ISO/TC 46/SC 9/Working Group 1for ISO Project 15706: International Standard Audiovisual Number (ISAN)

Web page: http://www.nlc-bnc.ca/iso/tc46sc9/isan.htm

ISO International Organization for Standardization ISO Organisation internationale de normalisationTC 46 Information and Documentation TC 46 Information et documentationSC 9 Presentation, Identification and Description of Documents SC 9 Présentation, identification et description des documentsWG 1 International Standard Audiovisual Number (ISAN) WG 1 Numéro international normalisé des oeuvres audiovisuelles (ISAN)

ISO/TC 46/SC 9/WG 1 N 941999-06-08

This document is an interesting presentation of labeling and watermarking which are obviousways to use and attach ISAN to works. The second part of the document is a more questionablescheme for commercial transactions between business partners on line: if such a practice isalready becoming common for still images, the prospects of commercial exchanges betweenprofessionals (including distributors and broadcasters) on line may be science fiction for quitesome years.

One can notice in this document prepared by very competent technical people present ideaswhich can only shock legal people: references to "copyright registration" and, although the worditself is not used, "copyright notices". It is however and interesting indication of whattechnicians and telecoms want to drag us into: those "people of the future" are trying to take usmore than a century back and re-invent copyright notices...

If we leave aside those aspects, this document can be of valuable interest to the members of WG1. It is of particular interest if put in relation with document N95 as far as it comes from peopleinvolved in the Octalis project. It can show also some of the concerns of Broadcasters.

How to secure IPR over existing open networksOn the feasibility to protect IPR in existing open networks

J.-M. BoucqueauD, J.-F. DelaigleD, E. Goray*DLaboratoire de Télécommunication et

TélédétectionUniversité catholique de Louvain

Bâtiment Stévin - Place du Levant, 2B-1348 Louvain-la-Neuve

Tel: +32 10 47 41 05 - fax: +32 10 47 20 89Emails: {Boucqueau, Delaigle}@tele.ucl.ac.be

*Radio Télévision Belge de la communautéFrançaise

Boulevard Reyers 52Local 8M60

B-1044 BruxellesTel: +32 2 7372989 – fax: +32 2 7374747

Email: eddy.goray@euronet.be

Abstract. This paper presents a new approach, based on existing techniques and specified supporting services, toefficiently protect IPR in real businesses over open networks. A system to control the access to the material and totrack the material once accessed is proposed. This system has been designed in a bottom-to-up approach reflectedin the paper structure. At first the system has been implemented and tested, during the football WorldCup inFrance amongst other. Afterwards, a mapping with existing business models has been achieved. Some lacks and

ISO/TC 46/SC 9/WG 1 N 941999-06-08

page 2

open issues in such models, in the standardisation work and in the technologies are highlighted. A formalisation ofthe IPR management and presentation is proposed.

1. Introduction

The current rapid development and deployment of new IT technologies for the fast provision of commercialmultimedia services has resulted in a strong demand for reliable and secure IPR protection techniques formultimedia data. Laboratories all over the world have been working on the development of such techniques. Inparallel, multidisciplinary experts groups have been working on the specifications of new business modelssupporting IPR protection all over the multimedia services distribution chain. Standardisation bodies have alsobeen working on the subject for a while. This paper presents a new approach, based on existing techniques andspecified supporting services, to efficiently protect IPR in real businesses over open networks. A representativeEuropean consortium, gathered within the OCTALIS1 project[1], has specified, implemented, tested and validatedthis novel approach during these last two years thanks to the ACTS2[2] funding.The second section skims through the existing means for IPR protection. The third one skims through theemerging ones. It concludes with the prerequisites for the understanding of the approach. The third one presentsthe two concrete business scenarios considered by the Octalis projects. Actors concerned in these scenarios wereinvolved in the project. The solutions, also introduced in this section, have been conceived with their support andcritics. The next section generalises the solutions in one common scheme and point out the few lacks in theexisting means. An original formalisation is proposed and defended. The arguments for such a model arecompleted by a mapping with a hugely accepted business model. Finally, conclusions stress the innovations,underline the lacks in the techniques and list open issues

2. Existing Protection Means

Copyright and author’s rights are secured automatically as soon as the work is created, and a work is “created”when it is fixed in material objects. This fact is common to major copyright and author’s right laws. There existnational and international legislation that aim at protecting intellectual property (US Copyright Act, BernConvention, WIPO treaties etc.). Nevertheless, additional means are needed in order to facilitate the application ofthe laws, such as technical protection tools.

2.1.Copyright/IPR Registration and Legal Deposit

In general, copyright registration is a legal formality intended to make a public record of the basic facts of aparticular copyright. Even though registration is not a requirement for protection, the copyright law providesseveral inducements or advantages to encourage copyright owners to make registration. The US Copyright Act alsopromotes the legal Deposit of copies of a work. The same concepts are currently promoted internationally by ISOthrough the notion of Registration Authorities. A registration and a fortiori a deposit can be advantageously used asevidence in case of copyright disputes. In European countries where moral rights are important, the deposit canalso be used to prove that the integrity of a work has not been respected. In order to be as general as possible and torespect the terminology of both legislation, we will use the term Intellectual Property Rights (IPR). Thisterminology is becoming more and more widely used so as to avoid the distinction between copyright and author’srights, even if IPR also include other topics such as patents and trade marks.

1 Offer of Content through Trusted Access LInkS2 Advanced Communication Technologies and Services

ISO/TC 46/SC 9/WG 1 N 941999-06-08

page 3

2.2. Identification

Persistent identification can be defined as the ability to manage the association of identifiers with digital content.This will achieve the critical link between the one or more component creations that may exist within a piece ofdigital content and the environment which stores the related descriptive data, current rights holders, licenseconditions and enforcement mechanisms. The importance of standardizing internationally identification numbersis crucial for protecting and managing intellectual property.Copyright owners and Rights management Organization are already managing materials subject to intellectualproperty rights by means of existing international standard numbering schemes, such as:• ISBN (International Standard Book Number) ISO 2108 – International Book Agency in Berlin• ISSN (International Standard Serial Number) ISO 3297 – International Serial Agency in Paris (for periodical

publications)• ISRC (International Standard Recording Code) ISO 3901 – International Federation of the Phonographic

Industry in London• ISMN (International Standard Music Number) ISO 10957 – International Published Music Agency in Berlin• ISAN (International Standard Audiovisual Number) project number 15706 – ISAN Agency in Geneva• ISWC- (T)/(L)/(S)/(V) (International Standard Work Code for tune/literary/scientific/visual) project 15707 –

Confederation Internationale des Sociétés d’Auteurs et Compositeurs in Paris.• IMLP (International Multimedia License Plate)

The necessity to support identification schemes has been recognized by major standardization bodies in charge ofcoding representation of digital content. Rooms have been defined in their specifications in order to includeidentifiers either in the bitstream (for transport) or in the file (for storage).In the domain of still images, FlashPix and SPIFF3 have dedicated tags and fields in the headers of their fileformats to identification purposes. Facilities to add information about intellectual property have also been granted.In the domain of moving images, MPEG24 has also specified a Copyright Id. and a Copyright Number to carryidentification of the registration authority and the identifier of the content. MPEG4, the new coming standard formultimedia content, will include in its specifications the possibility to attach Intellectual Property Information Datato each object defined by the standard, such as video, stills, audio, synthetic content, objects inside a video or a stillimage etc. There will be reserved fields dedicated to identifiers.

