"The end result will be failure. All digital copy protection schemes can be broken, and once they are, the breaks will be distributed ... law or no law. ... Digital files cannot be made uncopyable, any more than water can be made not wet."
- Bruce Schneier, cryptography expert
Digital copy protection is emerging as one of the highest stakes issues for consumer rights, digital age business models, and copyright law. To date, content owners have been on the losing end of a battle over bits. Hackers have successfully cracked copy protection on everything from DVD's to copy-proof CD's. Conscious of the failure of digital content protection systems and the widespread distribution of unauthorized copyrighted works over the Internet, the content industry is desperate for a technological or legal content protection. Most eyes are currently turned to advances in "digital rights management" ("DRM") technologies, which offer an unprecedented level of control over digital content. Additionally, such control could create new, restrictive business models and revenue sources for the content industry.
At the same time, the technology industry has been quietly developing an initiative known as "trusted computing," which aims to put security features deep into the hardware of personal computers. Trusted computing systems can protect individual files and sensitive data on a computer, as well as verify that the components of a computer are in a known state. Such a technology poses great advantages for electronic commerce as well as personal security for the public.
At the intersection of digital rights management and trusted computing lies a new approach to content protection. The content industry is pressing the technology industry to utilize the security of trusted systems to enable a much stronger digital rights management scheme. Many academics and civil libertarians worry that absolute content control could cause collateral damage. Advancing digital rights management technology could render many public freedoms under copyright law technologically impossible, while expanding liability under the controversial Digital Millennium Copyright Act.
This paper will examine the copyright implications of the new trusted computing initiative as a building block for stronger digital rights management schemes. As with many emerging technologies, there is some debate as to what "trusted computing" actually describes. In the academic and civil libertarian circles, "trusted systems" and "trusted computing" are often synonymous with digital rights management. Moreover, even knowledgeable critics within the technology community often "improperly lump together TCPA [the Trusted Computing Platform Alliance specification], Palladium, and DRM, considering them as one thing." To these commentators, "trusted systems" are a tool that absolutely enforces copyright restrictions and prevents digital copyright infringement on an end user's computer.
However, the actual technology dubbed "trusted systems" and "trusted computers" presently under development doesn't quite fit this bill. Technological "trusted systems," as touted by the Trusted Computing Group and Microsoft's Next Generation Secure Base, will not restrict digital content, nor implement any digital rights management scheme. Technological trusted systems are only security sub-systems, available for any number of security purposes. Consequently, this paper will use the phrases "trusted computing" and "trusted systems" to describe emerging trusted technologies, and not the academic / civil libertarian general concept of digital rights management.
This paper contends that advent of trusted computing has serious implications for the evolving digital copyright debate. First, digital rights systems based on trusted computing technology will create additional complications for anti-circumvention liability under the Digital Millennium Copyright Act. Second, trusted computing and digital rights management could effectively "privatize" copyright law - enabling content owners, and not the law, to decide exactly what rights the public will have over digital content. Finally, trusted computing cannot completely secure digital rights management over the long term. In fact, this paper will argue that no technology can. The content industry's continuing legal and technological pursuit of digital rights management in the face of technological impossibility threatens to create a digital "arms race," harming the balance of copyright law and consumer rights.
Part I of this paper explores digital rights management, both as a possible cure for digital piracy and as a tool for absolute control of copyrighted works. Part II explores the rapidly developing field of trusted computing, and follows the two most prevalent initiatives - the Trusted Computing Group and Microsoft's Next Generation Secure Computing Base. Part III contends that despite the security advances of trusted computing, no technology can enforce perfect digital rights management. Part IV gives a brief background of traditional copyright law and specifically examines liability for the circumvention of technological protection measures under the Digital Millennium Copyright Act and the DeCSS case. Part V examines the legal and technological intersection of copyright law and digital management, as trusted computing quickly becomes a tangible reality.
"Digital Rights Management" ("DRM") generally describes technologies that "restrict the use of digital files in order to protect the interests of copyright holders." DRM can prevent or restrict a computer from "altering, sharing, copying, printing, [or] saving" protected digital files. More importantly, DRM allows copyright owners very detailed control over the ways in which a user may access their files (how long they may view a file, how many times a file may be accessed, etc.). By way of example, Disney could release a digital video file of "Snow White" to a fictional user, Joe, that could only be played on the Joe's personal computer ("PC"), could not be copied or altered, and could be watched only 3 times. After watching the video twice, Joe could not loan his file to a friend to use up his third viewing, as it only plays on Joe's specific computer. If Joe wanted to watch the video file more than 3 times, or wished to save a portion of the video for his own personal use, he would be unable to do so without further permission from Disney.
