Digital Watermarking

Watermark Enforcement

Digital watermarking does not prevent any copying nor apply enforcement. Its effectiveness depends on providing evidence of illicit copying and dissemination by extracting watermarks from stolen images. Individual examination of suspicious images very costly, although it is unavoidable in some cases where human interference is needed to restore a seriously distorted image in order to detect watermarks from it.

The efficient and low-cost detection of watermarks from hundreds of millions of pieces of multimedia information is crucial to the success of digital watermarking technology. In this section we will discuss two types of watermark enforcement tools to automate the watermark detection, namely, the watermark monitor and the watermark tracking agent.

Watermark Monitoring:

A watermark monitor is a device which sits at a place and receives images broadcasted from satellites, or downloaded images via surfing the Internet. With the wide availability of many spiders or robots that are indexing millions of images and documents, it would be easy for the watermark monitor to check the images that have been indexed by these spiders in their databases (such as Alta visa or Yahoo).

Since watermark extraction requires the secret key, a monitor can be only dedicated to the detection of the watermarks that were embedded in one secret key or a few secret keys. In the latter case, the monitor must apply each of all secret keys to each image in order to find the watermark.

When a watermark monitor is dedicated to detect the owner-watermark, it produces reports to identify who is using the images that are owned by a specific entity. By comparing the reports with licensing records, copyright infringements can be easily found.

When a watermark monitor is dedicated to detect the recipient-watermark, it reports who has illegally distributed copyright protected images.

Implementation of a watermark monitor is straight forward with no requirements for reconfiguration or add-ons to the server. The drawback is the heavy burden on the monitor: each image has to be downloaded to the monitor and all operations will be performed locally within the monitor.

Watermark Tracking Agent:

A watermark agent acts like a ‘detective’ to trace any illicit copies by detecting watermarks and verifying the usage rights of the image. A watermark agent is mobile, for example, travelling from server to server on the Internet to check watermarks of the local images and send reports to the watermark clearinghouse and/or dispatcher.

The security of this agent is a major concern. In addition to the security concerns at the language level, i.e. determination of whether the agent will attempt to infect the server, deny service to other agents, an important security and privacy issue which has arisen from the watermark agent is: to prevent anyone on the network from detecting confidential information, such as watermarking key and detection reports, while the agent is migrating from host to host.

Both watermark monitoring and tracking agent require a knowledge database to guide them in locating and travelling. We propose centralized knowledge base built incrementally by the feedback and analysis of the watermark monitoring and tracking. For example, a site where illicit copies were detected should be visited frequently by the agent in the future.

All watermark agents bring reports to the watermark agent clearinghouse, which is responsible to dispatch the agents, collect and compile reports, and take appropriate actions. Like a watermark monitor, an agent is usually dedicated to one right holder with one or few recipient-watermark.

To determine copyright infringement automatically, a watermark agent must collaborate with other components such as a copyright registration agency, a copyright management information system, a watermark deposit service, and perhaps with other watermark agents as well as watermark monitors.

IMPLEMENTATION

As depicted in Figure 2, a web-based prototype image commerce system has been implemented and integrated into the Netscape environment. For great portability and extendibility, this prototype has been developed in Java and related tools. A set of Java classes provides the service provider with a flexible and powerful means to

1. Define the user interface within the browser
2. render the image; and
3. Invoke SysCoP API (as Java native methods) to perform watermark embedding and retrieval functions.

This prototype implements online image commerce on the World Wide Web as Illustrated in Figure 2. Assume that a public watermark containing a copyright notice and a Secret watermark with the author name and image identifier have been embedded into each Image stored in the online image database on the server. The whole transaction consists of

Five steps:

1. Once a client makes a request for an image, the server verifies the identity of the client through the built-in SSL (SecureSoa simple application-defined password based authentication.
2. If the verification succeeds, the customer’s identity and/or a transaction ID will Be embedded into the image as another hierarchical watermark (In addition to public and ownership water markings).
3. Optionally, the image can be encrypted using a one-time session secret key Generated on the server. We use TIE (Tool for Image Encryption)
   Developed at the Fraunhofer Institute for Computer Graphics to encrypt the Image. We will briefly discuss TIE later in this section.
4. Finally, the server sends the scrambled image and the session key which is Encrypted with the client’s public key, to the client.
5. On the client side, the browser activates the embedded Java interpreter to Execute Java programs and native methods to perform decryption, Rendering, and watermark retrieval.

The failure of detection of the watermark indicates that the image has been significantly tampered with and a window pops up to inform the user of the alteration.
To offer a preview and at the same time to avoid duplicate transmission of the image data, TIE[8] encrypts partial image data using the standard DES and replaces these data with low-resolution patterns, storing the encrypted data in the extension block of the image. Thus, the customer can convert a preview (scrambled image) into a full-value image by decrypting the data stored in the extension area once he or she obtains the secret encryption key.

In addition to the copyright notice, the public watermark contains a control message for copy protection. Two types of use-controls have been implemented: permission for printing; and, permission for storing an image into a local file. Embedding the use-control message into the image provides a universal solution to binding the rights with the content in a secure and robust way for the entire life of the content.

A Java-based mobile agent is underdevelopment to monitor and analyze images on the Web aiming to identify and track copyright infringements and the illicit use of copyrighted images. Moreover, by inserting the rank of confidentiality and/or the scope of distribution into electronic confidential documents, the watermark monitor and agent can detect and respond to any leakage of the confidential information in an unsecure Internet or a trusted Intranet domain.

Conclusion

Digital watermarking has emerged as an essential security technology, fully complementary to encryption-based secure transmission and copy protection. This technology is widely applicable to almost all electronic information services and systems.

Where data protection, security, and intellectual property rights are needed. However, effective uses in fields such as digital camera, DVD, digital broadcasting, electronic commerce, and law enforcement have not been explored. This paper presented a scheme to effectively apply the digital watermarking technology to online service for multimedia data. Three types of watermarking, namely, public, ownership and recipient watermarking, cover all concerns of parties in the multimedia chain. Two forms of watermark enforcement tools enable service providers and right holders to defend their intellectual rights and to deter illicit copying in a low-cost and efficient way. Combined with a partial image encryption method, the prototype described in this paper provides a multiple-level security solution to online multimedia commerce. I would like to thank Eckhart Koch and Dittmar Stork for their fruitful discussions and Xuedong Liu and Cheng Dong for the implementation of a prototype of the Sysco Java.

BIBLIOGRAPHY

1. Shahrzad Esmaili, Sridhar Krishnan and Kaamran Raahemifar, “Audio watermarking using time-frequency characteristics”, IEEE journal, Can. J. Elect. Compute. Eng., Vol. 28, No. 2, April 2003.
2. Hyoung Joong Kim, “Audio watermarking techniques”, Kangwon National University.
3. Koch, E. and Zhao, J. Towards Robust and Hidden Image Copyright Labeling. In:
   Proc. of 1995 IEEE Workshop on Nonlinear Signal and Image Processing Neos Marmaras, Halkidiki, Greece, June

INTERNET REFERENCES

1. www.wikipedia.org
2. http://www.ece.uvic.ca/~aupward/w/watermarking.htm