Content based Zero-Watermarking Algorithm for Authentication of Text Documents

Content based Zero-Watermarking Algorithm for Authentication of Text Documents Zunera Jalil 1 , Anwar M. Mirza 1 and Maria Sabir 2 1 FAST National Universit y of Computer and E merging Sciences, Islamabad, P akistan 2 Air University, Islamabad, Pakistan Abstract- Copyright protection and authentication of digital contents has become a significant issue in the current digital epoch with efficient communication mediums such as internet. Plain text is the rampantly used medium used over the internet for information exchange and it is very crucial to verify the authenticity of infor ma tion. There are very limited techniques available for plain text waterm arking and authentication. This paper presents a novel zero-w aterma rking algorithm for authentication of plain text. The algorithm generates a w aterma rk based on the text contents and this w atermark can later be extracted using extraction algorithm to prove the authenticity of text document. Experimental results de mon strate the effectiveness of the algorithm against tampering attacks identifying watermark accuracy and distortion rate on 10 different text samples of varying length and attacks. Keywords-watermarking; copyright protection; authentication; security; algorithm I. I NTRODUCTIO N Copyright protectio n and authentication of digital contents has gained more importance with the increasing use of internet, e-commerce, and other efficient communicatio n technologies. Besides, making it easier to access infor mation within a very short span of time, it has become difficult to protect copyright of digital contents and to prove the authenticit y of the obtained information. Digital contents mostly compr ises of text, image, audio, and video. Authentica tion and copyright protection of digital images, audio, and video has been given due thought by the researchers in past. However, authentication and cop yright protection of plain text has bee n neglected. Most of the digital contents like websites, e-books, articles, news, chats, SMS, are in the form of plain text. The threats of illegal copying, tampering, forgery, plagiarism, falsification, and other forms of possible sabotages need to be specifically addres sed. Digital watermarking is one of the solutions which have been used to authenticate and to protect d igital contents. Digital watermarking method s are used to identify the original cop yright owner (s) of the conte nts which can be an image, a plain text, an audio, a video or a combination of all. A digital watermark can be described as a visible or an invisible, preferabl y the latter, identification code that permanently is embedded in the data. It means that unlike conventional crypto graphic techniques, it remai ns prese nt within the data even after the decryption proce ss [1]. A text, being the simplest mode of com munication and information exchange, brings various challen ges when it comes to copyright protection and authentication. Any cha nges on text should preserve the value, usefulness, meaning, and grammaticality of the text. Short documents are more difficult to p rotect and aut henticate since a simple a nalysis would easily reveal the watermark, thus making text insecure. In image, audio, and video watermarking the li mitations of Human Visual and/or Human Auditory System are exploited for watermark embedding along with the inherent redundan cy. It is difficult to find such limitations and redundancy in plain text, since text is sensitive to any modification required to be made for water mark embedding. Text is easier to cop y, repro duce and tamper as compare d with images, audio and video. Text being a specialized medium req uires specialized copyright protectio n and authentication solutions. Tra ditional watermarking algorithms modify the contents of the digital medium to be protected by embedding a watermark. This traditional watermarking approach is not applicable for plain text. A specialized watermarking appro ach such as zero-water marking would do the needful for text. In this paper, we prop ose a novel zero - watermarking algorithm which utilizes the contents of text itself for its authentication. A zero-watermarking algorith m does not change the character s of original data, but utilize the characters of original data to construct ori ginal watermark information [2-3]. The paper is organized as follows: Section 2 provides an overview of the previo us work done on te xt watermarking. The proposed embedding and extr action algorithm are described in