2.3.Conditional Access Systems

Access control is the denial of access to unauthorized users. A conditional access system (CAS)[3] aims atmanaging access control for a specified set of users. This set of users is set-up and maintained through aregistration process (e.g. when you buy a decoder, a SIM card or an X.509 certificate). In most cases, those usersare identified.In general, well designed CAS offer the following functionality:• Creation access protection: obviously, while the creation access is unauthorized, copyrights can not be

violated.• Recipient a-priori identification and authentication: recipients are potential sources for copyright

infringements. This list of « suspects » might be useful if the accessed creations were not all exactly the same.

3 SPIFF : Still Picture Interchange File Format Annex F of ITU-T Recommendation T84 | ISO/IEC IS 10918-34 MPEG : ISO/IEC JTC1/SC29/WG11 coding of motion picture and audio

ISO/TC 46/SC 9/WG 1 N 941999-06-08

page 4

• Data secure transfer: the denial of access is valid for any kind of data. CAS system can be used for sensitivedata secure transfer, e.g. watermarking payload. CAS relies on cryptography. It can thus be customized andextended to various operations requiring cryptographic functions like signature, encryption, etc.

This makes CAS powerful but incomplete tools for IPR protection. Through its use, the access to creations can becontrolled. Rights to its use can be a-priori negotiated and legally formalized. But when it has been accessed, thereis absolutely no more technical means to control its use. Other technologies to track the accessed creations arenecessary.

2.4.Copy Control and Copy Management Systems

Copy Control consists in the ability to prevent copies from being made. Copy Management Systems is an extensionthat can allow a given number of copies.Serial Copy Management System (SCMS) was one of the first means to achieve Copy Control. It has been appliedto DAT and also embedded into some recordable CD players. This kind of protection is very simple. A counter isincremented each time a material is copied. This counter is only a few non-protected bits. In some systems, theidentification of the copying device cans also be associated with the copied content. In the simplest systems, suchas DAT), a bit is simply put to 1 when a copy has been done, so that the player knows that it is trying to read acopy. Of course, the efficiency of Copy Control mechanisms based such systems is very limited. They are not reallysecure since trained users can easily modify copy bits.In DVD5, an evolution of SCMS, called Copy Generation Management Systems (CGMS), will be adopted in thecoming specifications. CGMS allows no copies, single copy or multiple copies to be made from digital content.The main drawback of this kind of application is that is application oriented. For instance, CGMS is specificallydeveloped for DVD, but it does not apply to other storage media nor to other means to distribute the content.CGMS specifications are not finalized yet in DVD, but there is great chance they will be associated with encryptionsystems (Digital Transmission Content protection) or watermark based systems to enhance the security.

3. Emerging Protection Means

Improving management and protection of digital content requires the ability to associate Intellectual PropertyRights or contractual information (let us call them IPR information) with this content. This can be achieved byattaching either this IPR information to the content or only data that refer to it. In this latter case, only part of theinformation or simply a pointer is attached to the content, the whole IPR information being stored in remotedatabases. There exist some initiatives to promote and standardize the development of such databases, but thisissue is out of the scope of this paper6 [Error! Reference source not found.]. The referring pointer can be public(see the work of DOI [Error! Reference source not found.]), but it can also be private and reserved to entitiesresponsible for management and protection. In the same manner, IPR information stored in the databases caneither be public or private. These choices clearly depend on the Business Model (BM). This paper presents a BMthat tries to make clever combinations of public, private, attached and remote IPR information.Basically, there are two ways to technically associate IPR information with the content, we call them labeling andwatermarking. The major distinction between them is that watermarking modifies the content itself at the oppositeof labeling. These technologies have different functionality and offer different levels of protection.

5 DVD: Digital Versatile Disk is the next generation of optical disc storage technology. DVD has widespreadsupport from all major electronics companies, all major computer hardware companies, and about half of the majormovie and music studios.6 CIS : Common Information System ( Cf. Error! Reference source not found. )

ISO/TC 46/SC 9/WG 1 N 941999-06-08

page 5

3.1.Labeling

3.1.1. Definition

In this paper, we call label additional IPR protection data that are stored and carried along with the content inorder to enable its protection and management, without modifying this content and labeling the fact to attach oneor more label to the content. In certain circumstances, the label can be stored in a remote database in a duplicatedor extended version.

3.1.2. Features3.1.2.1. Techniques used

Among them, digital signatures are good candidates. Digital signatures are common techniques in cryptography.They allow the verification of the origin and of the integrity of the content if they are combined with validcertificates and corresponding cryptographic keys. It can also be interesting to include other generic IPRinformation data, such as simplified terms of contract, or identifiers etc.The functionality offered by labeling depends on the use of signatures. They can be applied on the content itself orpart of it. In this case, it is possible to verify the strict integrity of the bitstream after transmission. The majorproblem is that these signatures are useful to protect the bitstream, but they can not protect a work from itscreation. As a matter of fact, signatures allow detecting the modification of one single bit of the data they areapplied to. During its existence, a work will be submitted to several levels of editing processes, different standardsof coding, diverse compression ratios. Signatures would not be valid anymore after such processing. It is thereforenot possible to make the distinction between a forgery and an authorized processing.Another option is to apply signatures on features extracted from the content. These features can be objectscontours, textures or spectral characteristics of the content but must represent a certain semantic of the content.Features extraction (FE) is an area that is still under investigation [S. Bhattacharjee, M. Kutter, CompressionTolerant Image Authentication, in Proceedings of the IEEE ICIP , the IEEE ICIP 98, Volume 1 of 3, pp 435-439,Chicago, October 98], but there are interesting primary results that could yield to robust systems. When signaturesare applied after FE, it is possible to detect a forgery from a simple processing. This issue is essential in Europeancountries where moral rights have a great importance.Finally, signatures could be applied on identifiers, Cf. 2.2. Identification or to any kind of more generic IPRdata in order to guarantee their integrity.

3.1.2.2. The structure and the content of the label

It depends on the BM. The first data that must be included in labels are the public cryptographic key that permitsthe verification of the signatures, accompanied with a certificate and the respective signatures. To do this we haveto add data that are signed apart from the content, such as IPR data, features and possibly identifiers even if theyare already located in a dedicated location. Finally, as mentioned above, the label can be duplicated in a database,but we can also devise a system where only some parts of the labels are carried along with the content and thecomplete labels are stored remotely. In practice, extracted features are too large to be transmitted. So, the part ofthe label concerning these features would be stored in a database while the other parts, referring to the bitstreamintegrity and the integrity of identifiers would be transmitted.

