Watermark Design Pattern for Intellectual Property Protection in Electronic Commerce Applications

Transcription

1 Design Pattern for Intellectual Property Protection in Electronic Commerce Applications S.H. Kwok Department of Information and Systems Management The Hong Kong University of Science and Technology Clear Water Bay, Kowloon, Hong Kong C.C. Yang Department of Systems Engineering and Engineering Management The Chinese University of Hong Kong Shatin, New Territories, Hong Kong K.Y. Tam Department of Information and Systems Management The Hong Kong University of Science and Technology Clear Water Bay, Kowloon, Hong Kong Abstract In this paper, we propose watermark design patterns to model the characteristic of digital watermark in electronic commerce applications. A watermark design pattern preserves the properties of a digital watermark in binary tree representation. Applying the survey results and characteristics of existing watermarking techniques, watermark design patterns for four media data - video, image, audio, and text are derived and expressed as WDP video, WDP image, WDP audio and WDP text respectively. Our results show that watermarks for video and audio data have to be invisible and robust, and watermarks for text data has to be invisible and fragile, while watermark for image has to be either visible and fragile, or invisible and robust for electronic commerce applications. The watermark design pattern is further used to relate electronic commerce applications and digital watermarking techniques. This facilitates electronic commerce application developers to select adequate digital watermarking techniques and off-the-shelf systems for their development. 1. Introduction With the growing number of electronic commerce web sites or applications, intellectual property protection is a pressing concern for content owners who are exhibiting digital representations of photographs, rare books and manuscripts, and original artworks on the WWW. Electronic commerce web sites or applications include electronic publishing and advertisement, real-time information delivery, product ordering, transaction processing, photograph galleries, digital libraries, Web newspapers and magazines, network video and audio, personal communication and so on. In electronic commerce web sites or applications, digital contents can be categorized into four basic types of media data, and they are image, audio, video and text. Multimedia data will not be included in this paper, as this is considered to be a combination of the basic data types. ing technique [25] is viewed as an enabling technology to protect these media data from re-use without adequate credit or in an unauthorized way. In general, watermarking [14] enables ownership assertion, fingerprinting, authentication and integrity verification, content labeling, usage control and content protection. Hawkins [11] addressed that many watermarking techniques have been proposed for intellectual property and copyright protection in the literature, but different media data apply different digital watermarking techniques. Moreover, [14] technical requirements of watermarking technique also vary from functions of applications. Recent research trend in watermarking technique has been focusing on image data [5, 6, 9, 15, 29] but watermarking is not limited to images; there are also watermarking techniques for audio [3, 23], video [20, 21, 22], and text [4, 12, 13, 24, 26, 28] data. ing for black and white text data; e.g., electronic documents and manuscripts, is so-called binary watermarks [17], and is /00 $10.00 (c) 2000 IEEE 1

2 similar to visual cryptography [16], which was a technique proposed for information hiding. In audio watermarking, some proprietary commercial solutions are found promising. MusiCode [2] is a commercial system that uses an inaudible and indelible watermarking technology for audio data. Another watermarking technique, such as Cox et al. [7] targets a wide spectrum of media data, but only the fundamental concepts of the technique are given. Many commercial watermarking systems usually equip with different watermarking techniques for specific applications and specific media data, and their watermarks are of different characteristics. For example, MusiCode [2] uses robust and inaudible watermarking to protect audio data. Signafy [18] protects image data using invisible and robust watermark. Digimarc [8] applies watermarking for copyright protection, as well as promotion and marketing purposes. Its watermark is robust and invisible for image and video data. Digital watermarking is still an evolving and developing technology, although many watermarking techniques and systems are available. It is difficult for electronic commerce application developers, who are usually not familiar with digital watermarking to choose appropriate watermarking techniques for their applications. In this paper, we propose watermark design pattern to describe the characteristic of digital watermark for intellectual property protection. Survey results enable us to derive different watermark design patterns for different media. These watermark design patterns are