detail in section 3. Section 4 presents the experi mental res ults for the tampering (insertion, deletion and re-ordering) attacks with differe nt keywords on. Performance of the proposed algorithm is evaluated by co multiple text samples. The last section concl udes the paper along with directions for future work. II. P REVI OUS W ORK Text watermarking for a uthentication of text docume nts is an important area of researc h; however, the w ork done in this domain in past is very inadequate. The work on text watermarking initially started in 1991 . A number of text watermarking tec hniques have been pro posed since then. These include text watermarking using text images, synonym based , pre-supposition based, syntactic tree based, noun-verb based, (IJCSIS) International Journal of Computer Science and Information Security, Vol. 7, No. 2, February 2010 212 http://sites.google.com/site/ijcsis/ ISSN 1947-5500 word and sentence based, acro nym based, typo error based methods etc. The previous work on digital text w atermarking can be classified in the following categories; an image based approach, a syntactic appro ach, a semantic appro ach and the structural approach. Description of each category and the work done accordingly is as follows : A . A n I m a ge - B as e d A pp r o ac h In image based appro ach towards text watermarking, the image of text is takes as source for watermark embedding. Brassil, et al. were the first to propose a few te xt watermarki ng methods utilizing text image[4 ]-[5]. Later Maxemchuk, et al. [6]-[8] anal yzed the perfor mance of t hese methods. Low, et al. [9]-[10] further analyzed the efficiency of these methods. The first method was the line-shi ft algorith m which moves a line upward or downward (left or right) based on watermark bit values. The word-shift algorith m used the inter-word spaces to embed the watermark. The last method was the feature coding algorithm in which specific text features are tampered to encode w atermark bits i n the text. Huang and Yan [11] prop osed an algorithm based on an average inter -word distance in each line. The distances are adjusted according to the sine-wa ve of a specific phase and frequency. The feature and the pixel level algorithms were also developed which mark the documents by modifying the stroke features such as width or serif [12]. Text watermarking algorith ms using binary text image are not robust against re-typing attack and have li mited applicability. Authentication of text becomes eas y with text image, but it is mostly impractical to treat text as an image. In some algorith ms, watermark can be destroyed by a simple OCR (Optical Character Recognitions) analysis. The use of OCR obliterate the changes made to the spaces, margins and fonts of a text to embed watermark. B . A Sy n t ac t i c A pp r oa c h In this appro ach towards text watermarking, the synta ctic structure of text is used to embed watermark. Mikhail J. Atallah, et al. first p roposed th e natural lang uage watermarking scheme by using syntactic structure of text [13]-[14] where the syntactic tree is built and transfor mations are applied to it in order to embed t he watermark keeping all t he pro perties of text intact. The NLP techniques are used to analyze the syntactic and the semantic structure of text while perfor ming any transformations to embed the watermark bits. Hassan et al. perfor med morp ho-syntactic alterations to the text to watermark it [15]. The text is first transfor med into a syntactic tree diagram where text hierarch y and dependencies are analyzed to embed watermark bits. Hassan et al. provi ded an overview of available syntactic tools for text watermarking [16]. Text water marking by using syntactic structure of text and natural language processi ng algorith ms, is an efficient approach for text authenticatio n and copyright protection but progress in this do main is slower than the requirement. NLP is an immat ure area of researc h so far and using in-effici ent algorithms, efficie nt results in text watermarking cannot be obtained. C . A S e ma n ti c A pp r o ac h The semantic watermarking schemes focus on using the semantic contents of text to embed the watermark. Atallah et al. were the first to propose the semantic watermarking schemes in the year 2000 [17]-[1 9]. Later, the synonym substitution method was prop osed, in which watermark is embedded by replacing certain words with their synonyms [20]. Xingming, et al. p roposed nou n-verb based technique for text watermarking [21] where nouns and verbs in a sentence are parsed using grammar p