3.1.2.3. The location of the label

It is a difficult issue, especially during the transport of the content, that is, when the content is streamed. This issueis easier in the case of storage, file format specifications allow for additional user data, also referred to as metadata,which can contain labels, e.g. in headers. If not, it is easy to associate a label with a content by an appropriate filemanagement in the database. Similarly, interface and transport streams specifications allow conveying data in

ISO/TC 46/SC 9/WG 1 N 941999-06-08

page 6

addition to the content. However, in practice, equipment manufacturer and application designers do not alwaysimplement this facility. For instance, today, most of broadcasters professional equipment skips these data, becausethey are optional in the standards and specifications are never implemented entirely. The same remark applies forapplications designed for consumers’ devices. Fortunately, there exist initiatives that intend to promote the use ofthese data during transport and storage, such as the EBU/SMPTE task force7. Nevertheless, it could take a longtime and lots of efforts in standardization groups before the use of labels is widely accepted and supported.

3.1.2.4. The functionality of the label

They are of course limited by the fact that it can easily be removed. It is indeed impossible to prevent the removalor replacement of labels, since they are separate from the content. The only way to enforce usage of labels would belegal actions at an international level, but this is difficulty conceivable. Yet, labeling presents some interests. Theroom allocated to labels data are not too much limited in space. Actually, they can be limited by the bitrate whenlabels are transmitted along with content, but this limit generally leaves room enough for most IPR protection data.In conclusion, the functionality that can be offered by labeling are the following:• Authentication of the origin of the content• Strict integrity of the bitstream• Integrity of identification numbers and IPR data• Integrity of the meaning of the content

bitstream Originalcontent

Hashing Featuresextraction

IdentifiersIPR data.

Hashing

multiplesignatures

+ certification

Labeling

OR

Figure 1: Labels computation

Public Key Certificate Signatures IPR data/ identifiers Extracted FeaturesNot always present (optional)

Figure 2: Content of the label

7 EBU/SMPTE task force: ajouter reference

ISO/TC 46/SC 9/WG 1 N 941999-06-08

page 7

3.2.Watermarking

3.2.1. Definition

In this paper, we call label watermark IPR protection data that are embedded directly into the content. That is,watermarking implies a modification of the content itself. The signal is changed, and in digital, the bitsrepresenting the content are partly or integrally changed. Nonetheless, the resulting degradations of the qualitymust be minimized under a required level to keep this content valuable. In the following, we will focus on images(still and moving) but the same concepts exist for audio [M. D. Swanson, B. Zhu, A.H. Tewfik and L. Boney,Robust Audio Watermarking Using Perceptual Masking, Special Issue on Watermarking, Signal Processing,volume 66, N°3, pp 337-356, Elsevier, May 98] and other kind of multimedia content [R. Ohbuchi, H. Masuda andM. Aono, Watermarking 3D Polygonal Models Through Geometric and Topological Modifications, IEEE Journalof Selected Areas in Communication, 16(4) , pp 551-560, May 98].

3.2.2. Visible watermarks

Visible watermarks [G. W. Braudaway, K.A. Magerlein and F. Mintzer, Protecting Publicly Available Images witha Visible Image Watermark, In Proceedings of the SPIE/IS&T on Optical Security and Counterfeit DeterrenceTechniques, Electronic Imaging, volume 3016, pp 126-133, February 97] are an extension to the concept of logos.These logos are inlayed into the image, but they are transparent. The major difference with classical logos is thatthey recover a great part of the image, so that it is not possible to remove visible watermarks by simply croppingthe center part of the picture. Moreover, visible watermarks are ideally protected against attacks such as statisticalanalysis of the picture. The first drawbacks of visible watermarks is that first they degrade the quality of theoriginal content, having a logo in the center of the image can be quite disturbing. The other major drawback is thatvisible watermarks are can only be detected visually, which means watermarks can never be detected automaticallyby dedicated programs or devices. Visible watermarks are used, however, for very specific contents, e.g. maps,graphics, software user interfaces.

3.2.3. Invisible watermarks

On the opposite, invisible watermarks do not degrade the content, when correctly designed. The objective ofinvisible watermarks being to hide data into the content in a transparent way, the development of thesetechnologies must be done carefully. Invisible watermarks are submitted to conflicting constraints. On one side thequality of the content must not loose commercial value after watermarks embedding but on the other side, theembedded data rate must be as high as possible or when the data rate is fixed, the robustness of embedded datamust be as high as possible. This is contradicting since the invisibility constraint limits the capacity of the channelavailable for these data. A good approach is to estimate the capacity of these perceptual channels with help ofHuman Visual Model [C. Podilchuk and W. zeng, Image-Adaptive Watermarks Using Visual Models, IEEEJournal of Slected Areas in Communication, 16(4) , pp 525-540, May 98,J.F. Delaigle, C. DE Vleeschouwer andB. Macq, Watermarking Algorithm Based on a Human Visual Model, Special Issue on Watermarking, SignalProcessing, volume 66, N°3, pp 319-336, Elsevier, May 98]. Finally, the development of watermark technologies isalso submitted to cost and efficiency constraints (real time in the case of video).

ISO/TC 46/SC 9/WG 1 N 941999-06-08

page 8

There exist many different methods to hide data into a picture8, they have their respective performances. Theprinciple is to modify some of its characteristics, such as pixels luminance, DCT coefficients [Cox, J. Kilian, F.Leighton and T. Shamoon, Secure Spread Spectrum Watermarking for Multimedia, IEEE Transaction on ImageProcessing, volume 6, n°12, pp 1673-1687, 1997,F. Hartung and B. Girod, Watermarking of Uncompressed andCompressed Video, Special Issue on Watermarking, Signal Processing, volume 66, N°3, pp 283-302, Elsevier,May 98], wavelet coefficients [C. Podilchuk and W. zeng, Image-Adaptive Watermarks Using Visual Models, IEEEJournal of Slected Areas in Communication, 16(4) , pp 525-540, May 98,M.D. Swanson, B. Zhu and A.H. Tewfik,Multiresolution Scene-Based Video Watermarking Using Perceptual Models, IEEE Journal of Slected Areas inCommunication, 16(4) , pp 540-550, May 98], fractals or motion vectors in videos.

3.2.3.1. Public watermarks

We call public watermarks, watermarks that can potentially be read or retrieved by anyone, with the onlyknowledge of the algorithms used. The first commercial products available on the market were actually publicwatermarks [Error! Reference source not found.,Error! Reference source not found.]. The security of this kindof watermarks simply relied on the obscurity of their technologies. Thus, at the moment, one user acquiressufficient knowledge on the algorithm, the whole system is potentially insecure. Public watermarks are not securein the sense of the Kirkhoff law9. Public watermarks still present some interest when used adequately, that is, notfor securing the content but simply to carry IPR information and facilitate copyright clearance. Public watermarksare good alternatives to labels. They allow to transport data along with the content without being skipped byequipment or interfaces. Their major drawback is the limited space for IPR data because of the invisibilityconstraint.