DRM relies on two primary aspects to secure content for digital rights owners: "containment" and "marking." The content industry's concern is that once a digital file is sent to the user's computer, the user may attack and extract the information from the file for unrestricted use. Thus, for DRM to work, the digital content must be "contained" so that it may only be accessed in authorized ways. "Containment" is generally accomplished by encrypting distributed digital content so that only programs authorized by rights owners may decrypt, and thus access the information. Additionally, DRM systems must "mark" which uses of a digital file are authorized. This may be done with a "watermark," "flag" or an XrML tag. Returning to the previous example, Joe's digital video file of Snow White could be encrypted ("contained") so that Joe would be unable to access the data by any means other than a Disney-approved viewing program. Joe's video file would be "marked" in XrML with the instructions to: (1) play only on Joe's computer, (2) disallow any copying or alteration, and (3) play only 3 times. The video program would understand these instructions and limit Joe's viewing of Snow White accordingly.
Existing DRM systems are susceptible to attacks or reverse engineering that render digital content unprotected. Presently, digital content must be downloaded or streamed to an end user's computer. Once that information resides on an end user's computer, it is particularly vulnerable. Professor Ed Felten of Princeton University contends that both the "containment" and "marking" DRM techniques can be defeated by end users with even "moderate" programming skills.
Ed Felten classifies two "threat" models for digital content owners. The first is the "Napsterization" threat model, which "assumes that there are many people, some of them technically skilled, who want to redistribute [a copyrighted] work via peer-to-peer networks; and it assumes further that once [the digital] content appears on a p2p network, there is no stopping these people from infringing." Security under the Napsterization model must ultimately prevent even one copy of a digital file from breaking the security of the DRM system, because one copy may be widely distributed and used over the Internet. This threat model requires that a "DRM technology must be strong enough to stymie even the most clever and determined adversary."
The second threat model is the "casual-copying" model, which "assumes that you are worried about widespread, but small-scale and unorganized, copying among small groups of ordinary consumers." The casual-copying model isn't concerned with de minimis digital file cracks, provided that the majority of distributed copyrighted files remain secure. Securing against a casual copying threat model is a much more feasible goal for DRM systems.
Professor Felten notes that most digital content owners complain about the Napsterization model, while proposing DRM solutions that address the casual-copying threat model. However, solving the Napsterization threat model would require a DRM system to prevent single attacks from threatening the entire system. Peter Biddle et al., of Microsoft Corporation, notes that such a DRM system must "strive to be BOBE (break-once, break everywhere)-resistant." Biddle concludes that the ultimate goal in DRM design is to create a situation where "knowledge gained breaking one client cannot be applied elsewhere."
The growing concern is that a BOBE-resistant DRM system, capable of defeating the Napsterization threat model, may well be impossible. The emerging consensus among security experts is that DRM is "fundamentally insecure." Ed Felten observes that although theoretically unbreakable encryption codes exist, DRM systems cannot utilize these effectively to secure their content . DRM systems must ultimately leave digital content on the computer of an end user, a "presumed adversary," who has any length of time to attack a DRM file. Moreover, digital content must eventually be decrypted into a usable format (sound, video, etc) by an authorized DRM program, creating another opportunity for interception and attack.
Abstractly, the hopes for secure DRM are bleak. In practice, the situation is much worse. To date, most DRM systems have been circumvented with trivial effort. In September, 2000, the Secure Digital Music Initiative ("SDMI") opened up a public contest to see if any developers could break its DRM watermarking scheme for digital music files. Theoretically, this "watermark" would prevent any unauthorized use or illicit copying, because the music file would be unplayable under unauthorized circumstances. By November 2000, Professor Ed Felten and a group at Princeton University successfully broke the entire scheme. Their attack removed the watermark and converted the digital file into an unprotected format without degrading the music quality of the file. Consequently, a single DRM-protected file released in SDMI format can be (1) cracked, so that an end user can play the file in an unauthorized manner, (2) cracked using techniques that should work on any SDMI file, and (3) ultimately released to the public at large in the popular (and unprotected) MP3 format, thwarting all existing SDMI protection for that specific music file. Professor Felten postulates that the experience with SDMI is indicative of a greater problem with current DRM technologies: "The underlying problem that SDMI is trying to solve, that of protecting content from a hostile platform while allowing the platform to 'play' the content, is inherent very difficult, both in theory and in practice. To overhaul their system, SDMI may well have to overhaul their business model."
Despite the readily apparent problems with copyright controls in the digital wild, the pursuit of and conflicts over DRM will only continue to escalate.