arser and semantic networks. Later Mercan, et al. proposed an algorith m of the text watermarking by using typos, acronyms and abbreviation to embed the watermark [22]. Algorithms were develop ed to watermark the text using the linguistic semantic phenomena of presuppositions [23] by obse rving the discourse structure, meanings and representatio ns. The text pr uning and the grafting algorithms were also developed in the past. The algorithm based on text meaning represe ntation (TMR) strings has also been propo sed [24]. The text watermarking, based on semantics, is language dependent. The synonym bas ed techniques are not resilient to the random synonym substitutio n attacks. Sensitive nature of some documents e.g. legal do cuments, poetr y and quotes do not allow us to make semantic transfor mations rando mly because in these forms of text a simple transformation sometimes destro ys both the semantic connotatio n and the value of text[25]. D . A S t r u c t u r a l Ap p r o a ch This is the most recent appro ach used for copyright protectio n of text documents. A text watermarking algorith m for copyright protectio n of text using occurrences of doub le letters (aa-zz) in te xt to e mbed the watermark ha s recently been proposed [25]. The algorithm is a blend of encryption, steganography and watermarking. Howe ver, groups are formed by using full stop period in this algorith m. Text like poetr y, quotes, web contents, legal document may not essenti ally contain full stops; which makes this algorith m inapplicab le to all types of text. To overcome the shortco mings of this algorithm, another algorithm which use prepo sition besides double letters to watermark text ha s been prop osed [26]. The structural algorith ms are not applicab le to all types of text docume nts and are not designed specificall y to solve authentication problem; hence we prop ose a zero- watermarking algorithm which incorporates the contents of text for its protection. III. P ROPOSED A LG ORITHM The semantic and synta ctic watermarking algorith ms developed in past for plain text embed a watermark in the host text document itself which results in text qualit y, meaning and value degradation. We prop ose a zero-watermarking appro ach in which the host text doc ument is not altered to emb ed (IJCSIS) International Journal of Computer Science and Information Security, Vol. 7, No. 2, February 2010 213 http://sites.google.com/site/ijcsis/ ISSN 1947-5500 watermark, rather the chara cteristics of text are utilized to generate a watermark. T he watermark is fragi le in nature and is used to authenticate text documents. The watermark generation and extraction process is illustrated is fig. 1. Watermark is registered with the Certifying Authority (CA) and is used is the extractio n algorith m to authentica te text document. Fig 1: Overview o f Watermark Generation and Extraction Pro cesses The proposed algorith m utilizes the contents of text to protect it. A keyword from the text is selected based on author choice and a w atermark is generated based on the length of proceeding and next word length, to and from the keyword occurrences in text. This process is illustrated in fig. 1, where ‘is’ is the keyword and based on text contents, a watermark is generated. Fig 2: W atermark Generation It is a zero- watermarking scheme, since watermark is not actually embedded in the text itself; rather it is generated by using the character istics of text. The watermarking process involves two stages: (1) embedding algorith m and (2) extraction algorit hm. Watermark embedd ing is done by the original author and extraction done later by a Certify ing Authority (CA) to prove ownership. A trusted certifying authority is an essential requirement in this algorith m with whom, the original copyright owner register s his/her watermark. Whene ver the content/text ownership is in question, this trusted third p arty acts as a decisio n authority. A. Embedding Algorithm The algorithm which embeds the watermark in the text is called embedding algorithm. The watermark embedd ing algorithm require s original text file as input and keyword is selected by the original author /copyright owner. Keyword should be word having frequent occurrence in the text. A watermark is generated as output by this algorithm. This watermark is then re gistered with the certifying a uthority along with the original text d ocument, author name, keyword, current date and ti me. The algorithm proceed s as follows: 1. Read T O . 2. Count Occurrence of each word in T O . 