JOURNAL OF MULTIMEDIA, VOL. 8, NO. 2, APRIL 2013 161 Robust Blnd Vdeo Watermark Algorthm n Transform Doman Combnng wth 3D Vdeo Correlaton DING Ha-yang 1,3 1. Informaton Securty Center, Bejng Unversty of Posts and Telecommuncatons, Bejng 100876, Chna; 3. College of Informaton & Mechancal Engneerng, Bejng Insttute of Graphc Communcaton, Bejng 102600, Chna Emal: o_dhy@163.com ZHOU Ya-jan 1,2, YANG Y-xan 1,2, ZHANG Ru 1,2 1. Informaton Securty Center, Bejng Unversty of Posts and Telecommuncatons, Bejng 100876, Chna; 2. Key Laboratory of Trustworthy Dstrbuted Computng and Servce (BUPT), Mnstry of Educaton, Bejng Unversty of Posts and Telecommuncatons, Bejng 100876, Chna; Emal: yajan@bupt.edu.cn, yxyang@bupt.edu.cn, zhangru@bupt.edu.cn Abstract Accordng to the features of DCT-based and DFT-based vdeo watermark algorthms aganst attacks and consderng 3D vdeo structural characterstc, a new robust blnd vdeo watermark algorthm n transform doman combnng wth 3D vdeo correlaton s proposed. DCT-based watermark and DFT-based watermark are embedded n left and rght mages of a vdeo frame usng opposte embeddng rules, and the watermark can be extracted by ntegratng 3D vdeo correlaton, and more robust watermark s extracted from the DCT-based watermark and the DFT-based watermark by comparng ther performance. The algorthm can sgnfcantly mprove the robustness of 3D vdeo watermark aganst resoluton attack and codng attack. The expermental results show that the proposed algorthm has excellent performance aganst resoluton attack and codng attack, and s obvously better than DCT-based and DFT-based algorthms. Index Terms Vdeo Watermark; DCT; DFT; correlaton I. INTRODUCTION Vdeo watermark s an emergng technology for copyrght protecton and authentcaton of dgtal vdeo. Accordng whether need the orgnal vdeo when extractng the watermark, vdeo watermark can be dvded nto blnd watermark and non-blnd watermark [1]. If extractng the watermark wthout the orgnal vdeo, then known as blnd watermark, on the contrary, called non-blnd watermark. In ths aspect, the method n [2] utlzes a blnd vdeo watermark scheme based on ndependent component analyss; the approach n [3] explots a blnd watermark algorthm based on three-dmensonal DWT vdeo. Embeddng doman of watermark can be categorzed nto two domans: spatal doman and transform doman [4]. LSB algorthm [5] s a common method n spatal doman. Bascally, watermark algorthm n transform doman can be dvded nto DFT-based algorthm, DCT-based algorthm and DWT-based algorthm, and the dea of realzaton s smlar, watermark value can be embedded by control specfc coeffcents of transform doman. The approach n [6] presents a watermark algorthm based on DFT transform whch selects groups of DFT coeffcents n the low-mddle frequency band, and the coeffcents n every group are dvded nto two sub-groups based on a pre-defned pattern, and the energy relatonshp between these two sub-groups s used to hde the watermark. The method n [7] present an approach usng the fast Fourer transform, whch added a template n the Fourer transform doman to render the method robust aganst rotaton and scalng, or aspect rato changes. The method n [8] proposes the dfferental energy watermark (DEW) algorthm for JPEG/MPEG streams whch embeds label bts by selectvely dscardng hgh frequency DCT coeffcents n certan mage regons. The approach n [9] presents a DCT-based Watson's mperceptve watermark model ressted downsamplng attack. The approach n [10] proposes a novel watermark scheme that can combat the MC-TFA attack. The approach n [11] presents an algorthm whch embeds watermark by usng DC coeffcent of block DCT. The method n [12] proposes a scheme whch ndcates watermark value through odd-even of the number of hgh frequency coeffcents equal 0. The method n [13] presents a method whch embeds watermark usng relatonshp of one coeffcent and average of some coeffcents. The method n [14] presents an algorthm can resst the JPEG attach and avod the some weaknesses of JPEG quantfcaton. The method n [15] do:10.4304/jmm.8.2.161-167
