This paper provides an overview on the techniques and reference models used in digital watermarking. This paper proposes some technological means of authentication of multimedia content over the internet to enable secure connectivity while exploring the use of Digital watermarking framework for their realization. Examples are provided on some of the areas where this emerging technology concept is applicable including the Digitisation of government records which is ongoing in administrations worldwide. Mechanism by which digital watermarking can be carried out to provide means of security management in general, and multimedia copyright protection and content integrity verification in particular are explored
One of the stories in the Histories of Herodotus in which Histiaeus tattooed a message in the shaven head of a slave and waited for the new hair to grow before sending him to Aristagoras at Miletus with instructions to shave the slave’s head again is recounted by D. Kahn in his book “The Codebreakers” . Obviously, bandwidth was not a concern during that period, but methods haven’t changed so much when compared with the state of the art in digital watermarking.
Since the publication of a seminal work by Tanaka et al in 1990 , extraordinary growth of techniques for copyright protection of different types of data, especially multimedia information have been realized. This interest is not surprising in view of the simplicity of digital copying and dissemination: digital copies can be made identical to the original and later reused or even manipulated. Cryptography is an effective solution to the distribution problem, but in most instances has to be tied to use of specialized and costly hard ware configured to create tamper-proof devices that avoid direct access to data in digital format. Moreover, most cryptographic protocols are concerned with secured communications instead of ulterior copyright infringements. For instance, access control in set-top-boxes used for digital television demodulation and decoding succeed in avoiding unauthorized access to programs that are being broadcast in scrambled form  but fail in precluding further storage and illegal dissemination actions.
There are increasing needs for software and hardware that provides means of protection of ownership rights, and it is in this context where watermarking techniques come in handy.
Perceptible marks of ownership or authenticity have been around for centuries in the form of stamps, seals, signatures or classical watermarks. However, given current data manipulation technologies, imperceptible digital water- marks are mandatory in most applications.
Digital watermarks are distinguishing piece of information that are adhered to the data that it is intended to protect, meaning that it should be very difficult to extract or remove the watermark from the watermarked object. Since watermarking can be applied to various types of data, the imperceptibility constraint will take different forms, depending on the properties of the recipient.
In addition to imperceptibility there are some desirable characteristics that a watermark should possess, which are somewhat related to the so called robustness issue. First, the watermark should be resilient to standard manipulations of unintentional as well as intentional nature. Second, it should be statistically un-removable, that is, a statistical analysis should not produce any advantage from the attacking point of view. Finally, the watermark should withstand multiple watermarking to facilitate traitor tracing.
However, the type of manipulations and the attacker expected computational power heavily depend on the application. Watermarking, like cryptography, also uses secret keys to map information to owners, although the way this mapping is actually performed considerably differs from what is done in cryptography, mainly because the watermarked object should keep its intelligibility. In most watermarking applications embedment of additional information is necessary. This information includes identifiers of the owner, recipient and/or distributor, transaction dates, serial numbers, which play a crucial role in adding value to watermarking products.
2. Structure of a typical watermarking system
Every watermarking system consists at least of two different parts: watermark embedding unit and watermark detection and extraction unit. Figure 1.1 shows an example of embedding unit for still images. The unmarked image is passed through a perceptual analysis block that determines how much a certain pixel can be altered so that the resulting watermarked image is indistinguishable from the original. This takes into account the human eye sensitivity to changes in flat areas and its relatively high tolerance to small changes in edges. After this so-called perceptual-mask has been computed, the information to be hidden is shaped by this mask and spread all over the original image. This spreading technique is similar to the interleaving used in other applications involving coding, such as compact disc storage, to prevent damage of the information caused by scratches or dust. In our case, the main reason for this spreading is to ensure that the hidden information survives cropping of the image. Moreover, the way this spreading is performed depends on the secret key, so it is difficult to recover the hidden information if one is not in possession of this key. In fact, a similar technique is used in spread spectrum systems (more precisely, in Code-Division Multiple Access) to extract the desired information from noise or other users.
Additional key dependent uncertainty can be introduced in pixel amplitudes and finally, watermark is added to the original image.