Digital technology simultaneously implicates the content industry's worst nightmares and its greatest hopes. In the pre-digital era, the unauthorized distribution of copyrighted works was cumbersome. Only professional counterfeiters achieved any continuous and widespread distribution. Computers and the Internet changed everything. Digital technology brought perfect copying functionality and easy worldwide distribution to even the most casual computer user. One only needs look as far as Napster and Peer-to-Peer file sharing to see that the public indeed wants to copy, share and distribute digital works.
The content industry sees DRM as the only possible hope in combating the continued, widespread copying of copyrighted digital works. George Colony of Forrester Research observes that "[w]e're entering a period of three to seven years where entertainment companies keep trying to control and consumers keep trying to escape it. ... There's a lot of money at stake here and Hollywood doesn't want to lose it." Jack Valenti, the prolific president of the Motion Picture Association of America, contends stronger digital protection is necessary for content providers: "We need to put in speed bumps to keep people honest[.] ... If we don't, our future is bleak."
Not just a defensive measure, DRM also gives the content industry the possibility of reaping far more money and exercising far greater control over copyrighted works than was ever feasible in the pre-digital era. Once a paper book is sold, the content industry is powerless to stop a consumer from reading the new book a particular number of times, owning the book for extended periods of times, and/or eventually selling the used book to someone else. DRM allows content owners to release the same book in digital file format, while retaining absolute control over the number of times the same consumer can view the file, keep the file on their computer, or transfer the digital file. Larry Kenswil, president of the eLabs division of Universal Music Group notes that under this new paradigm, "[y]ou're not buying [content], you're buying a key[.] ... That's what digital rights management does: it enables business models."
Thus, creating more effective DRM is the veritable "holy grail" of technology for the content industry. DRM offers to plug the copyright holes of the digital age and give copyright owners unprecedented power over the public's use of digital content.
Despite repeated failures, DRM technologies continue to proliferate, most notably in the area of digital music. Ed Felten and other researchers quickly dashed the hopes of SDMI as a DRM solution for music. John Borland, a technology writer, describes the same result for almost all CD copy protection schemes:
[P]revious versions of the antipiracy technologies from SunnComm, Macrovision and others have proven flawed. CDs protected with the technology have been unable to play in some CD players or computers, potentially even damaging some machines. Hackers have been able to break through much of the protection technologies using techniques as simple as drawing on the CD with a felt-tipped pen.
Nonetheless, imperfect DRM solutions are viewed as "deterrent[s]" until better technologies become available. SunnComm and other DRM producers press on with new DRM technologies, hoping that maybe one time, it won't break.
While most CD DRM solutions have been confined to test markets, this is soon to change as well. A recent research note from J.P. Morgan analyst Sterling Auty expects "volume shipments of protected CDs to ship commercially in the U.S. as early as the May-June time frame using the SunnComm solution," and concludes that "[t]his will be the first major step in the growth of the CD audio protection market." Likely responding to Auty's note, Macrovision claims that already over 100 million CD's in distribution worldwide are protected by its technologies. Additionally, DRM deployment continues in a multitude of other digital fronts, like DVD's, e-mail and word processing document files and high-definition TV's ("HDTV's").
The burning desire of the content industry for DRM is met by continued development from the technology industry. Technology companies with the most effective DRM stand to gain advantage in the digital content arena. Professor Ross Anderson summarizes the technology-entertainment industry relationship by way of an example of Intel:
[Intel] makes most of its money from the PC microprocessor; they have most of the market; so to grow the company they need to grow the overall market for PCs; that means making sure the PC is the hub of the future home network; and if entertainment's the killer app, and DRM is the key technology for entertainment, then the PC must do DRM.
However, not all of the technology industry tows the content industry line. Many technology companies make products that are attractive for their use of unprotected digital formats, and see DRM as antithetical to their interests. Also, some companies attempt for the best of both worlds. Although recently developing its own moderate DRM system for music files, Daniel Steinberg notes that Apple "has been in the forefront of creative freedom and continues to enable creative expression with the suite of iApps," which enable editing and playback of a variety of digital content.
Thus, despite its failures, DRM is the near-term and long-term goals of the content industry. And much of the technology industry has vested interests in creating DRM to meet these goals. Hilary Rosen, the former chief executive of the Recording Industry Association of America ("RIAA"), has the optimistic view that "[w]hile the technology is apparently not quite ready, there is promise for some protective technologies,"
Weary of the technological limits of DRM solutions and the reluctance of some technology producers to embrace a complementary vision, the content industry has additionally pursued DRM through more malleable arena of politics. The content industry's failsafe plan is to politically reign in the public and dissenting technology companies that presently benefit from refusing DRM in popular products. J.D. Lasica, senior editor of the Online Journalism Review, observes that for content-embracing companies, "any opposition to content control is frowned upon in the halls of Congress. The film industry has the attention of the Congress and has thrown 60 million dollars at it."