then extend to describe the relationship between electronic commerce applications and watermarking techniques. The rest of the paper is organized as follows: Section 2 describes parameters of watermark, general watermarking procedures, general applications and requirements, as well as masking schemes of digital watermarking. In Section 3, we discuss the property and copyright protection issue in electronic commerce applications and introduce common copyright protection schemes used in existing electronic commerce applications. Section 4 proposes a watermark design pattern to describe the characteristic of digital watermark for a particular media data. After consolidating the survey results, watermark design patterns for four different media data are derived. In Section 5, we apply our results of watermark design pattern to form the relationship between electronic commerce applications and digital watermarking techniques. Finally, Section 6 summarizes the contributions of our study and highlights the future research directions in this area. 2. ing Techniques Digital watermarking is a new and quite promising technology that can contribute to the battle against piracy in the information age. Although the development of digital watermarking is still at its infant stage, digital watermarking has already attracted many firstrate researchers. Many watermarking techniques [3, 4, 5, 6, 9, 12, 13, 15, 20, 21, 22, 23, 24, 26, 28, 29] and commercial watermarking systems [2, 8, 18, 24] have been proposed to focused on the copyright protection of digital media data. Four parameters are used in the categorization of digital watermark [14]; they are perceptible, imperceptible, robust and fragile. Some of these parameters are mutually exclusive; for example, a watermark cannot be perceptible and also imperceptible. The four parameters can be defined as follows. Perceptible - Perceptible watermarking embeds data intended to be visible or audible. [14] There are two important criteria for a good perceptible watermark. First, it must be difficult for an unauthorized person to remove it. A good perceptible watermark also has to resist falsification. Since it is relatively easy to embed a pattern or logo into a data, we have to ensure the perceptible watermark was indeed inserted by the claimed user. Imperceptible - Imperceptible watermarking embeds data intended to be invisible or inaudible but that can be extracted by a computer. Robust - Robust watermarks resist common signal processing procedures, such as cropping, filtering and are useful for ownership assertion purposes. Two major classes of robust watermarking techniques are private and public (or oblivious). A private scheme requires an original or reference content in the watermark detection procedure; a public scheme does not. Public schemes are attractive for many electronic commerce applications, as it is impractical to use private scheme. An application in which a digital watermark is used to identify proper ownership needs a very robust watermark [1]. Fragile - Fragile watermarks are easily corrupted by any form of processing procedures. An application used only to indicate modifications of the content needs only a fragile watermark [1] /00 $10.00 (c) 2000 IEEE 2

3 2.1 ing Procedures Three basic watermarking procedures are required in digital watermarking and they are watermark insertion, watermark detection and watermark extraction. In general, watermark insertion requires (i) an original data, (ii) a watermark, and (iii) and a private key. The output is the watermarked data as shown in Figure 1. The watermark can be information about ownership, user identity, description of the original data, etc. The watermark insertion procedure is to embed a watermark into the original data. The watermark can be perceptible or imperceptible in the watermarked data depending on the applications. For applications requiring the original data not being distorted, imperceptible watermark is desired. For some other applications, which require displaying the embedded data, a perceptible watermark is preferred. Original Data Insertion ed Data Private Key Figure 1. Insertion. extraction and watermark detection as shown in Figure 2 and Figure 3 are used to retrieve the embedded watermark from the watermarked data. For watermark extraction, a public key is used together with the watermarked data to retrieve the embedded watermark. For watermark detection, a public key and a specified ID - watermark are used together with the watermarked data to determine whether the watermark is correct. ed Data Extraction Public Key Figure 2. Extraction. ed Data A specified ID, watermark Detection Public Key Correct / Incorrect Figure 3. Detection. 