Informaton Assurance and Securty Letters 1 (2010) 024-028 Enhanced Watermarkng Technque for Color Images usng Vsual Cryptography Enas F. Al rawashdeh 1, Rawan I.Zaghloul 2 1 Balqa Appled Unversty, MIS Department, Amman,Jordan Enas_fasal@yahoo.com 2 Balqa Appled Unversty, MIS Department, Amman,Jordan rawanzaghloul@yahoo.com Abstract: In s paper we consder a watermarkng technque based on vsual cryptography at offers enhancng e securty ssues n prevous works. The proposed meod s appled over color mages, where e RGB mage s converted to e HSV color scheme. The embeddng process depends on e saturaton plane (S) on whch we apply logcal X-OR, most sgnfcant, and party operatons; to generate e actual plane n whch we want to embed our watermark -upon e rules of vsual cryptography (as nvented by Noar and Shamr, 1995). Fnally e generated master share s regstered n e rd party. In e extracton process, applyng e same meod over e watermarked mage wll generate e master share agan; combne t w at of e rd party to generate e embedded watermark. Dfferent attack scenaros are taken nto consderaton such as flterng, addtve nose, compresson, rotaton, flppng, croppng, and dstorton. Keywords: Watermarkng, Secret Sharng, Vsual Cryptography, Party bt. 1. Introducton The rapd grow of multmeda content n dgtal form made e dgtal mages easy to steal, modfy, or used llegally by unauorzed people, s creates a pressng need for buldng more secure meods for legal dstrbuton of e dgtal content. So, dgtal watermarkng has been proposed to protect e ntegrty of multmeda. An mage watermark meod s now drawng e need for good schemes to protect copyrghts for dgtal data. But e chosen meod must be hard to detect and remove. There are many artcles reveal and proposes some watermark meods [2, 4, 9, 10], however ere are several watermarkng meods at are not robust or easy to remove [5]. In s paper we devse a watermarkng meod based on e vsual cryptography at ntroduced by Naor and Shamr [1]. The paper s organzed as follows: Vsual Cryptography s dscussed n Secton 2. In secton 3 e used measures are presented. Secton 4 dscusses e proposed algorm. The expermental results are provded n Secton 5. Fnally, e conclusons are presented n Secton 6. 2. Vsual Cryptography Cryptography at can be decoded drectly by e human vsual system wout any specal calculaton for decrypton s called Vsual Cryptography. It was proposed for e frst tme by Naor and Shamr. In vsual cryptography a secret message can be vewed as nong more an a collecton of black and whte pxels as depcted on fgure. 1. Each pxel n e orgnal mage s represented by at least one subpxel n each of e n generated shares. Fgure 1. The basc scheme of vsual cryptography Each share s comprsed of collectons of m black and whte subpxels where each collecton represents a partcular orgnal pxel [1, 2, 3]. R.J. Hwang proposed a watermark meod based on vsual cryptography, accordng to hs meod, e owner should select an h n black/whte mage as e watermark. In e embeddng processes, e owner should randomly select a number as hs secret key S, to embed e watermark nto e mage K at s a k l (256 gray-leveled) mage. The owner embeds e watermark pattern W nto e mage K by generatng e secret key S, and e verfcaton nformaton V. As llustrated n table 1, f e pxel n e watermark W s zero and e left most sgnfcant bt (LMB) of e R pxel n mage M s 1 en e verfcaton nformaton equals (0,1), and so on. In e extracton process, e notaral organzaton retreves e verfcaton nformaton V and e watermark pattern W, whch e owner has regstered, and verfes e ownershp of e watermarked mage K, for more detals see [2]. Table 1. Verfcaton Rules On 2009, Surekha proposed a vsual cryptography algorm n mage watermarkng at offers better securty an Hwang s meod, so at, attackers wll not be able to detect ownershp nformaton [3]. Our proposed meod offers an enhancement for Surekha s algorm. Receved June 21, 2010 2150-7996 $ 03.50 Dynamc Publshers, Inc.