3.2.3.2. Tamper-proofing watermarks

We call tamper-proofing watermarks [D. Kundur, D. Hatzinakos, Towards a Telltalt Watermarking Technique fortamper-Proofing, in Proceedings of the IEEE ICIP 98, Volume 2 of 3, pp 409-413 ,Chicago, October 98], or alsofragile watermarks, watermarks that allow the verification of the integrity of the content. In a way, they are alsopublic, because any user can verify the verification of the integrity. We distinguish them from public watermarksbecause they have different functionality. We have striven to respect the terminology of the literature in this paper.In a few words, fragile watermarks allow to detect significant modifications in a picture. The integrity check basedon fragile watermarks is able to make the distinction between artifacts introduced by a compression andmodification resulting from a forgery. The major drawback of this technology is the fact that it can be difficult todistinguish a modified content from a content that has not been watermarked. Fragile watermarks are still underinvestigation in the literature.

3.2.3.3. Private watermarks3.2.3.3.1.Definition

We call private watermarks, but we could also use the term secure, watermarks that are secured by a secret key.The IPR data embedded into the picture are only accessible with the use of this secret key. Moreover, it must bevery difficult to remove or alter private watermarks without degrading the content. In this paper, we will mainlyfocus on private watermarks.

3.2.3.3.2.Features

In addition to the above-mentioned constraints, watermark must also be very robust. Above all, watermarks mustresist the treatments imposed to the content during its existence, such as editing processes (e.g. re-scaling, contrast

8 ajouter reference au papier Kutter et Hartung9 ajouter reference crypto

ISO/TC 46/SC 9/WG 1 N 941999-06-08

page 9

enhancement etc.). This point is crucial, since Copyright and author’s rights apply during a long period of time.The content will be coded, re-encoded and distributed through several kinds of networks, on stored on variouskinds of media. Anyway, it must be possible to retrieve the IPR data watermarked in the content after itsredistribution.Some malevolent users, sometimes, try to infringe copyright, we call them pirates. Most of the time, they do notcare about the presence of a watermark. In this simple case, watermarks allow to track that kind of users. However,when they know the content is watermarked they will probably try to break it. There are many ways to breakwatermarks [Fabien Peticolas, Ross J. Anderson and Markus G. Kuhn, Attacks on Copyright Marking Systems, inInformation Hiding : second international workshop, Lecture notes in Computer Science. Springer Verlag, Berlin,Germany, 1998]. It is almost impossible to remove or to replace a watermark, because this requires the secret key.It is more clever to make it inefficient by modifying the watermarked content. Most of these attacks are verydifficult to deal with and sometimes not realistic. Fortunately, most of these attacks are effective when they degradethe content, e.g. low pass filtering, cropping, noise addition. Besides, the pirate has a great disadvantage, due to thefact that he does not posses the secret key. He can never be sure that the watermark can not be retrieved anymore.As a consequence, even if it is not possible to resist all the attacks, watermark application designers just have tomake sure that costs and risks taken to break the watermark are high enough to induce users to clear and respectcopyrights.

3.2.3.3.3.The content of the watermark

This really depends on the BM into which it is integrated. Copyright notices and images are obviouslyinappropriate. The use of identification scheme, being international standards or proprietary, is clearly moreefficient if it tends to maximize the entropy of identifiers and if they are unique. These identifiers must identifyuniquely a copyrighted work, a copyright holder or a consumer device, depending on when the watermark isapplied. In practice, their length is situated in a range from 64 to 160 bits. There is no need to go further. In someapplications, this length limits to a few bits (8 bits in DVD).

3.2.3.3.4.Functionality

There is confusion in the literature terminology between the technology and the applications of the watermarks. Atthe origin, watermarks were only used to identify the ownership of the content. This is why watermark applicationsgenerally refer to watermarks embedded during or just after the creation. When watermarks technologies wereapplied to identify the consumer device or the consumer himself, the term fingerprinting emerged, even if thetechnologies are actually the same.Besides, it is important to mention the retrieval of watermarks and fingerprints, also called monitoring or tracing.Monitoring the content consists in analyzing watermarks and trying to detect copyright violations. The efficiencyof monitoring is greatly increased when it is done automatically by dedicated hardware or software. The entityresponsible for monitoring, also called monitors, must be in possession of some information. The secret key isrequired to read the watermarks. This implies that monitors are only able to analyze watermarks on behalf of agiven number of copyright holders he has contracts with. Monitors record the result of the analysis and then thedetection of copyright violations can start.Finally, there is a still room for doubt about the use of watermarks as evidence in front of a judge. It is conceivable,but there is little chance that breakable technologies will serve to convict someone. Moreover, the retrieval ofwatermarks is never one hundred percent sure. Watermarks are nothing else but signals, they are detected whenthey are above a given threshold, and there is thus a certain degree of uncertainty in the retrieval process.In summary, when associated with appropriate monitoring, watermarks can offer these functionality:• Identification of ownership of a work• Tracing of the distribution of the content• Identification of a copying device (fingerprints)• Support for Copy Control Systems (DVD)

Next to this, we can add the possibility to serve as evidence before a court.

ISO/TC 46/SC 9/WG 1 N 941999-06-08

page 10

4. Business scenario

For several years, researches towards efficient means for IPR protection of digital creations have been conducted allover the world. These means have to be adapted to the business scenario. For example, DVD specified a system forDVD content protection, when accessed by the end-consumer.The following picture presents the different status of a digital material.

Figure 3: Material status, from the creator to the product

The creations and works correspond to the creation and shaping phases. During these stages, the material is underproduction. Once it is done, the work enters the distribution step. Distributors receive an instantiation of thework, called object. A last packaging, in respect with IPR, transforms it into a product, which will be distributed toend-consumers. This linear process is quite far from the real life for most of the creations. If you consider a CD-ROM or a TV program, the production is usually followed by a first distribution. It is then completed by newcreations before a second distribution, followed by a post-production etc. This issue is tackled in the next section.The common feature for all businesses is that these means have to make realistic the following sentence:

“I sell remotely my creation to you in a secure way.”

This expectation implies two steps. The first one is the work protection. This must be achieved once for eachcreation. It corresponds to the set-up of the preconditions for the proof of the ownership. The creator or thedesignated Service Producer deposits a unique description of the original (e.g. a hash-value, or a textualdescription) to the registration authority. This authority will attribute a unique identification number to the image,archive the two and send the number back to the owner.The owner will have to:• Securely and secretly merge something related to this identification number and the creation itself. In case of

ownership conflict, the actual beneficiary, in front of a court in the worst case, will prove the creation origin.He will be the only one able to retrieve the information from the creation.

• Securely and publicly attach something related to this identification number and the creation itself. By thisway, creation’s users can not deny being aware that this work is copyright protected.