3. Select KW based on occurrence frequency 4. KWCOUNT = Total occurrence count of KW in text T o 5. for i=1 to KWCOUNT, repeat step 6 to 8. 6. WM [j] = length (P i ) 7. WM [j+1] = length (N i ) 8. i=i+1 and j=j+1 9. Output WM T O = Original text; KW=keyword; KWCOUNT= keyword count ; WM= Watermark; P i = ‘Proceeding word’ of the ith occurrence of keyword (KW); N i = ‘Next word’ of the ith occurrence of keyword (KW) The original text (T O ) is first obtained from the author and occurrence frequency of each word in text is analyzed. A keyword is selected by the author which is typical a word with maximum occ urrence count in text. The procee ding and next word length for all occ urrence of keyword in text is analy zed and a numeric water mark is generated. This watermark is then registered with the CA with c urrent date and ti me. B. Extraction Algorithm The algorithm which extracts the watermark from the text is called extraction algorith m. The prop osed extraction algorithm takes the plain text and keyword as input. The text may be attacked or un-attacked . The watermark is generated from the te xt b y the extraction algorithm a nd is then, compared with the original watermark registered with the CA. We have also record ed author name, cur rent date and time with the CA. Multiple watermark registrat ion conflicts with CA can be resolved b y keeping record of ti me and date. T he author having former registration entry will be regarded as the original author. The watermark will be accuratel y detected by this algorith m in the absence of attack on text, and text docume nt will be called authentic text without tampering. The w atermark will get distor ted in the presence of tamperin g attacks with text. Tampering can be insertio n, deletion, paraphrasi ng or re- ordering of words and sentences in text. The extraction algorithm is as follo ws: (IJCSIS) International Journal of Computer Science and Information Security, Vol. 7, No. 2, February 2010 214 http://sites.google.com/site/ijcsis/ ISSN 1947-5500 1. Read T O or T A , WM and KW. 2. Count frequency of KW in given text. 3. KWCOUNT = Total occurrence count of KW in text 4. for i=1 to KWCOUNT, repeat step 5 to 7. 5. E WM [ j ] = length (P i ) 6. EWM [j+1] = length (N i ) 7. i=i+1 and j=j+1 8. if (EWM not equals WM) Tamper = YES 9. Output EWM. T O = Original text; T A = Attacked text; KW=keyword; KWCOUNT= keyword count ; EWM= Extracted Watermark; P i = ‘Proceeding word’ of the ith occurrence of keyword (KW); N i = ‘Next word’ of the ith occurrence of keyword (KW) IV. E XPERIMENTAL R ESULT S We used 10 samples of variable size text from the data set designed in [25] for our experiments. These samples have b een collected from Reuters’ corpus, e-books, and web pages. Insertion and deletion of words and sentences was performed at multiple rando mly selected locations in text. Table I show the sample label number as in dataset [25], number of words in original text, the insertion and deletion volume, and the number of words in the text after attack. TABLE I O RIGINAL AND A TTACKED T EXT S AMPLES W ITH I NSERTION AND D ELETION R ATIOS Sample No. Original Text Attack Attacke d Text WC Insertion Deletion Word Count 1 : [SST2] 421 26% 25% 425 2 : [SST4] 179 44% 54% 161 3: [MST2] 559 49% 25% 696 4: [MST4] 2018 14% 12% 2048 5: [MST5] 469 57% 53% 491 6: [LST1] 7993 9% 6% 8259 7: [LST3] 1824 26% 16% 2008 8: [LST5] 16076 9% 5% 16727 9: [VL ST3] 51800 11% 7% 53603 10: [VLST5] 67214 7% 5% 68853 The number of occ urrences of three different keywords “and”, “of”, and “in” was analyzed in the original and attacked text samples. These keywor ds were selected because of frequent occurrence s in all text samples. Watermark Accurac y Rate (WAR) and Water mark Distortio n Rate (WDR) are calculated as per the followin g formulas: WAR = N umber of characters corre ctly detected Number of watermark characters WDR = 1 - WAR The values of WAR ranges between 0 (the lowest) and 1(the highest) with desira ble value close to 1. The values of WDR also ranges between 0 (the highest) and 1(the lowest) with value close to 0 as desirabl e value. WAR of the extrac ted watermark was compared with the original watermark and tamper detection state was analyzed. Tab le II, III, and IV shows the WAR with keywords ‘and’, ‘of’, and ‘in respectively. WC 0 and WC A indicate the keyword count in original and attacked text respe ctively. TABLE II A CCURACY OF EXTRACTED WATERM ARK WITH KEYWORD ' AND ' Sample No. ‘and’ Tamper Detection WAR WC O WC A 1 : [SST2] 12 10 Yes 0.1538 2 : [SST4] 8 6 Yes 0.4736 3: [MST2] 8 7 Yes 0.2941 4: [MST4] 59 55 Yes 0.1935 5: [MST5] 