2.2 General Applications and Requirements Digital watermarking has many different techniques. ing techniques are usually designed for specific applications and may not be applicable for other applications. We shall first review a list of possible applications and then identify the requirement [19] for the specific applications Applications Copyright protection: Embedding the ownership of the information when the information is being duplicated or abused. Usage/copy tracking: Verify the usage and copy of the information by the embedded data. Metadata or additional information: Embedding data to describe the information, e.g., structure, indexing terms, etc. Multiple data embedding: Embedding smaller image in a larger image or multiple audio data in a video. Post-processing of watermarked data: Some watermarked data are required to undergo some lossy signal-processing operations. For examples, images for storage and transmission may perform lossy coding operation in order to reduce bit rates and increase efficiency Requirements It is noted that one of the basic requirements for digital watermarking is to maintain the quality of the original data not being distorted when a watermark is embedded into it. Besides, there are other requirements that are necessary for specific applications. Perceptual transparency and robustness are the most common requirements for applications. Some other requirements, such as capability to recover data without original data, bit rate of data embedding algorithm, security, unambiguous proof of ownership, may be required for specific applications. Perceptual transparency: Perceptual transparency means that the embedded watermark is undetectable perceptually. If humans cannot differentiate between the original data and the watermarked data, the watermark is imperceptible. Blind tests are usually conducted to measure the perceptual transparency, where human subjects are /00 $10.00 (c) 2000 IEEE 3

4 presented with data with or without watermark randomly and asked to determine the quality. Perceptual transparency is required for applications in copyright protection, usage tracking and embedded metadata. Robustness: The watermarked data may be processed by some signal processing operations. After operations, the watermarked data is modified or manipulated, the embedded watermark may have been destroyed. The robustness of watermarking ensures that the embedded watermark will not be destroyed after such operations. As a result, third parties are not able to modify the watermarked data to thwart detection of the embedded watermark. A watermarking technique that is not robust is called fragile watermarking. Applications, such as, data storage and transmission may perform coding operations in order to reduce bit rates. Robustness is important in these applications. Capability to recover data without original data: The amount of watermark embedded in the original host data greatly depends on the availability of the original data during the extraction of the watermark. It also affects the difficulties of the data extraction. The capability of recovering data without original data is required for applications in copyright protection, and copy tracking. Bit rate of watermarking algorithm: The ability of embedding a significant fraction of the amount of data in the original data depends critically on the capability of the watermarking algorithm in adapting its insertion strategy to the underlying original data. In some applications, such as embedding a smaller image in a larger image or embedding multiple speech signals in a video, a large fraction of amount of data are embedded in the original data. Security: In order to protect information from unauthorized user, the presence of the watermark is only detectable with a secret key. The secret key controls the insertion of the watermark in the original data. In this situation, knowing the exact watermarking algorithm does not help in extracting the watermark unless the authorized person offers the secret key. Security is necessary for applications that require protection of the watermark. In convert communication scenarios, the watermarking procedures may be interrupted in the same way as encryption techniques. Unambiguous proof of ownership: The proof of ownership is the application that draws most attention in watermarking technologies. Unfortunately, no current watermarking technique is able to unambiguously determine the ownership of a data if it does not use the original or copy in the detection process. can be added to the watermarked data that correlates well. In such case, multiple ownership may be claimed, which is also known as a deadlock problem. Unambiguous proof of ownership is critical for copyright protection. 2.3 Masking: Data Insertion Data is inserted to the original data in watermarking. In general, data insertion is a procedure, called masking. Since the human auditory and visual systems are imperfect, data can be added to the original without being detected by human. Masking is a technique that makes a component in a multimedia data becomes imperceptible or inaudible in the presence of another data, called the masker. Data embedding and watermarking are possible because of the limitation of human visual and auditory systems. Human can only detect visual or audio signals with intensity or contrast above a minimum level. Therefore, most data embedding and watermarking techniques exploit the characteristics of the human auditory systems (HAS) or human visual systems (HVS) directly or indirectly. In audio masking effect, a faint data becomes inaudible in the