162 JOURNAL OF MULTIMEDIA, VOL. 8, NO. 2, APRIL 2013 proposes a novel oblvous mage watermarkng technque for the copyrght protecton and authentcaton of stll mages, based on Dscrete Multwavelet Transform (DMWT) and Quantzaton Index Modulus Modulaton (QIMM). The method n [16] proposes a novel audo zero-watermarkng scheme based on dscrete wavelet transform (DWT). The paper [17] has studed and mplemented three knds of DCT-based vdeo watermark algorthms and one DFT-based vdeo watermark algorthm, and experments show that the transparency of DCT algorthms s superor to the DFT algorthm; and the robustness of the DFT algorthm aganst geometrcal attacks s better than DCT algorthms, the robustness of DCT algorthms aganst codng attacks s better than the DFT algorthm. In order to meet people s ncreasng vsual demand, 3D vdeo s to be more wdely used. Nowadays there s lttle research on 3D vdeo watermark. The approach n [18] proposes a vdeo watermark scheme for 3D vdeo watermark, whch embeds watermark by modfyng the dsparty vectors n 3D vdeo. In ths paper, a new 3D vdeo watermark algorthm s proposed. Accordng the features of DCT-based and DFT-based vdeo watermark algorthms aganst attacks, and consderng 3D vdeo structural characterstc, a new robust blnd vdeo watermark algorthm n transform doman combnng wth 3D vdeo correlaton was proposed. DCT-based watermark and DFT-based watermark are embedded n left and rght channels of a vdeo frame by usng opposte embeddng rules, and the watermark can be extracted by ntegratng 3D vdeo correlaton, and more robust watermark s extracted from the DCT-based watermark and the DFT-based watermark by comparng ther performance. The extracton process of the algorthm does not requre the orgnal vdeo, so the algorthm s a blnd watermark algorthm. The algorthm can sgnfcantly mprove the robustness of 3D vdeo watermark aganst resoluton attack and codng attack. The paper s organzed as follows. Secton 2 descrbes three parts of our mprovement of exstng algorthm. Secton 3 s the descrpton of our watermark embeddng and extracton algorthm. Expermental results and conclusons are presented n Secton 4 and 5 respectvely. II.COMPOSITION OF ALGORITHM IMPROVEMENT 3D vdeo s bnocular stereo vdeo, and each frame contans left and rght vew mages whch s to left and rght eye vew respectvely, and two mages are smlar, but wth a certan parallax. In ths paper, accordng to the structural characterstcs of the 3D vdeo, DCT-based vdeo watermark algorthms and the DFT-based vdeo watermark algorthm wll be mproved, and the mproved algorthm conssts of three parts: A. Preferred to Use Better one of DCT and DFT Algorthm Test results n [14] show that the DCT algorthm and the DFT algorthm have dfferent robustness aganst dfferent attacks. In order to mprove the robustness of vdeo watermark, ths paper presents the watermark algorthm preferred to use better one of DCT and DFT algorthm. Algorthm dea: (1) Add standard nformaton at the begnnng of the orgnal watermark, and new watermark nformaton wth standard nformaton s obtaned. (2) The vdeo frame to be watermarked s dvded nto left part and rght part, whch were embedded the DCT-based watermark and DFT-based watermark respectvely; (3) When extractng watermark, the standard nformaton s extracted respectvely from left part and rght part usng DCT algorthm and DFT algorthm; (4) Calculatng and comparng the correct rate of standard nformaton obtaned from two algorthms, f the correct rate of the DCT algorthm s better, and then extractng all watermark nformaton from the left part usng DCT algorthm; otherwse, extractng all watermark nformaton from the rght part usng DFT algorthm. B. DCT Watermark Algorthm combnng wth 3D Vdeo Correlaton Embeddng watermark n left and rght mages of a frame usng DCT algorthm, and extractng watermark after attacks, and then t can be found that the BER obtaned from two mages under the same attack s not always the same. The test results show that the mpact of the same attack s dfferent n the same locaton of two mages. So sometmes t happens that n one mage an error occurs, but n the other mage the watermark data s stll correct. Therefore, n order to mprove the robustness of the DCT algorthm, DCT watermark algorthm combnng wth 3D vdeo correlaton s proposed. Algorthm dea: (1) Left and rght mages of the frame to be watermarked are dvded nto 4 4 blocks, and coeffcent matrx can be gotten after 2D DCT transform on each block, and C13 denotes the 13th coeffcent of the -th block, and C13 s chosen to embed watermark. (2) Opposte embeddng rules were used n left and rght mages, specfcally, n left mage uses C13 >0 to ndcate watermark value 1, and n rght mage uses C13 >0 to ndcate watermark value 0; (3) When extractng the watermark, extractng watermark respectvely from left and rght mages usng the DCT algorthm and opposte extracton rules; (4) When two watermark values from two mages are dfferent, ndcatng an error occurs n one mage, and the watermark value can be obtaned correspondng to relatonshp between C13 of left mage and C13 of rght mage.