19 13 Yes 0.3333 6: [LST 1] 257 264 Yes 0.1248 7: [LST 3] 45 51 Yes 0.4190 8: [LST 5] 286 299 Yes 0.1868 9: [VLST3] 858 915 Yes 0.1717 10: [VLST5] 1031 1053 Yes 0.1326 It can be observed in table II and III that tampering with text is always detected and low accuracy of watermark indicates that the extent to which text has been attacked. In table IV, the accuracy rate of watermark in sample 4 is 0.225 4, even with same frequency counter of keyword ‘in’ in both original and attacked texts. It depicts the fact that even if the frequency counters of keyword remain intact, the prob ability of getting same proceeding and next word length for all occurre nces of keyword is ver y low. TABLE III A CCURACY OF EXTRACTED WATERM ARK WITH KEYWORD ' OF ' Sample No. ‘of’ Tamper Detection WAR WC O WC A 1 : [SST2] 18 23 Yes 0.2380 2 : [SST4] 7 5 Yes 0.6153 3: [MST2] 9 12 Yes 0.0952 4: [MST4] 64 76 Yes 0.1582 5: [MST5] 7 9 Yes 0.2352 6: [LST 1] 237 255 Yes 0.1526 7: [LST 3] 38 51 Yes 0.1125 8: [LST 5] 571 599 Yes 0.1680 9: [VLST3] 2110 2229 Yes 0.1323 10: [VLST5] 2251 2351 Yes 0.1407 Figure 3 (a), (b), and (c) sho ws the watermark distortion rate (WDR) with keyword ‘and’, ’of’, and ‘in’ on all text sample s. (IJCSIS) International Journal of Computer Science and Information Security, Vol. 7, No. 2, February 2010 215 http://sites.google.com/site/ijcsis/ ISSN 1947-5500 It can be clearly o bserved that watermark distortion rate is ver y high eve n when insertion and deletion volume is low (as in case of sample 8, 9, and 10) for all three keywords. Text is sensitive to any modifications made by the attacker. High distortion rate indicates that the text has been tampered and is not authentic. This proves that the accurac y of watermark gets adversely affected even with minor tampering and w atermark fragility proves that te xt has been attacked. TABLE IV A CCURACY OF EXTRACTED WATERM ARK WITH KEYWORD ' IN ' Sample No. ‘in’ Tamper Detection WAR WC O WC A 1 : [SST2] 12 11 Yes 0.2962 2 : [SST4] 4 5 Yes 0.3636 3: [MST2] 11 15 Yes 0.40 4: [MST4] 34 34 Yes 0.2253 5: [MST5] 4 14 Yes 0.1818 6: [LST 1] 169 162 Yes 0.1680 7: [LST 3] 25 27 Yes 0.0909 8: [LST 5] 287 291 Yes 0.3117 9: [VLST3] 904 929 Yes 0.1354 10: [VLST5] 1162 1206 Yes 0.1266 V. C ONCLU S ION The existing text watermarking solutions for text authentication are not applicab le under random tamper ing attacks and on all types of text. With the small volume of attack, it becomes impossible to id entify the existence of at tack and to prove authenticity of infor mation. We have developed a zero-text watermarking algori thm, which utilizes the contents of text to generate a watermark and this watermark is later extracted to prove the authenticity of text document. We evaluated the performance of the algorith m for random tampering attack in dispe rsed form on 10 variable size text samples. Results show that our algorith m always detects tampering even when the tampering volume is lo w. A CKNOWLEDGMENT One of the authors, Ms. Zunera Jalil , 041-101673 -Cu-014 would like to acknowledge the Higher Educatio n Commission of Pakistan for providing the funding and resource s to complete this w or k under Indigenous Fello wship Program. R EFERENCES [1]. Asifullah Khan, Anwar M. Mirza and Abdul Majid, “Optimizing Perceptual Shaping of a Digital Watermark Using Genetic Fig. 3. Watermar k distortion rate w ith keyword (a)'and',(b) ‘o f’, and (c) ‘in’, with all text sampl es. (IJCSIS) International Journal of Computer Science and Information Security, Vol. 7, No. 2, February 2010 216 http://sites.google.com/site/ijcsis/ ISSN 1947-5500 Programming”, Iranian Journal of Electrical and Computer Engineering, vol. 3, pp. 144-150, 2004. [2]. Anbo Li, Bing-xian Lin, Ying Chen, “Study on copyright authentication of GI S vector data based on Ze ro-watermarking”, The I nternational Archives of the Photogramme try, Remote Sensing and Spatial Information Sciences. Vol . VII. Part B4, pp.1783-1786, 2 008. [3]. Zhou, Xinmin , Zhao, Weidong, Wang, Zhicheng, Pan, Li , “Security theory and attack analy sis for text wate rmarking”, 2009 International Conference on E-Business and Information System Security, EBISS 2009. [4]. J. T. Brassil, S. Low, N. F. Maxemchuk, and L. O’Gorman, “Electronic Marking and Identification Techniques to Discourage Document Copying,” IEEE Journal on Selected Areas in Communications, vol. 13, no. 8, October 1995, pp. 1 495-1504. [5]. J. T. Brassil, S. Low, and N. F. Maxemchuk, "Copyright Protection for the Electronic Distribution of Text Documents," Proce edings of the IEEE, vol. 87, no. 7, July 1999, pp.1181-1196. [6]. N. F. Maxemchuk, S. H. Low, “Performance Comparison of Two Text Marking Methods,” IEEE Journal of Selected Areas in Communications (JSAC) , May 1998. vol . 