025 Rawashdeh and Zaghloul 3. Measures To prove e effect of our meod eoretcally by measures, e followng metrcs are used to measure dstorton and smlarty between e orgnal watermark and e extracted watermark: 1. Root Mean Square Error (RMSE), as defned n equaton 1, (1) Where m n s e sze of e watermark, Wj s e watermark pxel n e row and j column, and W'j s e W' n e row and j column. 2. Peak Sgnal to Nose Rato (PSNR), as defned n equaton 2.[7,6] (2) Where e MSE s e Mean Square Error whch defned by equaton 3: (3) Where m n s e sze of e watermark, Wj s e watermark pxel n e row and j column, and W'j s e W' n e row and j column. [7,8] Fgure 2. Embeddng Model Fgure.3 shows e extracton Model, n whch we start by readng e watermarked mage I' and convert t to e HSV scheme, en we extract e saturaton from e HSV mage. Then we start e extractng process as dscussed n secton 4.3. Fnally e secret mage s extracted. 3. Correlaton (Corr), as defned n (3), (4) Where and are e mean for e orgnal watermark and e extracted watermark respectvely. [7,8] 4. Proposed Algorm Ths secton reveals e proposed meod for watermark embeddng and detecton. 4.1 General Model The general dagrams of our meod are descrbed n Fgure.2 and Fgure.3. Where Fgure.2 shows e watermark embeddng Model, n whch we convert e host mage from RGB color scheme to e HSV color scheme. Then we extract e S (saturaton) from HSV mage. After at we start e embeddng process as llustrated n secton 4.2, at e end we generate e Master share M, at s regstered n a trusted rd party. Fgure 3. Extracton Model 4.2 Watermark Embeddng As shown n Fgure.5 e embeddng process s as follows: Read e RGB mage I of sze rxc Convert e mage to e HSV scheme Extract e saturaton plane (S) from e HSV mage Read e black-whte watermark mage (W) of sze wxh Select a secret key K as a seed to generate wxh random numbers over e nterval [1 to rxc]. Let R be e random number.
Enhanced Watermarkng Technque for Color Images usng Vsual Cryptography 026 Create a bnary matrx MSB of sze wxh, t contans e most sgnfcant bts of R pxel of e host mage. Create a bnary matrx P of sze wxh at represents e party for e frst ree bts of R pxel of e host mage. Compute Y such at t fnds e logcal xor between P and MSB as n e followng equaton, Y = P MSB Create a bnary matrx Z of sze wxh such at e entres n e array are e most sgnfcant bts of e R random number. Create a bnary matrx Y of sze wxh such at F = XOR (Y, Z ) Generate e Master Share M by assgnng a par of bts for each element n e bnary matrx F accordng to e encrypton rules of VC. As dscussed before n table 1. M s regstered to e trusted rd party. Fgure 5. Watermark embeddng Fgure 6. Watermark extracton
027 Rawashdeh and Zaghloul 4.3 Watermark Extracton As shown n Fgure.6 e extracton process s as follows: Read e watermarked mage I of sze rxc Convert I to HSV scheme Extract e saturaton plane (S) from I Read e black-whte watermark mage (W) of sze wxh Use e secret key K as a seed to generate wxh random numbers over e nterval [1 to rxc]. Let R be e random number. Create a bnary matrx MSB of sze wxh, t contans e most sgnfcant bts of R pxel of e host mage. Create a bnary matrx P of sze wxh at represents e party for e frst ree bts of R pxel of e host mage. Create a bnary matrx Z of sze wxh such at e entres n e array are e most sgnfcant bts of e R random number. Create a bnary matrx F of sze wxh such at F = XOR (Y, Z ) Create e Verfcaton Share such at, f e element n e bnary matrx F s 0 en V = (0, 1), else V = (1, 0) The secret mage can be extracted by performng logcal OR between e master share M and e Verfcaton share V. 5. Expermental Results The robustness of our approach s tested by applyng several attacks over e Peppers mage of sze (512x512) and e bnary watermark mage of sze (144 x 148). Fgures 7,8 show e orgnal peppers mage and e bnary watermark