JOURNAL OF MULTIMEDIA, VOL. 8, NO. 2, APRIL 2013 163 C. DFT Watermark Algorthm combnng wth 3D Vdeo Correlaton Embeddng watermark n left and rght mages of a frame usng DFT algorthm reference [6], and extractng watermark after attacks, and fndng a smlar concluson wth 2.2 that the mpact of the same attack s dfferent n the same locaton of two mages. Therefore, n order to mprove the robustness of the DFT algorthm, DFT watermark algorthm combnng wth 3D vdeo correlaton s proposed. Algorthm dea: (1) Frequency doman data of Left and rght mages s obtaned after the whole 2D DFT of each mage, and the data s dvded nto 4 4 block, and the -th blocks of two mages are selected to embed watermark. (2) Coeffcents n each block are dvded nto two sub-groups based on a pre-defned pattern; fnally the energy relatonshp between these two sub-groups s used to hde the watermark. Fg. 1 shows the locaton of the two sub-groups n a 4 4 block. mages of a vdeo frame usng opposte embeddng rules, and the watermark can be extracted by ntegratng 3D vdeo correlaton, and more robust watermark s extracted from the DCT-based watermark and the DFT-based watermark by comparng ther performance. A. Watermark Embeddng The dagram of watermark embeddng s shown n Fg.2, and the process of watermark embeddng s descrbed as follows: (1) Add standard nformaton at the begnnng of the orgnal watermark, and new watermark nformaton wth standard nformaton s obtaned. (2) Left mage of the frame to be watermarked s dvded nto left part and rght part, and rght mage of the frame s also dvded nto left part and rght part; (3) Left part of left mage and left part of rght mage compose a new 3D vdeo data, whch s embedded watermark usng DCT watermark algorthm combnng wth 3D vdeo correlaton. Fgure 1. the locaton of the two sub-groups In the -th 4 4 block, "O" ndcates the four elements of sub-group 1, and E1 denotes energy of sub-group 1, and the calculaton method s shown n (1); "X" ndcates the four elements of sub-group 2, and E2 denotes energy of sub-group 2, and the calculaton method s shown n (2) : E1 = E2 = 4 2 m= 3 n= 1 2 4 e( m, n) e( m, n) m n= 3 (2) (3) The opposte embeddng rules were used n left and rght mages, specfcally, n left mage uses E2 E1 to ndcate watermark value 1, and n rght mage uses E2 E1 to ndcate watermark value 0; (4) When extractng the watermark, extractng watermark respectvely from left and rght mages usng the DFT algorthm and opposte extracton rules; (5) When two watermark values from two mages are dfferent, ndcatng an error occurs n one mage, and the watermark value can be obtaned accordng to relatonshp between E1 and E2 of two mages. (1) III.PROPOSED WATERMARKING SCHEME In ths paper, a new robust blnd vdeo watermark algorthm n transform doman combnng wth 3D vdeo correlaton was proposed. DCT-based watermark and DFT-based watermark are embedded n left and rght Fgure 2. the dagram of watermark embeddng Embeddng process: (a) denotes the 13th coeffcent of the -th block n left mage, and RC13 denotes the 13th coeffcent of the -th block n rght mage, and W denotes the -th watermark value, and opposte embeddng rules were used n left and rght mages, and embeddng rule n left mage s defned n (3), and embeddng rule n rght mage s defned n (4) k W k W (3) RC13 k W RC13 k W (4) k s the strength of watermark, and k> 0. (b) Embeddng process n left mage When W =1, to ensure k, the embeddng rule s defned as follows: > k = k k k < k (5) When W =0, to ensure -k, the embeddng rule s