Figure 2.1: Original ‘Fabric’ image
Figure 2.2: Perceptual Mask
Figure 2.2 represents the perceptual mask that results after analysing the image presented in Figure 2.1. Higher intensity (i.e., whiter) levels imply that higher perturbations can be made at those pixels without perceptible distortion. Thus, the higher capacity areas for hiding information correspond to edges. These masks are computed by using some known results on how the human eye works in the spatial domain. Different results are obtained when working on other domains, such as the DCT (Discrete Cosine Transform) or Wavelet transform. In fact, when working on the DCT coefficients domain one may take advantage of the relative independence between the maximum allowable perturbations at every coefficient. This is useful when dealing with the mask for water- marking purposes. Figure 2.3 shows the typical configuration of a watermark detection and extraction unit. Water- mark detection involves deciding whether a certain image has been watermarked with a given key. Note then that a watermark detector produces a binary output. Important considerations here are the probability of correct detection PD (i.e., the probability of correctly deciding that a watermark is present) and the probability of false alarm PF (i.e., the probability of incorrectly deciding that an image has been watermarked with a certain key).
These two measures allow us to compare different watermarking schemes: One method will be superior if achieves a higher PD for a fixed PF. Note also that for a watermarking algorithm to be useful it must work with extremely low probabilities of false alarm. Watermark detection is usually done by correlating the watermarked image with a locally generated version of the watermark at the receiver side. This correlation yields a high value when the watermark has been obtained with the proper key. As we have shown in , it is possible to improve the performance of the detector by eliminating original image-induced noise with signal processing. It is worthy of remark that some authors  propose using the original image in the detection process. Although this simplifies further treatment of the watermark in the receiver end, it is quite unrealistic for most applications, particularly those related to e-commerce.
Once the presence of the watermark has been correctly detected, it is possible to extract the hidden information. The procedure is also generally done by means of a cross-correlation but in this case, an independent decision has to be taken for every information bit with a sign slicer. In fact, it has been shown that this correlation structure has not been well-founded and significant improvements are achievable when image statistics are available.
For instance, the widely-used DCT coefficients used in the JPEG and MPEG-2 standards are well approximated by generalized Gaussian probability density functions that yield a considerably different extraction scheme. Obviously, when extracting the information the most adequate parameter for comparison purposes is the probability of bit error Pb, identical to that used in digital communications. This is not surprising because water- marking creates a hidden (sometimes called steganographic) channel on which information is conveyed.
3. Watermarking schemes and their applications
Digital watermarking schemes can be broadly classified into four categories namely: robust, fragile, semi-fragile, and reversible. While imperceptibility/low embedding distortion and security are the common requirements of all classes, each different category of scheme has different characteristics and, thus, is suitable for different applications. For example, while robustness is an essential requirement for copyright applications, it has no role in most authentication applications.
3.1 Robust Watermarking Schemes
Watermarks of robust schemes are required to survive manipulations unless they have rendered the content valueless in some senses. This class of schemes has found its applications in the following areas.
• Ownership proof and identification: A watermark containing the identification information of the content owner can be embedded in the host media for proving or identifying copyright ownership.
• Transaction tracking/fingerprinting: The copyright owner could insert a unique watermark, which, for example, identifies the recipient, into each copy of the media and uses it to trace the source should illegal redistribution occur.
• Broadcast monitoring: In the advertisement applications, by embedding watermark that is to be broadcasted along with the host media, the advertisers can monitor whether or not the commercials they have paid for are aired by the broadcasters according to the contracts. The embedded data can be used to verify whether the actual broadcasting of commercials tool place a scheduled, for devising an automated royalty collection scheme for copyright material (music, movies) that is aired by broadcasting operators. Other applications include Copy control/copy prevention as well as Copy Protection and Device Control
3.2 Fragile Watermarking Schemes
In contrary to robust watermarking, fragile watermarks are sensitive to all kind of malicious and non-malicious manipulations i.e. when manipulated the watermarks are expected to be completely destroyed and are therefore, they are useful for the following applications:
• Data Authentication: Digital watermarking is often used to prove the authenticity of a specific digital document where the watermark contains information that can be used to prove that the content has not been changed. If the watermark information can be extracted without errors, the authenticity can be proved. In order to design and effective watermarking algorithm, the watermarking data or procedure can be linked to the content of the digital document and such watermarks are called fragile watermarks .In the areas of military intelligence and news broadcasting, authenticity of the media sources is a key concern. By embedding a fragile watermark which identifies the source or producer in the media, the legitimate recipients of the marked media would be able to verify the authenticity of the received media by checking the presence of the source’s or the producer’s watermark. If the marked media is manipulated, the embedded watermark will become undetectable, the recipient will, thus, know that the media is not trustworthy. Although the main application of digital watermarking is to secure the intellectual property, it can also be used in the medical field as container for information about patients and their diagnosis. Medical records such as X-Ray images and Magnetic Resonance Imaging (MRI) could have such information embedded as watermarks for secure storage .