presence of the masker, a louder audible sound. The audio masking exploits spectral and temporal characteristics. Frequency masking masks between frequency components. The masking threshold of a masker depends on frequency, sound pressure level, tone-like or noise-like characteristics of both the masker and the masked data. Temporal masking includes pre-masking and post-masking. Pre-masking/post-masking render weaker signals inaudible before/after the stronger masker is turned on/off. In visual masking effect, it works similar to the audio masking effect. Frequency masking and spatial masking are used in visual masking. Frequency masking uses a spatial sinusoidal pattern to lower the detectability of another sinusoidal pattern which frequency is close to that of the original spatial sinusoidal pattern. The spatial masking predicts the masking effects near edges and in uniform background. The upper channel of the spatial masking model can be linearized to obtain the tolerable error level for each spatial coefficient given that the modification of the image is small /00 $10.00 (c) 2000 IEEE 4

5 3. Electronic Commerce Applications Electronic Commerce Web Sites NIST Gallery ublic_affairs/gallery/ galindex.htm Aperture Gallery om/gallery/970516pix /970516A2.html National Library of Medicine OnLine s from the History of Medicine ov/ Apple Computer /quicktime/showcase/l ucasfilm/index.html ABC News m/sections/us/video_i ndex/video_index.htm l MP3 Radio Free Music coolsite.com/ Truman Library rary.org/photos/avphoto.htm IBM Intellectual Protection Network m.com/ Data Type Video Video Audio Audio Text Text Digital Contents Scientific Graphic Gallery Medicine Trailer and Music Video NewsClip Music and Song Music and Song Document ary Material Patent Document Table 1. Some electronic commerce web sites with their digital media data types and their digital contents. When digital watermarks are used for intellectual property protection, many electronic commerce applications are beneficial and they include the online and offline distribution of multimedia content, broadcast services, document verification, ownership identification and so on. This also benefits content creators artists, authors, and movie studios; content providers photo stock archives, libraries, and professional photographers; electronic commerce and graphics software vendors; and manufacturers of digital still images, video cameras and digital video discs (DVDs). Numerous electronic business web sites are found in the Internet and nine general and typical web sites were chosen in our discussion. These web sites were chosen for discussion because they contain all types of data, as well as they are typical electronic commerce applications. The locations of the web site together with the media data types and types of digital contents are listed in Table 1. a. A sample image from National Library of Medicine OnLine s from the History of Medicine b. A sample image from Aperture Figure 4. Samples of perceptible watermark. These web sites all offer digital contents for download, yet only a few appear to take any precautionary measures to prevent unauthorized or illegal further use of their digital contents. Some of these sites only include copyright statements and conditions of use which explained in explicit detail what kind of uses are, and are not, allowable for that particular assortment of digital contents. However, this appears not to be an aggressive approach for protecting their intellectual properties. A relatively interesting copyright protection attempt is found at the site of the /00 $10.00 (c) 2000 IEEE 5

6 National Library of Medicine OnLine s from the History of Medicine. To prevent unauthorized re-use of the image files they provide, they have resorted to drawing diagonal lines through all the images known or thought to be copyrighted and a sample image is given in Figure 4a. In the terminology of digital watermarking, this refers to a visible watermark. Use of proper visible digital watermark is also found at the Aperture Gallery, which stamps the symbol of Copyright, on all images in the gallery, to protect against unauthorized use. A sample image is shown in Figure 4b. 3.1 Types of Digital Contents Nine electronic web sites, together with the data types of their corresponding digital contents are tabulated in Table 1. There are four basic data types of digital contents in these electronic commerce sites, namely image, video, audio and text. Common file formats for the four basic data types include JPEG and GIF for image; AVI, MOV and RA for video; MP3, WAV, and RA for audio, and PDF, PS and TIF for text. 4. Design Pattern As watermarking technology is still a relatively young research area and the past works were mainly devoted to specific applications and media data, it is difficult for developers who are new to the technology to select suitable watermarking technology to protect digital contents in their applications. To facilitate and encourage them to use the watermarking technology in more convenient way, this paper proposes a mechanism or tool to relate off-the-shelf watermarking techniques and systems, and electronic