Curran, and P.McKevitt, “Digital image steganography:Survey and analysis of current methods,”Signal Process., vol. 90, pp. 727–752, 2010.[5] T. Filler, J. Judas, and J. Fridrich,“Minimizing embedding impact insteganography using trellis-codedquantization,” in Proc. SPIE, Media Forensicsand Security, 2010, vol. 7541, DOI:10.1117/12.838002.[6] C. K. Chan and L. M. Cheng, “Hiding data inimages by simple LSB substitution,” PatternRecognit., vol. 37, no. 3, pp. 469–474, 2004.[7] J. Wang, Y. Sun, H. Xu, K. Chen, H. J. Kim,and S. H. Joo, “An improved section-wiseexploiting modification direction method,”Signal Process., vol. 90, no. 11, pp. 2954–2964, 2010.[8] J. Fridrich and T. Filler, “Practical methodsfor minimizing embedding impact insteganography,” in Proc. SPIE, Security,Steganography., Watermarking ofMultimedia, 2007, vol. 6050, pp. 2–3.[9] J. Fridrich and D. Soukal, “Matrix embeddingfor large payloads,”IEEE Trans. Inf.Forensics Security, vol. 1, no. 3, pp. 390–394,Sep.2006.[10] Chi-Kwong Chan, L.M. Cheng, Improvedhiding data in images by optimal moderatelysigni7cant-bit replacement, IEEElectron. Lett.37 (16) (2001) 1017–1018
  • M.Lakshmi Prasanna, Mr. Sk.Mahaboob Basha / International Journal of EngineeringResearch and Applications (IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.1484-14901490 | P a g e[11] Ran-Zan Wang, Chi-Fang Lin, Ja-Chen Lin,Image hiding by optimal LSB substitutionand genetic algorithm, Pattern Recognition 34(3) (2001) 671–683.[12] G. Cancelli, G. Doërr, I. Cox, and M. Barni.“Detection of steganography based on theamplitude of histogram local extrema”. IEEEInternational Conference on ImageProcessing,ICIP,San Diego, California,October 12–15, 2008.[13] F. Hartung and M. Kutter, “Multimediawatermarking techniques,”Proceedings of theIEEE, vol. 87, no. 7, pp. 1079–1107,1999.[14] X. Zhang and S.Wang, “Vulnerability ofpixel-value differencing steganography tohistogram analysis and modification forenhanced security,” Pattern RecognitionLetters, vol. 25, no. 3,pp. 331–339, 2004.[15] Y.-C. Tseng, Y.-Y.Chen, and H.-K. Pan, “Asecure data hiding scheme for binaryimages,” IEEE Transactions onCommunications, vol. 50, no. 8, pp. 1227–1231, 2002.AUTHORS PROFILEM.LakshmiPrasanna was born inAndhra Pradesh, India, in 1988. Shereceived the Bachelor degree,B.Tech (ECE) from the Universityof JNTU, Ananthapoor, in 2009. Sheis currently pursuing Master degree,M.Tech (DECS) in SreeVidyanikethan Engineering College, Tirupathi.Mr. Sk.Mahaboob Basha receivedhis M.Tech degree. He is currentlyworking as Assistant Professor,Department of ECE inSreeVidyanikethan EngineeringCollege.