• Content-integrity verification: In the areas of medical image archiving, media recording of criminal events, accident scene capturing for insurance and forensic purposes, content integrity may have decisive impact on court ruling. The very presence of a fragile watermark in the original media allows the relevant parties to verify the integrity of the content.
An effective fragile watermarking scheme must have the capability of thwarting the attacks, such as cut-and-paste and vector quantization. However, fragile watermarks are sensitive not only to malicious manipulations but also to content-preserving operations such as lossy compression, transcoding, bit rate scaling, and frame rate conversion. Unfortunately, those content-preserving operations are sometimes necessary in many Internet and multimedia applications, making fragile watermarking feasible only in the applications, such as satellite imagery, military intelligence, and medical image archiving.
3.3 Semi-Fragile Watermarking Schemes
To facilitate the authentication and content-integrity verification for multimedia applications where content-preserving operations are a common practice, semi-fragile watermarking scheme have been proposed in the last few years. This class of watermarks is intended to be fragile only when the manipulations on the watermarked media are deemed malicious by the schemes.
However there are challenges posed in this form of watermarking scheme includes that a unique watermark may be extracted from many different media, which might have been subjected to some forms of content-preserving operations or malicious manipulations. Another challenge semi-fragile schemes faces is how to distinguish content-preserving operations from malicious attacks. For example, transcoding may be deemed acceptable for one application while it may be seen as malicious for another. Therefore, with these two issues, semi-fragile watermarking is usually not suitable for applications concerning legal and national security issues.
3.4 Reversible Watermarking Schemes
One limitation of the previously mentioned authentication schemes is that the distortion inflicted on the host media by the embedding process is permanent. Although the distortion is often insignificant, it may not be acceptable for some applications.
For example, any tiny distortion of an image, even if it were a result of the watermark embedding process itself, in the legal cases of medical malpractice would cause serious debate on the integrity of the image. Therefore, it is desirable that watermarking schemes are capable of perfectly recovering the original media after passing the authentication process. Schemes with this capability are often referred to as reversible watermarking schemes (also known as invertible or erasable watermarking.
4.0 General Authentication framework based on digital watermarking
Authentication is a variant of the previous application, in an area where cryptographic techniques have already made their way. However, there are two significant benefits that arise from using watermarking: first, as in the previous case, the signature becomes embedded in the message, second, it is possible to create ‘soft authentication’ algorithms that offer a multivalued ‘perceptual closeness’ measure that accounts for different unintentional transformations that the data may have suffered, instead of the classical yes/no answer given by cryptography-based authentication. Unfortunately, the major drawback of watermarking-based authentication is the lack of public key algorithms that force either to subject the secret keys at risk or to resort to trusted parties. Various types of watermarking schemes have been proposed for different applications. For the purpose of copyright protection, embedded watermarks are expected to survive various kinds of manipulation to some extent, provided that the altered media are still valuable in terms of commercial significance or acceptable in terms of visual quality. On the other hand, in medical, forensic, broadcasting, and military applications where content verification and identity authentication are much more of a concern, more emphases are focused on the capability of the watermarking schemes to detect forgeries and impersonation. This is intended to deal with watermarking schemes for authentication purpose. A general authentication framework based on digital watermarking is illustrated in Figure 4.
Usually, but not always, a secret key K available on both the embedding and authentication sides is used to generate a watermark to be embedded into the host media. The marked media is then delivered via the communication channel (e.g., internet, satellite, etc.) or stored in a database. To authenticate the marked media, the same secret key is used to generate the original watermark so as to be used for extracting and comparing against the embedded version. The difference map, the output of the authentication system, tells the authenticity of the media.
Figure 4: General authentication framework based on digital watermarking
Other applications where authentication based on digital watermarking plays a major role include:
Copy control/copy prevention: Illegal copying or recording is another common piracy scenario. One possible solution is to embed a never-copy watermark, that when detected in the recording device, disallows further recording. It can be used for notification of users that the item is copyrighted, for tracking illegal copies or for possible proof of ownership in case of legal dispute. However, this mechanism requires every recording device to have a watermark detector. It is difficult to persuade the consumers to pay more for a device that restrict their “freedom” of making copies. This commercially undesirable requirement is unlikely to be met without the support of global legislation.