commerce applications. Developers who are willing to use digital watermarking can easily identify appropriate techniques and commercial systems that best fit for their applications by referring to the relationship diagram produced in Section 5. A watermark design pattern (WDP) introduced in this paper describes the characteristic of a digital watermark for a particular media data. For instance, a watermark design pattern for image data can be invisible and fragile, whilst a watermark design pattern for audio data can be inaudible and robust. Different watermark design patterns are used in different applications due to their differences in digital contents and media data used in applications. A general WDP is described and presented in Figure 5. The general watermark design pattern - WDP media data is a complete structure that covers all possible watermark design patterns that may be found in electronic commerce applications, regardless the media data type of digital contents. The characteristic of digital watermark, described in the general WDP, expresses that it can be either perceptible or imperceptible, and for either case, it can also be robust or fragile. For example, a watermark is perceptible and robust if it follows the leftmost path, indicated by the thick arrow in Figure 5. We conducted a survey on nine different electronic commerce web sites listed in Table 1 to investigate the relationship between watermark design pattern and media data. The focus of this paper is on copyright protection. Among the nine web sites, three sites are devoted to image data, and two sites are devoted to each video, audio, and text data. These web sites are typical media distribution sites, and we assume other media distribution web sites are of the same characteristics and requirements and therefore the results of this paper are applicable to them as well. After visiting and studying these sites and their media data, we learnt the phenomenon of these sites, the properties of their digital contents and characteristics of required watermarks. Applying the survey results to the general WDP, and characteristics of existing watermarking technology, we derived a set of four WDPs for various media data. It is evident that although imperceptible watermarks dominate in most of these sites, perceptible watermarks were found to be useful for impressive samplings of digital image collections. Samples of perceptible watermark are already given in Figure 4. It is noted that watermarks for audio and video data have to be inaudible and invisible respectively, as listeners and watchers like to enjoy these types of media in a noise-free environment. It is because perceptible watermark could be a source of noise from the user s point of view. s for audio and video data also have to be robust, in the sense that the watermarks can survive after undergoing various common attacks, such as cropping and filtering. It is necessary because fragments of video and music can be easily extracted for re-use without proper credit or in an unauthorized way. s for rare documents and manuscripts, referring to Truman Library web site has to be invisible, as these documents are already blur after digitization. Any additional noise can likely turn the documents to be unreadable. Patent documents in IBM /00 $10.00 (c) 2000 IEEE 6

7 Intellectual Protection Network are also digitized documents, further distortion on the text and diagrams is not tolerable, and the attached bar-code requires a clear background for detection. Thus it is recommended that invisible watermarks are more appropriate to this type of document. In deed, text data in electronic commerce applications are usually produced by scanners and facsimiles and its readability has already been downgraded. Further degradation on the quality of the document is not encouraged. Besides, Acken [1] suggested that an application used only to indicate modifications of the content needs only a fragile watermark. It is because fragile watermarks are easily corrupted by any form of signal processing procedures. Originality is the prime asset of important documents, such as rare and patent documents. Any changes or modifications must be noticeable to the viewers and owners. Thus, invisible and fragile watermarking is preferable to text data. For image data, watermarks can be visible and invisible depending on the purposes of the application and digital contents. For promotion and demonstration purposes, digital photograph may carry a visible watermark that can be a company trademark or a copyright mark, like Figure 4a, so that content distributors can easily identify their properties without performing any watermark extraction process. When perceptibility of digital contents is a concern, an invisible watermark is a better choice. All existing invisible image watermarking techniques and systems in the literature are robust, while all visible image watermark techniques and systems are fragile. Video Robust WDP video = {invisible} invisible = {Robust} Invisible Figure 7. A Design Pattern for Video data. Text WDP text = {invisible} invisible = {Fragile} Invisible Fragile Figure 8. A Design Pattern for Text data. Using the above results, we may conclude the characteristics of watermarks for the four different media data and their corresponding WDPs are shown in Figures 6, 7, 8 and 9. Visible Invisible Audio Fragile Robust WDP image = {visible, invisible} Robust WDP audio = {inaudible} inaudible = {Robust} Inaudible Figure 6. A Design Pattern for Audio data. visible = {Fragile} invisible = {Robust} Figure 9. A Design Pattern for data. These four WDPs are subsets of the general WDP, and they can be expressed as follows. WDP media data = {WDP image, WDP video, WDP audio, WDP text } It is concluded that our results are the following /00 $10.00 (c) 2000 IEEE 7