ISO/TC 46/SC 9/WG 1 N 941999-06-08

page 11

The second phase, the secure exchange, can be, from the concepts point-of-view, repeated indefinitely. If a tradeoccurs, a contract must be signed. This contract may be implicit or not applied. Nevertheless, in general, suchsystem aims at providing the seller with the ability to sign a contract before sending the creation. Solutions to theissue we are dealing with have to respect some constraints. A major one is the decrease of the rights clearance cost.On-line creation trading with classical contract procedure would not necessarily be a relevant improvement.Therefore, and to go further into the selling process, a secure communication between the two actors is necessary.Later, exactly the same creation will probably be sold to another customer. Therefore, once the contract is signed,we still need to establish a link between this creation and this trading operation. This link has no public interest.The end-consumer does not care about the IPR details. If he wants to go through a more complex process (e.g. if hewants to re-use the creation for commercial purposes), he has to get in touch with the initial copyright owner.Thus, the seller has to merge a secure and secret information with the creation.Having in mind the previous sections, such a process may be realized by:• Applying a label when a secure and publicly accessible attachment to the creation is required,• Applying a watermark when a secure and secret merging of information with the image is required.10

Several business models related to this issue have been published these years [references]. The complexity of theprocess, the complexity of the creation concept and the numerous actors having their own interests and their ownpoints-of-view, led to a set of models sharing the core elements and diverging on the process details. We followedthe reversed approach. From real and specific scenarios, we built solutions, validated it and went towards a genericmodel. The rest of this section aims at presenting those scenarios.

4.1.Creations on-line trading over an open interactive network

Electronic commerce on the Web is now a reality. An Internet connection is enough to buy remotely anything youwant. For some material, this E-com take off is working properly. But for others like the audio files, it leads to animportant amount of piracy. Creations’ E-com incurs the same risks. Nevertheless, marketing studies [reference]and major E-com actors’ investments [21] demonstrate the urgent need for reliable solutions. That is the issue weare addressing in this business scenario. Used tools are those presented in the second and third sections, except theSCMS which is a dedicated system and the label (tested in the trials but the system security does not rely on)because there exists no secure and relevant label today. The solution specifically addresses still images trading overthe Internet, but the architecture and the protocols are applicable to any kind of data, assuming correspondingwatermarking algorithms are available (for sound, text, compressed data).

4.1.1. Common functional model

The figure below presents the corresponding functional model. Processes are also represented (number 1 to 6) andare explained in the rest of this section.

10In this section, we use equally merging and embedding in the context of watermarking

ISO/TC 46/SC 9/WG 1 N 941999-06-08

page 12

1 1 1

1

3

3

CO SPd

InfoDatabas

WatermarkingW1,

OriginalImages

2

U

InfDatabas

SPv

Image Bankofwatermarked

images

WatermarkingW2

ImageWW 2

1

5

2 5

6

6

4

RA CA/TTP1Registration

ImageW11

ImagesW1

Certificate

3

Figure 4: Creations secure trading over the Internet: Common Functional Model

Functions are classical and have already been defined [reference], and introduced in this section. Three functionclasses can be identified:• Management functions,• Production functions,• Diffusion functions.

Management functions correspond to the upper layer of the figure. The Certification Authorities (CA) and theTrusted Third Parties (TTP) are responsible for the actors’ certificates management (registration, re-validation andblacklisting). It corresponds to the first process. The Registration Authorities are responsible for the IPRregistration or the legal deposit. It corresponds to the third process. Other management functions exist, like theAuthor societies, but are not directly related to this paper main issue.Production functions correspond to the path towards the final, consumable, product. On one side, the creator (COfor Copyright Owner) will generate original creations. On the other side, Service Producers (SPd) will give someshape to this creation. By the way, he can generate a composite work composed with shaped creations and originalones.Diffusion functions correspond to the path towards the end-consumer. When entering into this phase, creationsshould not be modified anymore. In the figure above, it corresponds to the Service provider (SPv) and the User (U).The last two classes are easily split from the first one, but are often interleaved. The next section will come back onthis point.

4.1.2. ProcessesFirst of all, an initiation stage is necessary. All actors are registered to a CA or TTP. They obtain certificatesincluding their status (through a set of rights), a validity period and a public key. Ideally, the corresponding privatekey is distributed on a smart card to protect its access. It corresponds to the first process.For a first implementation of this business scenario, we only considered a linear distribution, from the creator tillthe user.• Once an original creation is achieved (process number 2), a registration or a legal deposit is mandatory for IPR

protection. It can be done by the CO or by the SPd. The best is to do it as soon as possible.

ISO/TC 46/SC 9/WG 1 N 941999-06-08

page 13

• In return, the creator receives a unique identifier for its work (third process).• To definitively guarantee its ownership, a secure and secret binding of the creation with the unique identifier

must be realized. Applying a first watermark (called W1 at this stage) as soon as it is in a digital format doesthis. It corresponds to the fourth process.

• At this stage, the work may enter the distribution phase. Therefore, it is transmitted to a re-seller, also calledSPv (process number five). IPR management between those CO, SPd and SPv has to be done off-line. Usually,it consists in business contracts established prior to the transactions. The SPv will propose thumbnails of thecreations on a web site [reference Odiss].

• U browses the thumbnails freely. He can express his interest by clicking one of those. At this stage, securecommunications are necessary because we enter into a business process and thus in the sixth process. Bothparties are authenticated through an exchange of certificates and a form is proposed to the U. This form will bethe basis for a trading contract. Its content is depending on the SPv. A minimal set of information is theidentity, the localization and the purpose of the U. This form is encrypted and sent to the SPv. The SPv cangenerate a contract, sign it electronically and propose it to the U. If U accepts the terms of the contract, hesigns it electronically too and send it back to the SPv and wait for the creation delivery. The SPv will probablysell the creation to other users. In order to track the use of it, he must be able to differentiate those. Therefore,a watermark is applied before the delivery. This secret watermark (called W2 at this stage) establishes a linkbetween the creation and the trading contract.

4.1.3. ConclusionsWith this simple system combining conditional access, cryptographic tools and watermarking, creators are able:• To control the access to their object,• To track the use of their object once accessed.Combined with an efficient legal net [reference AT&T – Data hiding], such a system may efficiently protect IPR.Nevertheless, if we consider a less linear process, several watermarks may be necessary and should degrade theimage quality.Finally, we must stress that such a system is only working in a point-to-point communication configuration.

4.2. IPR protection in a distribution networkIPR protection is fundamental for creator and creators are fundamental to food the multimedia services distributionchain. Consequently, IPR protection means deployment is a major argument in a competition framework. In thissecond business scenario, we focused on the competition between network operators. Once large amount of data,such as audio-visual material for broadcasting is concerned, the Internet is not sufficient at all. For such business,telecom operators and network operators are offering a wide range of alternatives. Our interest was in providing anadded value, through an efficient IPR protection system to such networks. The solution we conceived has beenimplemented and validated over one of them, the Eurovision network [reference site web UER].The Eurovision network is a satellite network managed by the European Broadcasting Union (EBU) [reference]. Itaims at exchanging material for TV programs. There are 176 members all over the World. These members are theonly ones with permanently authorized access to the network.The EBU is also responsible for major events rights negotiations. For instance, rights for sport events arenegotiated and bought by EBU, under the name of its members. Numerous cases exist and have been studied, butare out of the scope of the paper. This last activity is slightly marginal, compared with alternative network. It doesnot influence the solution but accelerates its demand. During those negotiations, arguments like “We guarantee therespect of your IPR” would be a major advantage, but quite unrealistic today. The solution we proposed allowsthem to say “We deploy means to protect your IPR, and in case of piracy, we will be able to prove that our membersare not responsible, or to point out the responsible member”.