164 JOURNAL OF MULTIMEDIA, VOL. 8, NO. 2, APRIL 2013 defned as follows: > k = k k k < k (6) (c) Embeddng process n rght mage When W =1, to ensure RC13 -k; When W =0, to ensure RC13 k. (4) Rght part of left mage and rght part of rght mage compose a new 3D vdeo data, whch s embedded watermark usng DFT watermark algorthm combnng wth 3D vdeo correlaton. Embeddng process: (a) LE1 and LE2 denote energy of sub-group 1 and sub-group 2 of the -th block n left mage; RE1 and RE2 denote energys of sub-group 1 and sub-group 2 of the -th block n rght mage; and opposte embeddng rules were used n left and rght mages, and embeddng rule n left mage s defned n (7), and embeddng rule n rght mage s defned n (8): LE2 LE1 + k W LE1 LE2 + k W (7) RE1 RE2 + k W RE2 RE1 + k W (8) k s the strength of watermark, and k> 0. (b) Embeddng process n left mage When W =1, to ensure LE1 and LE2 meet the demand n (7), If not meet the demand embeddng rule s defned as follows: delta = ( k + LE1 LE2 ) / 2 LE2 = LE2 + delta LE1 = LE1 delta (9) When W =0, to ensure LE1 and LE2 meet the demand n (7), If not meet the demand embeddng rule s defned as follows: delta = ( k + LE2 LE1 ) / 2 LE1 = LE1 + delta LE2 = LE2 delta (10) (c) Embeddng process n rght mage When W =1, to ensure RE1 RE2 +k; When W =0, to ensure RE2 RE1 +k. B. Watermark Extracton The dagram of watermark extracton s shown n Fg.3, and the process of watermark extracton s descrbed as follows: (1) Left mage of the frame to be watermarked s dvded nto left part and rght part, and rght mage of the frame s also dvded nto left part and rght part; (2) Left part of left mage and left part of rght mage compose a 3D vdeo data, n whch extractng standard nformaton usng DCT watermark algorthm combnng wth 3D vdeo correlaton, and usng DCT-SI to denote the extracted nformaton. Extractng process: (a) Extractng watermark respectvely from left mage and rght mage, and extractng rule n left mage s defned n (11), and LW denotes the watermark value extracted from left mage; extractng rule n rght mage s defned n (12), and RW denotes the watermark value extracted from rght mage. Fgure 3. 1 LW = 0 1 RW = 0 the dagram of watermark extracton 0 < 0 RC13 < 0 RC13 0 (11) (12) (b) When LW and RW are the same, and LW denotes the fnal watermark value W ; (C) When LW and RW are dfferent, the fnal watermark value W can be obtaned accordng to relatonshp between and RC13. The rule s defned as follows: 1 W = 0 RC13 < RC13 (13) (3) Rght part of left mage and rght part of rght mage compose a 3D vdeo data, n whch extractng standard nformaton usng DFT watermark algorthm combnng wth 3D vdeo correlaton, and usng DFT-SI to denote the extracted nformaton. Extractng process: (a) Extractng watermark respectvely from left mage and rght mage, and extractng rule n left mage s defned n (14), and LW denotes the watermark value extracted from left mage; extractng rule n rght mage s defned n (15), and RW denotes the watermark value extracted from rght mage. 1 LE2 LE1 LW = 0 LE2 < LE1 (14)