Copy Protection and Device Control: In this scenario, the recording device scans the digital data stream for an existing watermark and enables or disables the recording action for a specific movie or video stream. Such technology could extend the pay-per-view concept and close the gap between the applied cryptographic approach and its usability.
4.1 Secure Digital Watermarking-Based Authentication Techniques in networks
As the interconnected networks for instant transaction prevail and the power of digital multimedia processing tools for perfect duplication and manipulation increases, forgery and impersonation become major concerns of the information era. As a result, the importance of authentication and content verification became more apparent and acute. In response to these challenges, approaches conveying the authentication data in digital media have been proposed in the last decade.
4.2 Applications for multimedia authentication
These can be found in many areas such as;
• Imaging / sound recording of criminal events: Authentic imaging or recording of legally essential event or conversation could lead to breakthrough in criminal cases while maliciously tampered imaging / recording, if not detected, could result in wrong ruling.
• Broadcasting: During international crises, tampered or forged media could be used for propaganda and manipulating public opinion. Therefore, broadcasting is an area where multimedia authentication is applicable. Recent advances in networking and digital media technologies have created a large number of networked multimedia applications. Those applications are often deployed in a distributed network environment that makes multimedia contents vulnerable to privacy and malicious attacks. For insecure environments, it is possible for an enemy to tamper with images during transmission. To guarantee trustworthiness, image authentication techniques have emerged to confirm content integrity and prevent forgery.
These techniques are required to be robust against normal image processing and transmission errors, while being able to detect malevolent tampering on the image . Such authentication techniques have wide applicability in law, commerce, journalism and national defence. The work extending the digital signature scheme from data (fragile or hard) authentication to content (semi fragile or soft) authentication Discovery of routes is a major task and for image authentication, it is desired that the verification method be able to resist content preserving modifications while being sensitive to content changing modifications.
This paper has explored the use of digital watermarking as a scheme to address copyright challenges in the digitization process of multimedia records and security of system for multimedia communication within the Internet domain. However, despite its potentials, it is by no means a cure-all solution for multimedia security management. For some applications such as copy control, non-technical backing by legislation is also crucial. With so many challenges and potential, we expect that digital watermarking will continue to be an active area of research.
 D. Kahn, The Codebreakers; The Compr hensive History of Secret Communication from Ancient Times to the Internet. Scribner, December 1996.
 K. Tanaka, Y. Nakamura, and K. Matsui, “Embedding secret information into a dithered multi-level image,” in Proc. 1990 IEEE Military Communications Conference, pp. 216–220, 1990.
 B. Macq and J. Quisquater, “Cryptology for digital TV broadcasting,” Proceedings of the IEEE, vol. 83, pp. 944–957, February 1995.
 J. R. Herna´ndez, F. Pe´rez-Gonza´lez, J. M.Rodr´?guez, and G. Nieto, “Performance analysis of a 2d-multipulse amplitude mod- ulation scheme for data hiding and water- marking of still images,” IEEE J. Select. Ar eas Commun., vol. 16, pp. 510–524, May 1998.
 I. J. Cox, J. Kilian, T. Leighton, and T. Shamoon, “A secure, robust watermark for multimedia,” in Information Hiding (G. Goos, J. Hartmanis, and J. Leeuwen, eds.), vol. 1174 of Lecture Notes in Com puter Science, (University of Cambridge, UK), pp. 185–206, Springer-Verlag, May 1996.
 F. Petitcolas, R. Anderson, and M. Kuhn, “Attacks on copyright marking systems,” in Information Hiding (D. Aucsmith, ed.), vol. 1525 of Lecture Notes in Computer Science, (Berlin), pp. 218–238, Springer- Verlag, 1998.
 Panyindee, C. ; Dept. of Electrics, King Mongkut”s Inst. of Technol. Ladkrabang, Bangkok, Thailand ; Pintavirooj, C. Reversible watermarking algorithm in application for medical images” Biomedical Engineering International Conference (BMEiCON), 5-7 Dec. 2012,Page(s):1 – 5 ISBN:978-1-4673-4890-4,: Available at //ieeexplore.ieee.org/xpl/mostRecentIssue.jsp
 I.J. Cox, M.L. Miller, J.A. Bloom, J. Fridrich, and T.Kalker, Digital Watermarking and Steganography, 2nd Edition, San Francisco, CA, USA.: Morgan Kaufmann Publishers, 2008.