16 no. 4 1998. pp. 561-572. [7]. N. F. Maxemchuk, “Ele ctronic Document Distribution,” AT&T Technical Jour nal, September 1994 , pp. 73-80. 6. [8]. N. F. Maxemchuk and S. L ow, “Marking Text Documents,” Proceedings of the IEEE Internationa l Conference on Image Processing , Washington, DC, Oct. 26-2 9, 1997, pp. 13-16. [9]. S. H. L ow, N. F. Maxemchuk, and A. M. Lapone, "Document Identification for Copyr ight Protection Using Centroid Detection," IEEE Transactions on Communications , Ma r. 1998, vol. 46, no.3, pp 372-381. [10]. S. H. Low and N. F. Maxemchuk, “Capacity of Text Marking Channel,” IEEE Signal Processin g Letters , vol. 7, no. 12 , Dec. 2000, pp. 3 45 -347. [11]. D. Huang and H. Yan, "Interword distance changes represented by sine waves for watermarking text images," IEEE Trans. Circuits and Systems for Video Tec hnology , Vol.11, No. 12, pp.123 7-1245, Dec 2001. [12]. T. Am ano and D. Misaki, "A feature calibration met hod for watermarking of document images," Proceedings of ICDAR , pp.91-94, 1999. [13]. M. J. Atallah, C. McDonough, S. Nirenburg, and V. Raskin, “Natural Language Processing for Information Assurance and Security: An Overview and I mplementations”, Procee dings 9th ACM/SIGSA C New Security Paradigms Workshop, September, 2000, Cork, Ireland, pp. 51– 65. [14]. M. J. Atallah, V. Raskin, M. C. Crogan, C. F. Hempelmann, F. Kerschbaum, D. Mohamed, and S.Naik, “Natural language watermarking: Design,an alysis, and a proof-of-concept impleme ntation”, Proceedings of the Fourth Information HidingWorkshop, vol. LNCS 2137, 25-27 April 2001, Pitts burgh, PA. [15]. Hassan M. Meral et al., “Natural language wate rmarking via morphosyntactic alterations”, Computer Speech and Language, 23, 107- 125, 2009. [16]. Hasan M. Meral, Emre Sevinç, Ersin Ünkar, Bülent Sankur, A. Sumru Özsoy , Tunga Güngör, Syntactic tools for text watermarking, 19th SPI E Electronic Imaging Conf. 6505: Security, Steganography, and Watermarking of Multimed ia Contents, Jan. 2007, San Jose. [17]. M. Atallah, C. McDonough, S. Nirenburg, and V. Raskin, “Natural Language Processing for Information Assurance and Security: An Overview and I mplementations,” Procee dings 9th ACM/SIGSA C New Security Paradigms Workshop, September, 2000, Cork, Ireland, pp. 51– 65. [18]. M. Atallah, V. Raskin, C. F. Hempelmann, M. Kar ahan, R. Sion, U. Topkara, and K. E. Triezenberg, “Natural Language Watermarking and Tamperproofing”, Fifth Infor mation Hiding Workshop, vol. LNCS, 2578, October, 2002, Noordw ijkerhout, The Netherlands, Springer-V erlag. [19]. M. Topkara, C. M. Taskiran, and E. Delp, “Natural language watermarking,” Proceedings of the SPIE International Confere nce on Security, Steganography, and Water marking of Multimedia Contents VII, 2005. [20]. U. Topkara, M. Topkara, M. J. Atallah , “ The Hiding Virtues of Am biguity: Quantifiably Resilient Watermarking of Natural Language Text through Sy nonym Substitutions”. In Proceedings of ACM Multimedia and Security Conference, Geneva, 2006. [21]. Xingming Sun, Alex Jessey Asiimwe. Noun-Verb Based Technique of Text Watermarking Using Recursive Decent Semantic Net Parsers. Lecture Notes in Computer Science (LNCS) 3612: 958-961, Springer Press, August 2005. [22]. To pkara, M., Topraka, U., Atall ah, M.J., 2007. Information hiding through erro rs: a confusing approach. In: Delp III, E.J., Wong, P.W. (Eds.), Security, Steganography, and watermarking of Multimedia Contents IX. Proceedings of SPI E-IS&T Electronic Imaging SPIE 6505. pp. 65050V-1–65050V-1 2. [23]. B. Macq and O. Vybornova, “A method of text watermarking using presuppositions,” in Proceedings of the SPIE Inter nati onal Confere nce on Security, Steganography, and Watermarking of Multimedia Contents, January 2007. [24]. Peng L u et al., “An optimized natural language watermarking algorithm based on TMR”, on proceedings of 9 th I nternational Conference for Young Computer Scientists, 2009. [25]. Z. Jalil and A.M. Mirza, “A Novel Text Water marking Algorithm Based on Double Letters”, International Journal of Computer Mathematics. (Submitted) [26]. Z. Jalil and A. M. Mirza, “A Preposition based Al gorithm for Copyright Protection of Text Documen ts”, Jo urnal of the Chinese Institute of Engineers. (Submitted) (IJCSIS) International Journal of Computer Science and Information Security, Vol. 7, No. 2, February 2010 217 http://sites.google.com/site/ijcsis/ ISSN 1947-5500