8 for audio data has to be inaudible and robust. for video data has to be invisible and robust. for text data has to be invisible and fragile. for image data can be either 1. visible and fragile, or 2. invisible and robust. 5. Relationship between Electronic Commerce Application and Digital ing Technique Using the watermark design patterns derived in Section 4, we introduce a selection scheme for electronic commerce application developers to select adequate watermarking techniques and systems for different media data in their applications. Zhao et al. [30] addressed that the watermarking technology s goals or functions can be classified into four application categories: copyright protection, hidden annotations; authentication; and secure and invisible communications. Each application category generally requires applicationspecific watermarking techniques. These categories are mapped to the electronic commerce domain. Incorporating the concept of WDP and the application categories, the basic relation between electronic commerce applications and digital watermarking techniques can therefore be described as Figure 10. In the diagram, there are four layers; category, digital contents, WDP and digital watermarking technique. At the category layer, four watermarking technology s functions of electronic commerce applications are defined. For instance, following the branch of copyright protection, which is the focus of this paper, the application is enforced by copyright protection scheme. Hence, the watermarking can provide ownership verification and identification. As discussed in the Section 4, different media data of digital contents require different watermarking techniques. The characteristics of these digital watermarks are described by their associated WDPs, namely WDP video, WDP image, WDP audio and WDP text for video, image, audio and text data, respectively. Each WDP leads to a set of existing watermarking techniques and off-the-shelf watermarking systems. The category of copyright protection in general can be formulated as follows. Cat copyright = {video copyright, image copyright, audio copyright, text copyright } where video copyright WDP video, image copyright WDP image, audio copyright WDP audio, and text copyright WDP text. And, therefore Cat copyright {WDP video, WDP image, WDP audio, WDP text } The above formulation can also be extendible to other categories. However, they are not covered in this paper because it is out of the scope of this paper. The relationship diagram shown in Figure 10 is important for electronic commerce application developers to protect intellectual property of their distributed media, as it provides a look-up table for them to choose adequate watermarking techniques and systems for their electronic commerce applications. For example, to design a web-based video distributing system, which supplies video trailers in MOV format to its subscribers, and copyright protection is a great concern to the application. Referring to Figure 10, digital watermark for its digital contents must be invisible and robust, and watermarking techniques, WT1, WT2, and WT3 are desirable for this case. 6. Conclusions ing is undoubtedly important for protecting various forms of digital contents in the digital age. Electronic commerce applications require such protection to prevent the misuse of the material they mount for public consumption. However, only a few electronic commerce application developers apply efficient techniques to protect digital contents in their applications, mainly because they are unfamiliar to the technology. Our objective was to develop a scheme that can facilitate electronic commerce application developers to choose adequate digital watermarking techniques and off-the-shelf systems for their applications in an efficient way. In this paper, we proposed watermark design pattern (WDP) to describe the characteristic of a digital watermark for specific media data. A survey on nine general and typical media distribution web sites, which are exhibiting digital contents on the WWW, was conducted to investigate the relationship between watermark design patterns and media data, when copyright protection was a concern. Applying the survey results to the general WDP and using characteristics of existing watermarking techniques, four watermark /00 $10.00 (c) 2000 IEEE 8