ISO/TC 46/SC 9/WG 1 N 941999-06-08

page 14

4.2.1. Exchanges over a distribution networkContribution networks for professionals have specific constraints and features. Most of the constraints are relatedto the high Quality of Service demand. Customers usually require:• High bit rate levels to avoid any degradation before the post-production. This post-production is usually

applied on the CCIR 656 format. The minimal acceptable compressed bit rate for such transport is 24 Mb/s.• Quality is not subject to any compromise. The exchanged material is dedicated to professional post-production.

In consequence, any signal manipulation, such as watermarking, degrading the material is unacceptable.• Fast and reliable management, with permanent availability. Today, everywhere in the world, live-events are

demanded and obtained.• Real-time without any delay. For live programmes like duplex, no delay at all in the image transmission is

acceptable.Most of these networks broadcast the material and use a CAS. If it is not the case, the architecture of the solutionshould be closer to the one proposed for the previous business scenario. Otherwise, typical features are:• The CAS• A co-ordination center (a kind of TTP for the customers) for the transmission usually managed by the network

operator. This center commits on the transmission to occur, manages the CAS, and maintains an informationserver (IS) dedicated to the customers. Such information is a transmission unique Id, the transmitter and thereceiver(s), the time slot and the channel specifications. This information server normally use an othernetwork because the requirements are not at all the same and to avoid shared material amongst different staffs(operations and co-ordinations are split). Typically, it uses TCP/IP over Internet or closed networks such asdedicated lines.

Such a network also has its own characteristics:• It is symmetric. Any customer may, at a certain time, behave as a SPv and later has a user.• Data related to the customers are available in a more precise way than for the secondary network (from the TV

Company to the end-users). A secondary network typical configuration is one SPv and several thousands of U.A contribution network is used by several hundred customers and most of the transmission involve a few ofthose.

The figure below presents a generic scheme for such a network.

IS

Distribution Network

Originator

Cr. Scramb.Customer S(Cr) CrDescramb .Customer

SPdδδt SPvδδt Uδδt

U3

U2

Un

Figure 5: Distribution network generic scheme

ISO/TC 46/SC 9/WG 1 N 941999-06-08

page 15

4.2.2. IPR protection over the Eurovision network

Having in mind the objective presented above, it is the typical case of watermarked material exchanges that requirea limited access through scrambling and a second watermark at the reception. We assume that the creatorregistered and watermarked with the obtained unique Id its creation before any delivery. In order to avoid piracyfrom outside the customers’ community, conditional access is necessary and enough. In order to be able to identifythe customer responsible for a piracy act, a second watermark must be applied. But this second watermark must bedifferent from one receptor to another and thus be applied at the reception. Some researches for solutions toslightly modify the content through the descrambling operations are conducted [reference chameleon si je retrouve]but not applicable today.

Distrib.network

Customer

Customer

OriginatorW1

W2

W2

Figure 6: IPR protection in a distributed network

The above figure summarizes the situation. Once the material enters the distribution network, the creation step andthus the W1 encrustation are done. The network operator manages the exchanges between customers. Allparameters regarding those exchanges are available on the IS.The following picture presents the system has applied to the network. OPP stands for Octalis Planing Procedure. Itis an IS back up introduced for trial convenience. The Octalis clients are PCs with Internet connections. It isinstalled in each reception site and hosts the software used for the equipment management (encoders, decoders,multiplexes, up-converters, etc.). It also hosts the watermarking hardware, cryptographic engines and it isconnected to a smart card reader. The IS, the OPP and the Octalis clients are permanently connected through aTCP/IP link.

ISO/TC 46/SC 9/WG 1 N 941999-06-08

page 16

IS

OPP

Distribution Network.

Octalis

client

Originator

Cr. Cr.+W1W1 Scramb.Customer S(Cr.+W1)

Octalis

client

Cr.+W1+W2W2Descramb .Customer

SPdδδt SPvδδt Uδδt

TTPδδtR.A. Certificate

U3

U2

Un

Figure 7: IPR protection over the Eurovision network

The OPP manages a database with all clients’ identity and public keys. Once a transmission is published on the IS,the OPP generates a (de)scrambling key, a watermarking key and one watermark payload per receiver. Thoseparameters are encrypted with the clients’ public keys and published. The Octalis clients connect periodically tothe OPP and retrieve their dedicated set of parameters. In the client, parameters are decrypted and sent to thecorresponding devices (watermarking hardware and (de)scrambling engine). Once the transmission occurs, thecreation and its first watermark are scrambled and watermarked with a W2 as soon as descrambled.The watermark payload is a set of 64 bits specifying a transmission unique Id, the network and the receiveridentity. All these parameters are archived in a secure database in the network operator building. In case ofconflict, the absence of W2 would prove that the illegal material has not been transmitted over this network. Ifthere is a W2, the network operator is able to identify the responsible customer.For a real implementation, parts of the Octalis client should be embedded within tamperproof chips.

5. Business scenarios generic solution

The previous section demonstrated that a combination of conditional access, cryptographic tools and watermarkingmight, with the support of a management layer, offer acceptable solutions to the IPR protection issue. As stated inthe introduction, we decided to address the IPR issue in a bottom-up approach. From the two cases we solved andfrom the abundant related literature, some generic guidelines can be identified.

5.1. Production/Diffusion splittingThis paper does not address the illegal private copies done by the end-user. This issue has to be solved throughvarious solutions depending on the kind of product to protect like DVD did. Creations work or objects IPRviolations may lead to large-scale piracy acts. Once a product has been produced with objects illegal copies, it canbe marketed for a while and generate large profits. The means deployed to hide the piracy will be proportional tothe potential profits. Furthermore, such actions are mainly reserved to professional and specialists in the field. Wealready stated that the material would go through four steps; the creation, the work, the object and the product. Thesolution proposed in this paper intends to protect IPR for the three first steps. This solution has been conceived,implemented and validated with the assumption of a linear process. But most of the products are composed withnew creations combined with objects and works. The path towards the product can be long, complex and iterative.Nevertheless, the protection steps are clearly stated:The creation has to be registered or deposed and must obtain a unique Id,The object may require an up-date of the IPR related to the initial creation. If there is an added valuecorresponding to the passing from one status to the other, an IPR update must be achieved. In this case, twoalternatives must be considered:

ISO/TC 46/SC 9/WG 1 N 941999-06-08

page 17

• The W1 is kept but associated to new basic facts owned by the creation shaper. Rights for the new objects areshared between the creator and the re-shaper (the service producer).

• The W1 is changed. A trusted authority or the creator himself must retrieve the watermark. The Object isbecoming a new creation with another unique Id. In this case, all the rights are given to the SPd.