JOURNAL OF MULTIMEDIA, VOL. 8, NO. 2, APRIL 2013 165 1 RE2 < RE1 RW = (15) 0 RE2 RE1 (b) When LW and RW are the same, and LW denotes the fnal watermark value W ; (C) When LW and RW are dfferent, and the watermark value can be obtaned accordng to relatonshp of LE1, LE2, RE1, RE2. The rule s defned as follows: 1 LE2 + RE1 LE1 + RE2 W = 0 LE2 + RE1 < LE1 + RE2 (16) (4)Calculatng and comparng the correct rate of DCT-SI and DFT-SI. (5) If the correct rate of DCT-SI s better, and then extractng all watermark nformaton from the left part of left and rght mages usng DCT algorthm combnng wth 3D vdeo correlaton; Otherwse, extractng all watermark nformaton from the rght part of left and rght mages usng DFT algorthm combnng wth 3D vdeo correlaton. IV. EXPERIMENTAL RESULTS AND PERFORMANCE ANALYSIS In ths paper, cactus, Flower, rose_sbs_dvx and Ume614x2x384 s used as a test set, and frame rate s 29/sec, and the length of watermarked content s 288 bts. A. Transparency Fg. 4 shows the ffth frame of 4 vdeos and watermarked frame used the proposed algorthm: (a) cactus orgnal (c) flower orgnal (e)rose orgnal (b)cactus watermarked (c)flower watermarked (f)rose watermarked dfferent vdeo, but the average of PNSR s better than 35db, ndcatng that the proposed algorthm has outstandng transparency. TABLE I. PNSR STATISTICAL RESULTS Vdeo name Resuloton PSNR1 PSNR2 PSNR3 cactus 1500 524 53.52 45.31 41.01 flower 720 262 38.74 33.63 30.04 rose_sbs_dvx 1024 384 49.34 40.47 36.08 Ume614x2x384 1024 384 49.69 40.52 36.02 Average of PSNR 47.82 39.98 35.79 B. Robustness In ths paper, watermarked vdeo s suffered wth resoluton attack and encodng attack, and calculatng bt error rate (BER) after extractng the watermark, and statstcal results are shown n Table II: TABLE II. TABLE II BER (%) AGAINST ATTACKS Vdeo name 60% rescale 50% rescale h.264- h.264-8000kbps 1150kbp cactus 0 0.182 0 3.875 flower 0 0 0 0 rose_sbs_dvx 0.034 0.398 0 0.034 Ume614x2x384 0 0.233 0 0 Average of BER 0.008 0.203 0 0.977 Embeddng strength of DCT method s 30 and embeddng strength of DFT method s 30000. It can be seen from Table II that the propose algorthm has excellent robustness aganst resoluton attack and codng attack. Comparng the proposed algorthm and other 3 algorthms whch nclude DCT-based algorthm, DFT-based algorthm, preferred to use better one of DCT and DFT algorthm (mentoned n 2.1), and expermental results are shown n Fg.5, and DCT denotes DCT-based algorthm, and DFT denotes DFT-based algorthm, and DCT+DFT denotes preferred to use better one of DCT and DFT algorthm, and DCTR+DFTR denotes the proposed algorthm. (g)ume orgnal (h)ume watermarked Fgure 4. Frames of 4 vdeos Embeddng strength of DCT algorthm s 30, and Embeddng strength of DFT algorthm s 30000. From Fg.4 t can be seen that the transparency of the proposed algorthm s excellent. PSNR statstcal results of the proposed algorthm are shown n Table I. Embeddng strength of DCT method s 10,30,50 and embeddng strength of DFT method s 10000,30000,50000 correspondng wth PSNR1, PSNR2, PSNR3. It can be seen from Table I that the propose algorthm usng the same strength has dfferent mpact on (a) BER under 60% resoluton attack
