Digital watermarking, a popular technique for content authentication and forgery prevention, is viewed as an enabling technology to protect distributed content from unauthorized reuse, or reuse with inadequate credit. Digital watermarks of ownership embedded onto digital content offer copyright protection, ownership assertion, and integrity checks for digital content [4], and can provide evidence of copyright infringement after an attack. Watermark researchers always claim their watermarking techniques are robust and secure, but they fail to address the fundamental question of just how secure. Several information systems professors raised this issue after my presentation on watermark-based copyright protection at an e-commerce conference, and their queries motivated me to investigate this research question and present my results in this article.
A key issue to resolve in watermark-based copyright protection is the deadlock problem, which occurs when a pirate inserts a watermark in publicly available content and claims he or she is the legitimate owner of the copyright-protected content. Current watermark-based copyright schemes are unable to establish who watermarked the data first. Previous research tackling the deadlock problem was based solely on technology innovations, and unfortunately, those results were unable to persuade investors and executives that a watermark-based copyright protection system is fail-safe. While the research on this technology is far from complete, and a solid copyright protection infrastructure could, to a certain extent, solve the deadlock problem, this article takes the approach that technology cannot be the only direction from which to address the problem. Here, I respond to the question of the breakability of a watermark-based copyright protection system and address the deadlock problem from a non-watermarking angle. The contributions of this article include an anti-deadlock strategy to resolve the deadlock problem.
What makes the Internet an effective digital distribution channel, which reduces the marginal costs of digital content production and distribution, also creates opportunities for content and copyright infringement and piracy problems. This doubled-edged sword has affected the music industry more than any other content publishing industry. Record companies assert that widespread unauthorized distribution and reproduction of digital music facilitated by the Internet have resulted in losses of millions of dollars. On the other hand, record companies also believe the Internet could provide a new source of income. While music labels seek moneymaking business models and secure technologies for digital music distribution, music sharing and music dissemination activities are growing faster than ever. The major players in the online music distribution realm are eagerly seeking clarity about whether a watermark-based copyright protection system is breakable, for a truly secure watermark-based copyright protection system will allow them to thrive.
A copyright protection system is a mechanism to protect the rights of ownership of distributed content so no one can use the copyright-protected content from an online content distribution site without getting permission first, and if necessary, paying the copyright owners a royalty for usage. The owners generally include the content creator, distributor, and consumer [3]. The content distributor is usually responsible for inserting the owners' identities into the content. In a watermark-based system, the identities are represented as digital watermarks, and watermarking processes are performed to insert, detect, and extract the embedded watermarks. A digital watermark is basically an integer number, also known as watermark value. Therefore, all distributed content is copyright-protected.
The deadlock problem occurs when multiple ownership claims are made, and the rightful ownership of digital content cannot be resolved [1]. The inability to deal with the deadlock problem is independent of how the watermark is inserted into digital content or how robust it is to various types of modifications [6]. Publicly available and non-copyright protected content is the most vulnerable to ownership deadlocks. While some researchers have attempted to resolve the deadlock problem by applying multiple watermarks [5], it is the opinion of this author that watermark-based solutions, whether multiple watermarks or highly robust watermarks, are not the ultimate solutions to this problem. Instead, watermarking is simply a vehicle to drive copyright protection, and copyright protection should be enforced by concerned authorities.
Three ways exist to break a watermark-based copyright protection system. Attackers can remove embedded watermarks or cause them to be undetectable; they can replace a watermark with a false one; and they can insert an additional watermarka false oneinto copyright-protected content, which creates the deadlock problem.
The first approach involves performing signal-processing operations such as compression and file conversion on the copyright-protected content, which may change the embedded watermark and in the worst case render it undetectable. If the watermarking is robust enough, the copyright-protected content should be able to resist many intentional or unintentional signal-processing operations. The simplest way to remove an embedded watermark is to average the watermarked content and the original content. But the pirate must have access to the original content. Other easy ways to remove a watermark rely on publicly available programs and tools, such as image-processing programs, music jukeboxes, and signal-processing tools.
Since removing watermarks from copyright-protected content results in turning the copyright-protected content into ordinary content, attackers who wish to own the content may choose to replace the watermark. This approach is far more challenging than converting copyright-protected content to ordinary content. It requires replacing the rightful owner's watermark (usually the consumer's watermark) with the attacker's watermark. This can be achieved either by physically removing the owner's watermark and inserting another one using the same watermarking technique as the distributor, or modifying the embedded watermark so the watermark detection system returns a watermark value the attacker prefers. Both approaches require exceptional skills in and signal processing, and potential attackers are often insiders who know the watermarking processes well, and have access to confidential user information.
Since current watermarking technology cannot resolve multiple ownership claims, the deadlock approach is the easiest way to break the copyright protection system and it is the major problem a copyright protection system should address. Adding the attacker's watermark to the copyright-protected content and claiming ownership of the content effectively breaks the copyright protection system. Such activity requires no specific skills and knowledge about the watermarking technique used by the distributor. The attacker may apply any watermarking technique available and present the extracted watermark as evidence to a court or the corresponding government authority to claim ownership of the content.
Approaches that attempt to resolve the deadlock problem by watermarking technology [5] could fail if the attacker applies the same technique in embedding the attacker's watermark. Instead, the key to the anti-deadlock strategy should not depend on watermarking, but on a watermark clearance center (WCC), a well-known authority with a role analogous to a certificate authority (CA). This authority could be responsible for managing all watermark issues, especially judging the ownership of suspected copyright-protected content.