Such a situation may be reproduced several times. The unique common point is the production-related activity.In the same way, object exchange between service providers may occur several times. As son as this object is notmodify, it falls into the diffusion-related activities. In this case, there is no direct interference with the IPR.In conclusion, we suggest starting with a distinction between production activities and diffusion activities.Production activities are dealing with creations and objects as re-shaped creations. For those steps, thewatermarking appliances have their own requirements and specificity. In the business scenarios, such applianceswere presented as W1. This designation corresponds to a generic concept of WP. WP is the class of watermarksrelated to the ownership protection. It points out to the origin and beneficiaries of the bound material.Diffusion activities are dealing with objects still requiring a packaging and products dedicated to the end-user. Forthose steps, the watermarking appliances also have their own requirements and specificity. In the businessscenarios, such appliances were presented as W2. This designation corresponds to the generic concept of WD. WD isthe class of watermarks related to the tracking of the material. It points out to the various recipients and their rightson the bound material.

5.2. Watermarking: Application and implicationToday, watermarking is becoming an well-accepted technique. Progress in the field are going fast [reference JFD???] and the robustness / invisibility compromise limits are always pushed back. Nevertheless, two limits will stayvalid for a while:• The watermark payload: The shorter the payload of a watermark is, the better the chances are to communicate

it reliably. Furthermore, the difficulty for embedding in real time the watermarking is proportional to thispayload.

• The number of watermarks: The schemes we presented may be extended to more complex cases with variousphases of production and diffusion. It implies the multiplication of watermarks. Today, tests we did [referenceD23] showed an acceptable quality with three watermarks. Material will not support more for a while.

As stated in section four, some initiatives are promoting solutions relying on a binding of a pointer to the imageand on remote database for related IPR data storage. Only those types of solutions are viable today. A distinctionshould be done between the watermark application and its implications.The watermarking application process is the binding, through a watermarking algorithm, of the material and thewatermark payload. An optional cryptographic key may be part of the process.The watermarking implications are relying on:• The relevance of the watermarking payload as a pointer: Obviously, the pointer must be unique. To build an

efficient monitoring system behind, this pointer should be easily understandable and interpretable. Forinstance, first digits of a unique Id like the license plate indicate the country of origin of the registrationauthority, which delivered this pointer. In case of use of a secret key, a first publicly readable pointer mustpoint out the way to obtain this key. It motivates the support to initiative promoting the standardized pointers.Even more, the optional wrapping of this pointer into a watermarking payload should be standardized.

• The trusted and thus legal validity of the database pointed by the payload: Information accessed through thepointer must be relevant. Those information will mainly be useful in case of conflict and thus in front of acourt. It implies their legal validity. Authorities responsible for the maintenance of the database should betrusted. Their way of management, the protocols with implied parties should be well thought (authentication,confidentiality and integrity of the transmitted information, signed confirmations), validated and accepted bythe legal net. The information storage and splitting within inter-connected database require the same care.Some complementary cares like the up-to-date aspect and the public availability for a part of the informationhave to be considered.

ISO/TC 46/SC 9/WG 1 N 941999-06-08

page 18

As a consequence, we suggest to distingue the watermarking application, noted w and the watermark with all itsimplications, noted W. w is one technique. It may even be used to generate special effects in an image. W is onelegal protection mean, it is the set of information describing the rights, IPR among others, for this material.

An example:Let’s consider a TV program; A talk around some animals’ films. In this case, the TV Company will have to playthe roles of Service provider, service producer and creator in parallel as explained here below.First of all, animals’ films (film) are created by film-director (f).

Films are watermarked ( )( filmwf

P),

and registered with information like basic facts, right holders, etc. ( )( filmIf

f).

The results is a binding between the film and:

)()(:)( filmfilmfilm IwWf

f

f

P

f

P+= (1.1)

The TV Company (t) buys the films to their service producer. This action must be registered by adding information

linking both actors and the film ( )( filmIf

t). Updates implications for the watermark of the film are:

)()()(:)( filmfilmfilmfilm IIwWf

t

f

f

f

P

f

P++= (1.2)

Between the animal’s films, talks (talk) are broadcast. The TV Company creates these talks. It is a creationrequiring a registration, a unique Id and a WP.

)()(:)( talktalktalk IwWt

t

t

P

t

P+= (1.3)

Finally, the sequences are merged. It forms a composite object, a programme (prog). The ideal solution would be toconsider it as a creation and to go once again through the different steps. It would imply a new watermark and fallsin contradiction with the watermark limits stated above. Therefore, we suggest adding a watermark implication(and thus a unique Id) only as:

)()()(:)( progfilmtalkprog IWWWt

t

f

P

t

P

t

P++= ∑∑ (1.4)

5.3.FingerprintingThe same scheme may be applied to the diffusion stages. The set of information related to the production combinedwith the same type of information about the diffusion (of access protected content) would form a powerful tool forIPR management and clearance.Let’s re-consider the example of the previous section. When the TV Company buys the film, a diffusion watermarkshould be used:

)()(:)(,,

filmfilmfilm IwWf

t

f

tD

f

tD+= (2.1)

In this case, this buying action has no impact on the production database content. It stays in the 1.1 status and 1.2does not exist. 1.3 stays as such, it is only related to a production activity. Once the programme is created, theregistration process starts but 1.4 is becoming:

)()()()(:)(,

progfilmfilmtalkprog IWWWWt

t

f

tD

f

P

t

P

t

P+++= ∑∑∑ (2.2)

This solution offers new advantages:• The trusted database system would be more distributed. Database could keep manageable sizes, be more

reliable and answer faster.• Distribution network operators may gain value-added for their services. If they are allowed to obtain a trusted

status to manage such database, they could take in charge the diffusion watermarks management and simplify

ISO/TC 46/SC 9/WG 1 N 941999-06-08

page 19

the work for their customers. This is mandatory in a broadcast environment such as the case presented insection 5.2.

6. Business model mapping

6.1.The Imprimatur business modelFunded under the European Commission’s DGIII ESPRIT program IMPRIMATUR for Intellectual MultimediaProperty Rights Model And Terminology For Universal Reference. As its name implies, IMPRIMATUR isconcerned with the protection of Intellectual Property Rights exchanged in electronic trading, through thedefinition and adoption through consensus of a common architecture and standards.The IMPRIMATUR project11 has conducted considerable research into the different types of business modelswhich are likely to predominate through rigorous examination of the requirements of rights holders, by the analysisof established Web sites and through the distillation of the results from its consensus building activities. It hasproduced its own conceptual business model which identifies and defines what it believes to be the essential roleswhich will be required in an electronic trading environment; and which will require electronic copyrightmanagement system functionality for the management and protection of copyright material.The business model defines the different roles, or entities, of the participants in a trading environment, and its keycomponents, such as an electronic copyright management system. Having established the roles, it is necessary todefine their attributes. In other words, attributes are characteristics and features that uniquely identify each role. Arelationship represents the logical association between the roles and helps to explain the operation of the enterprisethat the model represents. Different types of transactions between the roles are identified, such as the passing ofcreations, monetary value, and licenses.A model is a helpful mechanism for illustrating scenarios with multi-dimensional layers; for example, paralleltransactions between two roles to support the purchase and sale of goods; obtaining the permission of rightsholders; and the payment transactions which may relate to both of these. For each of the roles defined within themodel there can be many actors or candidates to play them. We will also see that it is possible for one actor to playmore than one role. Once developed, the business model will enable us to take a view of all the individuals andorganizations which play an active part in the trading and protection of intellectual property in a physicalenvironment; and to map these against the roles which exist in a ‘virtual’ environment.The following conceptual representation of the model illustrates the different roles and the division between thecontrolled ECMS environment of the Creation Provider and the less controlled world of the Media Distributor.