166 JOURNAL OF MULTIMEDIA, VOL. 8, NO. 2, APRIL 2013 results show that the proposed algorthm has excellent performance aganst resoluton attack and codng attack. (b) BER under 50% resoluton attack ACKNOWLEDGMENT Ths work was supported by the Scentfc Research Common Program of Bejng Muncpal Commsson of Educaton (KM201210015006, KM201210015007), Bejng Natural Scence Foundaton(4122026, 4122053), General Admnstraton of Press and Publcaton of the People's Republc of Chna (GXTC-CZ-1015004/15-1), the NSFC project (61003284). REFERENCES (c) BER under 8000kbps encodng attack (d) BER under 1150kbps encodng attack Fgure 5. BER under attacks From Fg. 5 t can be seen that sufferng wth resoluton attack and codng attack the propose algorthm has excellent robustness, and s obvously better than other 3 algorthms. V. CONCLUSIONS In ths paper, a new 3D vdeo watermark algorthm s proposed. Accordng to the features of DCT-based and DFT-based vdeo watermark algorthms aganst attacks, and consderng 3D vdeo structural characterstc, a new robust blnd vdeo watermark algorthm n transform doman combnng wth 3D vdeo correlaton was proposed. DCT-based watermark and DFT-based watermark are embedded n left and rght channels of a vdeo frame by usng opposte embeddng rules, and the watermark can be extracted by ntegratng 3D vdeo correlaton, and more robust watermark s extracted from the DCT-based watermark and the DFT-based watermark by comparng ther performance. The expermental [1] Yuan Wugang, "Research of Robust Vdeo Watermark Technology", WuHan: Huazhong Unversty of Scence and Technology, 2007. [2] SUN Jande and LIU Ju, "A Blnd Vdeo Watermark Scheme Based on Independent Component Analyss", ACTA ELECTRONICA SINICA, Vol.32, No.9, pp.1507-1510, Sep.2004. [3] Ce Wang, Chao Zhang, and Pengwe Hao, A BLIND VIDEO WATERMARK DETECTION METHOD BASED ON 3D-DWT TRANSFORM, 2010 IEEE 17th Internatonal Conference on Image Processng, 2010(9):3693-3696 [4] T.JAYAMALAR, V. RADHA. Survey on Dgtal Vdeo Watermark Technques and Attacks on Watermarks, Internatonal Journal of Engneerng Scence and Technology, Vol. 2(12),pp. 6963-6967, 2010. [5] Eugene T. Ln and Edward J. Delp, A revew of data hdng n dgtal mages, Proceedngs of the conference on Image processng, 1999(4): 274-278 [6] Dajun He, Qbn Sun and Q Tan. AN OBJECT BASED WATERMARK SOLUTION FOR MPEG4 VIDEO AUTHENTICATION. ICASSP 2003.537-540 [7] Shelby Perera, Joseph J. K. ORuanadh and Frederc Degullaume, Template Based Recovery of Fourer-Based Watermarks Usng Log-polar and Log-log Maps, Conf on Multmeda Computng and Systems, vol.1, pp.870-874, Jul.1999. [8] Gerrt C. Langelaar and Regnald L. Lagendjk, Optmal Dfferental Energy Watermark of DCT Encoded Images and Vdeo, IEEE TRANSACTIONS ON IMAGE PROCESSING. VOL.10, NO.1, pp.148-158, Oct.2001. [9] DU Yao-gang, CAI An-n2 and SUN Jng-ao, "DCT-based Vdeo Watermark Technque Usng Downsamplng", ACTA ELECTRONICA SINICA, Vol.33, No.12, pp.2219-2221, Dec.2005. [10] LIU Shao-hu, HAN Lu and YAO Hong-xun, "Vdeo watermark algorthm for resstng colluson attacks", Journal on Communcatons, Vol. 31, No. 1, pp.14-19, January, 2010. [11] Lu Lanshan and L Renhou, "Bl nd Vdeo Watermark Scheme Based on Drect Current Component n Dscrete Cosne Transformaton Doman", JOURNAL OF XI'AN JIAO TONG UNIVERSITY, Vol.40, No.4, pp.402-405, Apr.2006. [12] Qn Janjun, "Research of Dgtal Vdeo Watermark Technology". Chang Sha: Hunan Normal Unversty,2010 [13] Cao Junme, "A Robust Dgtal Watermark Algorthm Base on DCT", Mcrocomputer Applcatons VoL.16, No.1, pp.11-12, 2010.