9 design patterns WDP video, WDP audio, WDP image, and WDP text were derived for four different media data video, audio, image and text, respectively. We extended and applied our results and findings to present the relationship between electronic commerce applications and digital watermarking techniques. The relationship diagram fulfills our objectives by closing the gap between developers and digital watermarking for copyright protection. WDPs derived in this paper hold, when typical and general media distribution web sites are in use. In special or extreme cases, their WDPs may differ from those derived from the nine web sites. In order to strengthen the argument of our results, we will carry on an extensive survey on widespread electronic business web sites and applications. Future research will also focus on the issue that versatile watermarking techniques are capable of applying to more than one media data, and how they affect the structure of the relationship between electronic commerce applications and watermarking techniques. Acknowledgements This research is supported by Industrial Support Fund of Hong Kong. References [1] J.M. Acken, How ing Adds Value to Digital Content, Communications of ACM, Vol. 41, No. 7, pp , July [2] ARIS Technologies, [3] L. Boney, A.H. Tewfik, and K.N. Hamdy, Digital s for Audio Signals, Proceedings of MULITMEDIA 96, pp , [4] J. Brassil, S. Low, N. Maxemchuk, and L. O Gorman, Electronic Marking and Identification Techniques to Discourage Document Copying, IEEE J. Select. Areas Commun., Vol. 13, pp , Oct [5] G. Braudaway, Results of attacks on a claimed robust digital image watermark, In Proceedings of the IS&T/SPIE Symposium on Optical Security and Counterfeit Deterrence Techniques, SPIE, pp , [6] G. Braudaway, K. Magerlein, and F. Mintzer, Color Correct Digital ing of s, U.S. Patent , June 25, [7] I.J. Cox, J. Kilian, T. Leightont, and T. Shamoon, Secure Spread Spectrum ing for s, Audio, and Video, IEEE Transactions on Processing, Vol. 6, No. 2, pp , 1997; see also Proc. ICIP 96, vol. III, pp [8] Digimarc, [9] H. Gladney, F. Mintzer, and F. Schiattarella, Safeguarding digital library contents and users: Digital images of treasured antiquities, DLIB Mag., 1997; see ml. [10] D. Goodin, A Market Waiting To Happen, IP Magazine (September 1997) Available at [11] D.T. Hawkins, Digital ing: Intellectual Property Protection for the Internet?, Online, pp , July/August [12] S. Low, N. Maxemchuk, J. Brassil, and L. O Gorman, Document Marking and Identification Using Both Line and Word Shifting, In Proc. Infocom 95, (Available WWW: ts/ecom.html.) [13] N. Maxemchuk, Electronic Document Distribution, AT&T Tech. J., pp , Sept [14] N. Memon and P.W. Wong, Protecting Digital Media Content, Communications of ACM, Vol. 41, No. 7, pp , July [15] F. Mintzer, G. Braudaway, and M. Yeung, Effective and ineffective image watermarks, In Proceedings of the IEEE ICIP 97, pp. 9-12, [16] B. Schneier, Cryptography: The Importance of Not Being Different, IEEE Computer, Vol. 32, No. 3, pp , March [17] A. Shamir, Method and Apparatus for Protecting Visual Information With Printed Cryptographic s, U.S. Patent , Jan [18] Signafy, [19] M. Swanson, M. Kobayashi, and A. Tewfik, Multimedia Data-Embedding and ing Technologies, IEEE Proceedings 86, 6, pp , June [20] M. Swanson, B. Zhu, and A.H. Tewfik, Data Hiding for Video in Video, Proceedings of the IEEE International Conference on Processing 1997, Vol. II, pp , [21] M. Swanson, B. Zhu, and A.H. Tewfik, Multiresolution Video ing using Perceptual Models and Scene Segmentation, Proceedings of the IEEE International Conference on Processing 1997, Vol. II, pp , /00 $10.00 (c) 2000 IEEE 9

10 [22] M. Swanson, B. Zhu, A.H. Tewfik, Object-based Transparent Video ing, Proceedings 1997 IEEE Multimedia Signal Processing Workshop, pp , [23] M. Swanson, B. Zhu, A.H. Tewfik, and L. Boney, Robust Audio ing using Perceptual Masking, Signal Processing, (66), pp , [24] SysCoP, ml. [25] C. Trowbridge, Protection for Archives, Special Collection Libraries and Museums in the WWW Environment, [Online]. Available WWW: Fall95papers/trowbridge.html. [26] P. Vogel, System For Altering Elements of a Text File to Mark Documents, U.S. Patent , Feb. 7, [27] M. Yeung, Digital ing, Communications of ACM, Vol. 41, No. 7, pp , July [28] J. Zhao and Fraunhofer Inst. For Computer Graphics (1996). [Online]. Available WWW: [29] J. Zhao, and E. Koch, Embedding Robust Labels into s for Copyright Protection, Proceedings of the KnowRight 95 conference, pp , [30] J. Zhao, E. Koch and C. Luo, Digital ing In Business Today and Tomorrow, Communications of ACM, Vol. 41, No. 7, pp , July Media Data Perceptible Imperceptible Robust Fragile Robust Fragile WDP media data = {perceptible, imperceptible} perceptible = imperceptible = {Robust, Fragile} Figure 5. A general Design Pattern for Media Data. Electronic Commerce Applications Category Hidden annotations Copyright protection Authentication Secure and invisible communications Media Data Video Audio Text Design Pattern (WDP) WDP video WDP image WDP audio WDP text Digital ing Technique WT1 WT3 WT2 where WT1, WT2, and WT3 are three different ing techniques or systems Figure 10. Relationship between electronic commerce applications and digital watermarking techniques /00 $10.00 (c) 2000 IEEE 10