A simple anti-deadlock strategy with a WCC could involve a mechanism to set the valid ranges of watermarks for various entities involved, including the content creator, distributor, and consumer. Varying content, creators (or publishers), and distributors may have their own watermark settings. Watermarks would be regarded as valid if their detected watermark values were within the valid ranges preset by the distributor and maintained in the WCC. Any detectable watermark outside the valid ranges would be considered an attacker's watermark. When an attacker's watermark was detected, the WCC and the content distributor would take follow-up action to stop the spread of the infringed content.
Attackers attempting to break the watermark-based copyright protection system must possess both the watermarking parameters and the valid range information for generating and inserting their watermark into the content. The latter information is difficult to obtain. As the figure here illustrates, the distributor can set the valid ranges of all three watermarks and then inform the WCC of the settings. The valid ranges are known only by the WCC and the distributor.
As stated previously, from the attacker's perspective, disabling an embedded watermark is far easier than deleting an embedded watermark from the copyright-protected content and replacing it with another. But in the future, when all downloadable and distributed content is copyright-protected, disabling the embedded watermark will lead to a fatal error in the media player. If the attacker manages to obtain all watermarking parameters and insert a watermark into the copyright-protected content, it is almost impossible for the attacker to notice the existence of the valid ranges of the watermarks. Lacking such information, the attacker's watermark may alert the distributor and the WCC of potential attacks when the media player tries to render the copyright-protected content.
The distributor and the WCC may take further actions to prevent the spread of the violated content by setting valid ranges for the content, invalidating the content with the previous valid range setting for further usage until a new protection scheme is applied, and tracking the potential attackers for prosecution, as the WCC may log the usage of the copyright-protected content when requested. However, these actions cannot fully resolve the deadlock problem, because the attacker's watermark may be within the valid range, and the deadlock problem still exists. But at least such approaches minimize the probability of having unambiguous proof of ownership.
By hiring a trustworthy WCC for the management of ownership identification, verification becomes the essence of the anti-deadlock strategy. The WCC can be a government authority responsible for issuing, renewing, and revoking digital watermarks to creators, distributors, and consumers; generating watermark revocation lists; and publishing watermarks through a directory server. In managing ownership, the WCC would act as a judge to verify the suspected distributed content files both online and offline [2]. Online copyright enforcement could be incorporated with the media player, while offline copyright enforcement could be handled within the WCC. The valid range settings of watermark values can be achieved. Unfortunately, no official WCC is currently being established, but as more online businesses realize the importance of copyright protection, such an authority will emerge.
The question of whether watermark-based copyright protection is breakable is of interest to many in the academic and business communities. Business owners are concerned about the risks to their online content distribution businesses, while researchers face the intricate deadlock problem in watermarking. Previous efforts to resolve this problem have focused on watermarking technology, with no success. This article suggests another angle to view the problem and proposes an anti-deadlock strategy to reduce the chance of incurring deadlocks in copyright protection systems. This analysis of the breakability of watermark-based copyright protection systems suggests not technology, but a well-known third party or government authority, or WCC, should manage copyright protection.
In summary, breaking a watermark means one can remove the embedded watermark so the copyright-protected content carries an incomplete set of watermarks; modify the embedded watermark so ownership can be transferred to another owner; or insert one or more watermarks using the same watermarking technique as the distributor into the copyright-protected content, so that the rightful owners are unable to claim ownership.
The ease of removing a watermark directly relates to the robustness of the watermarking technique. Many watermarking techniques claim they are robust to many operations and attacks, but attackers are usually knowledgeable and skillful in removing watermarks. However, this does not help when the media player requires watermark for check-in, as it will in the future.
Breaking the system by modifying the embedded watermark is not easy. An attacker must possess proper information about the watermarking technique and embed another watermark into the content by using the same technique. It is unlikely the attacker can gather such information and programs unless he is the distributor, or someone who works at the distribution site.
Interrupting a copyright protection system by adding more watermarks to the copyright-protected content is generally the easiest way to break the system. With the proposal of the WCC and valid range settings for watermarks, any invalid watermark or attacker's watermark can be detected, and follow-up actions can be carried out by the WCC and the distributor. The WCC should possess sufficient evidence to prove content ownership.
With a proper anti-deadlock strategy and a WCC in place, the watermark-based copyright protection system will not be easily breakable by attackers. It is foreseeable that online content distribution will be popular in the future, but it requires a solid copyright protection scheme. A common international standard should solidify the foundation of copyright protection. By then, all distributed contents will be copyright protected, and the roles of the WCC will be recognized and valued.
1. Craver, S., Memon, N., Yeo, B.L., and Yeung, M. Can invisible watermarks resolve rightful ownerships? IBM CyberJournal, July 1996.
2. Kwok, S.H. and Lui, S.M. An enhanced license management model in digital rights management for online music business. International Journal of Information Technology and Decision Making 1, 3 (2002) 541558.
3. Kwok S.H. and Yang, C.C. Copyright protection in media distribution e-service. In Proceedings of the Third IEEE Conference on Information Technology (Apr. 2002) 158163.
4. Memon, N. and Wong, P.W. Protecting digital media content. Commun. ACM 41, 7 (July 1998), 3543
5. Swanson, M.D., Zhu, B., and Tewfik, A.H. Multiresolution scene-based video watermarking using perceptual models. IEEE Journal on Selected Areas in Communications (May 1998), 540550.
6. Swanson, M.D. Zhu, B., Tewfik, A.H., and Boney, L. Robust audio watermarking using perceptual masking. Signal Processing 66, 3 (May 1998), 337355.
©2003 ACM 0002-0782/03/1000 $5.00
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee.
The Digital Library is published by the Association for Computing Machinery. Copyright © 2003 ACM, Inc.
No entries found