respectvely. Fgure 7. Peppers orgnal mage Fgure 8. Bnary watermark mage The robustness of e meod s confrmed by expermental results at dsplay e mmunty of e embedded watermark over dfferent types of attacks, such as compresson, flterng, croppng, rotaton, addtve nose, and dstorton. Table.2 shows e detaled results. Attack Watermarked Image RMSE PSNR Corr Extracted Watermark Blurred (gaussan, w flter sze=5) 42.4664 31.8503 0.9368 Flterng Lnear moton of a camera by len pxels=31, angle eta degree=11 49.5357 31.1816 0.9142 59.902 30.3564 0.8756 Addtve Nose 96.7933 28.2724 0.6867 Compresson 77.0675 29.2621 0.7981 Table 2. Expermental results
Enhanced Watermarkng Technque for Color Images usng Vsual Cryptography 028 Rotaton 95.362 28.337 0.698 Flppng 90.6041 28.5593 0.7225 Croppng 61.3170 30.2550 0.8708 Dstorton 92.2124 28.4829 0.7148 Table 2. Expermental results (contnued) 6. Conclusons The need of effcent meod at makes t hard to detect e marked mage, t wll become more prevalent. Ths paper, proposed a good watermark meod at s buld up on e vsual cryptography watermarkng technque for secure mage auentcaton for colored mages. The features of our meod are lsted below: Ths meod s appled over colored mages; especally we have used e saturaton plane from e HSV model for secure auentcaton. To ncrease e securty of e proposed model, e logcal X-OR and partes are used. Accordng to our model we acheve good results aganst several knds of attacks such as flterng, addtve nose, compresson, rotaton, flppng, croppng and dstorton. References [1] M. Noar and A. Shamr, Vsual Cryptography, Advances n Cryptography Eurocrypt 94, Lecture Notes n Computer Scence, Sprnger-Verlag, Berln, Vol. 950, pp. 1-12, 1995. [2] R. Hwang, A dgtal Image Copyrght Protecton Scheme Based on Vsual Cryptography, Tambang Journal of scence and Engneerng, vol.3, No.2, pp. 97-106, 2000. [3] B.Surekha,GN.Swamy, K.Rao and A.Kumar, A Watermarkng Technque based on Vsual Cryptography, Journal of Informaton Assurance and Securty No.4.pp. 470-473, 2009. [4] M.Prasad and S.Kolwad, A Comprehensve Survey of Contemporary Researches n Watermarkng for Copyrght Protecton of Dgtal Images, IJCSNS Internatonal Journal of Computer Scence and Network Securty, VOL.9 No.4, Aprl 2009 [5] Inoue H., Myazak A., Yamamoto A. and Katsura T., A Dgtal Watermark Technque Based on e Wavelet Transform and Its Robustness on Image Compresson and Transformaton, IEICE Transactons on Fundamentals of Electroncs, Communcatons and Computer Scences, Vol. E82- A, No. 1, pp.2-10(1999). [6] B.Verma, S.Jan, D. Agarwal, and A.Phadkar, A New Color Image Watermarkng Scheme ; INFOCOMP, vol. 5, No.2, pp. 37-42 (2006). [7] Rawan I. Zaghloul, Enas F. Al-Rawashdeh, "HSV Image Watermarkng Scheme Based on Vsual Cryptography", Proceedngs of World Academy of Scence, Engneerng and Technology, Vol. 34, October 2008. [8] MaWorks, Image Processng Toolbox User's Gude 4.2. [9] P.S.Revenkar, Ansa Anjum, and W.Z.Gandhare, Secure Irs Auentcaton Usng Vsual Cryptography, (IJCSIS) Internatonal Journal of Computer Scence and Informaton Securty, Vol. 7, No.3, 2010 [10] Er.Suprya Knger, Effcent Vsual Cryptography, Journal of emergng technology n web ntellgent, Vol. 2, No. 2, 2010 Auor Bographes E Rawashdeh receved a B.S. degree n Computer nformaton system from appled scence Unversty, Jordan n 2004 and e M.S. degrees n Computer scence from Jordanan Unversty, Jordan n 2006, where she s currently workng n MIS Department of Al-Balqa a Appled Unversty, Jordan. Her research nterests nclude mage processng. R Zaghloul receved a B.S. degree n Computer Scence from Al- Balqa a Appled Unversty, Jordan n 2003 and e M.S. degrees n Computer scence from Jordanan Unversty, Jordan n 2005, where she s currently workng n MIS Department of Al-Balqa a Appled Unversty, Jordan. Her research nterests nclude mage processng.