JOURNAL OF MULTIMEDIA, VOL. 8, NO. 2, APRIL 2013 167 [14] Chen Huang-Ch; Chang Yu-Wen; Hwang Rey-Chue. "A watermarkng technque based on the frequency doman". Journal of Multmeda (JMM).Vol.7, No.1, pp.82-89,2012. [15] Leelavathy N, Prasad E.V., Srnvas Kumar S., Chandra Mohan, B. "Oblvous mage watermarkng n dscrete multwavelet doman usng QIMM". Journal of Multmeda(JMM).Vol.6, No.4, pp.359-368,2011. [16] Yang Yu, Le Mn, Lu Huaqun, Zhou Yaan, Luo Oun."A novel robust zero-watermarkng scheme based on dscrete wavelet transform", Journal of Multmeda.Vol.7, No.4, pp. 303-308, 2012. [17] DING Ha-yang. "Research and Implementaton on Blnd Vdeo Watermarkng Algorthms Based on DCT and DFT". Journal of Bejng Insttute of Graphc Communcaton. Vol. 20, No. 4,pp: 32-35,2012 [18] Zengnan Zhang, Zhongje Zhu, Lfeng X. Novel Scheme for Watermark Stereo Vdeo, Internatonal Journal of Nonlnear Scence, Vol.3, No.1, pp.74-80, 2007. Dng Ha-yang receved hs M.S. degree n sgnal and nformaton processng from Bejng Jaotong Unversty, Bejng n 2004, and hs B.S. degree n electroncs and nformaton technology from Bejng Jaotong Unversty, Bejng n 2001. He s currently a Ph.D. canddate at Bejng Unversty of Posts and Telecommuncatons, Bejng, Chna. Hs current research nterests nclude nformaton securty, dgtal mage processng, DSP technology. Yang Y-Xan receved hs Ph.D. degree n sgnal and nformaton processng from Bejng Unversty of Posts and Telecommuncatons n 1988. He s now the drector of Natonal Engneerng Laboratory for Dsaster Backup and Recovery, the drector of Key Laboratory of Network and Informaton Attack Defence Technology of MOE, the drector of Informaton Securty Center, and the Deputy Drector of State Key Laboratory of Networkng and Swtchng Technology. Meanwhle, he s a Changjang Scholars Dstngushed Professor and the edtor-n-chef of "Journal on Communcatons" and "The Journal of Chna Unversty of Posts and Telecommuncatons". Hs current research nterests nclude network and nformaton securty theory and applcatons, network-based computer applcaton technology, codng theory and technology. Zhang Ru was born n Chna n 1976. She s currently an assocate professor n School of Computer Scence and Technology of Bejng Unversty of Posts and Telecommuncatons. Her man research nterests nclude nformaton securty, dgtal watermarkng, cryptography, etc. Zhou Ya-Jan was born n Chna n 1971. He receved hs Ph.D. degree n 2003 n communcatons engneerng from Xdan Unversty at X'an. He receved hs M.S. degree n 1996 and B.S. degree n 1993, both from the Department of Materal Scence and Engneerng of Bejng Unversty of Aeronautcs and Astronautcs. He s currently an assocate professor n School of Computer Scence and Technology of Bejng Unversty of Posts and Telecommuncatons. Hs man research nterests nclude moble communcatons, securty of wreless networks, securty of databases, cryptography theory and ts applcaton, etc. He s now the project leader of two projects from the Natonal Natural Scence Foundaton of Chna and the Natonal Hgh Technology Research and Development 863 Program of Chna, spectvely. Meanwhle, he s n charge of a project from Bejng Muncpal Natural Scence Foundaton. He has publshed about 20 techncal papers and two books.