11 For further information refer to the IMPRIMATUR Web site: www.imprimatur.alcs.co.uk

ISO/TC 46/SC 9/WG 1 N 941999-06-08

page 20

Creator

UniqueNumberIssuer

Rightsholder

CreationProvider

MediaDistributor

Purchaser

IPRDatabase

MonitoringServiceProvider

CertificationAuthority

Creation

Description

Unique

Num

ber

Value

ConditionalLicence

ApplicationforLicence

AssignmentofRi

CreationCreationD

Identity

Creation

CurrentRightsholde

andPaymentdetail

Log

Value

Check MDAuthorisation

IPR Info

Check ID ofMD

Log

$

$ $ $ $

N

Legend:Imprint of Purchaser ID

Imprint of Media Distributor ID

Imprint of Unique Number

Role exchanges Value withBank

$

P

D

N

Value

IPR Info

IPR Info

Creation

IPR Info

Protected Unprotected

P

Figure 5: IMPRIMATUR BUSINESS MODEL

6.2.Business scenario solution mappingThrough this paper, various names for the actors have been used as synonyms. First step of this mapping must bedone at this level.The creator corresponds to our Copyright Owner or Originator. In any case, this actor is the creations’ generator.He is feeding the rest of the chain.The creation provider corresponds to our service provider, the Media Distributor to the service provider and thePurchaser to the User. Through this paper, we demonstrated that the choices of the names are not easy. Mostfunctions can be merged in one entity, at the same time or not. In the example presented in section 6.2, the TVCompany endorses the four roles.The Unique Number Issuer is the registration authority and the CA is, in our case, extended to the Trusted ThirdParty. This TTP’s main goal is to offer on-line services and to simplify the CA’s role.We did not tackle the Rights holder or the Monitoring service Provider because it was out of scope. The re-distribution of the royalties and the monitoring are other complex issues. It is part of the other managementfunctions introduced in section 5.1.1.In our case, the IPR database is part of the RA. Within the second business scenario, the network operator alsomanages a kind of IPR database related to the diffusion activities. As stated in section 6, these inter-connecteddatabases are not yet specified enough. Initiatives are running and will probably converge soon.Regarding the dynamic of the system, similarities are also obvious:• The creation provider registers the material and imprints the unique number (N): That’s equation 1.1. We

suggest proceeding this operation as soon as possible. Therefore, we add an optional link between the RA and

ISO/TC 46/SC 9/WG 1 N 941999-06-08

page 21

the creator. In the professional world for instance, the idea of a watermarking chip in the camera and of a-priori obtaining unique Id stored on a tamperproof device such as a smart-card is seriously considered.

• The imprint of Media-Distributor Id: It corresponds to equation 2.1. In our scheme, we supposed a contractualand trusted relation between the SPd and the SPv for the first business scenario. The Imprimatur model ismore generic and complete. Nevertheless, our implementations may easily support it.

• The Imprint of purchaser Id: It is equation 2.1. We went further by adding a deposit or a trusted managementof this operation record. Once again, the IPR databases specifications and inter-connections should beimproved.

• Relations to the CA are the same.

7. ConclusionsTechnologies enable an acceptable IPR protection solution today. The label will be a very interesting added value.Nevertheless, with an important support of the management functions, conditional access and watermarking maybe sufficient.

8. References1. Error! Reference source not found.2. Error! Reference source not found.3. S. Bhattacharjee, M. Kutter, Compression Tolerant Image Authentication, in Proceedings of the IEEE ICIP ,

the IEEE ICIP 98, Volume 1 of 3, pp 435-439 ,Chicago, October 984. M. D. Swanson, B. Zhu, A.H. Tewfik and L. Boney, Robust Audio Watermarking Using Perceptual Masking,

Special Issue on Watermarking, Signal Processing, volume 66, N°3, pp 337-356, Elsevier, May 985. R. Ohbuchi, H. Masuda and M. Aono, Watermarking 3D Polygonal Models Through Geometric and

Topological Modifications, IEEE Journal of Selected Areas in Communication, 16(4) , pp 551-560, May 986. G. W. Braudaway, K.A. Magerlein and F. Mintzer, Protecting Publicly Available Images with a Visible Image

Watermark, In Proceedings of the SPIE/IS&T on Optical Security and Counterfeit Deterrence Techniques,Electronic Imaging, volume 3016, pp 126-133, February 97

7. Error! Reference source not found.8. Error! Reference source not found.9. D. Kundur, D. Hatzinakos, Towards a Telltalt Watermarking Technique for tamper-Proofing, in Proceedings

of the IEEE ICIP 98, Volume 2 of 3, pp 409-413 ,Chicago, October 9810. W. Bender, D. Gruhl and N. Morimoto, Techniques for Data Hiding, in Proceedings of the SPIE, San Jose,

California, pp 2420-2440, February 9511. Cox, J. Kilian, F. Leighton and T. Shamoon, Secure Spread Spectrum Watermarking for Multimedia, IEEE

Transaction on Image Processing, volume 6, n°12, pp 1673-1687, 199712. J.J.K. O’Ruanaidh and T. Pun, Rotation, Scale and Translation Invariant Spread Spectrum Digital Image

watermarking, Special Issue on Watermarking, Signal Processing, volume 66, N°3, pp 303-318, Elsevier,May 98

13. F. Hartung and B. Girod, Watermarking of Uncompressed and Compressed Video, Special Issue onWatermarking, Signal Processing, volume 66, N°3, pp 283-302, Elsevier, May 98

14. C. Podilchuk and W. zeng, Image-Adaptive Watermarks Using Visual Models, IEEE Journal of Slected Areasin Communication, 16(4) , pp 525-540, May 98

15. M.D. Swanson, B. Zhu and A.H. Tewfik, Multiresolution Scene-Based Video Watermarking Using PerceptualModels, IEEE Journal of Slected Areas in Communication, 16(4) , pp 540-550, May 98

16. J.F. Delaigle, C. DE Vleeschouwer and B. Macq, Watermarking Algorithm Based on a Human Visual Model,Special Issue on Watermarking, Signal Processing, volume 66, N°3, pp 319-336, Elsevier, May 98

17. Fabien Peticolas, Ross J. Anderson and Markus G. Kuhn, Attacks on Copyright Marking Systems, inInformation Hiding : second international workshop, Lecture notes in Computer Science. Springer Verlag,Berlin, Germany, 1998