The content industry's cries are falling on receptive ears. In September of 2001, Senator Ernest "Fritz" Hollings (D-SC) scheduled a Congressional introduction for his bill, the Security Systems Standards and Certification Act ("SSSCA"). The SSSCA draft mandated DRM technology by creating civil offenses for the sale or creation of any computer that "does not include and utilize certified security technologies." The SSSCA also enumerated new federal felonies for trafficking in copyrighted content with disabled "security measures." On March 21, 2002, Senator Hollings introduced a modified version of the bill under a new title, the Consumer Broadband and Digital Television Promotion Act ("CBDTPA"). The CBDTPA mandates that every new electronic device short of a digital "clock" have built-in DRM technology:
[The CBDTPA] would have movie studios, record labels and others attach digital tags to a movie, song or album that would encode rules about how it could be played, viewed or copied on devices such as computers or digital TVs.
Manufacturers and content owners would have a year to agree on technology to enforce these rules; after that, the Federal Communications Commission could impose a standard. It would then be illegal to manufacture devices that didn't implement it.
In addition to Hollings' far-reaching gambit, many politicians have offered related legislation. Proposed legislation and rules range from mandating copy protection codes in digital television signals, to permitting the content industry to pursue "government-sanctioned vigilantism" against online file sharing systems. In July of 2002, Representative Howard Berman (D-CA) introduced an anti-piracy bill, H.R. 5211, to combat Peer-to-Peer file sharing. H.R. 5211 would allow content owners to attack and hack file traders believed to be trafficking in unauthorized copyrighted works. The bill would offer individuals "almost no recourse" if they were wrongly attacked by copyright owners.
Washington, however, is rarely on one side of an issue. Several legislators have proposed bills limiting the ability of content owners to implement DRM as well as prohibiting DRM-friendly legislation. On March 4, 2003, Representative Zoe Lofgren (D-CA) introduced H.R. 1066, the Benefit Authors without Limiting Advancement or Net Consumer Expectations (BALANCE) Act of 2003. The BALANCE act would include analog and digital transmissions of copyrighted content within the public's fair use protections, allow end users to make digital backups, and create a digital "first sale" doctrine. Most importantly, the bill would allow the public to circumvent any copyright protection technology (like DRM) for purposes that are traditionally exempted under copyright law. On March 24, 2003, Senator Ron Wyden (D-OR) introduced S. 692, the Digital Consumer Right to Know Act ("DCRKA") , which would grant the Federal Trade Commission ("FTC") the power to regulate labeling of CD's with any form of content protection or DRM technology limiting public use.
Representative Rick Boucher's (D-VA) earlier January 1, 2003 bill, H.R.107, Digital Media Consumers' Rights Act of 2003 ("DMCRA"), encompasses elements of both the Wyden and Lofgren bills. The DMCRA requires "prominent and plainly legible" notice to consumers on copy-protected CD's, although Boucher's provision is not as wide-ranging as Wyden's DCRKA. Similar to Lofgren's BALANCE Act, the DMCRA would create exceptions to the DMCA, allowing users to circumvent technology protection (like DRM) for noninfringing uses and for scientific research. The DMCRA would further allow the manufacture and distribution of tools "capable of enabling significant noninfringing use of a copyrighted work."
The technology industry is understandably worried about any legislative DRM mandates. Technology coalitions have been active in opposing the possible reintroduction of Sen. Holling's CBDTPA, which threatens the most far-reaching mandates. Jack Valenti and the motion picture industry continue to support the bill, standing at odds with the technology community.
The music industry, however, has been able to find a shaky middle ground with the technology industry. On January 14, 2003, the music industry and many technology groups agreed on a "rhetorical peace accord." The music industry promised to forego pursuing DRM legislative mandates in exchange for the technology industry's restraint in seeking anti-DRM legislation. Both groups agreed to pursue mutual technology and public solutions to digital content piracy, rather than resort to appeals to Washington. For the time being, it appears that DRM legislation is being held at bay by the pushes and pulls from various interested coalitions.
The present technological failures of DRM are forcing the technology industry to consider new approaches. Most DRM security is implemented to secure software, but not hardware. The underlying hardware and operating system allow an end user to access every bit of information in a digital file, even when protected by software DRM. With this low-level access, end users can attack the digital file itself, intercept digital information as a program executes (through an emulator or debugger), or access the end result (through screen or audio capture programs).
In December 1996, Bill Arbaugh, Dave Farber and Jonathan Smith published a paper entitled "A Secure and Reliable Bootstrap Architecture," describing the technical means to accomplish a secure hardware environment. The idea, dubbed "trusted computing" or "trustworthy computing," proposes adding several hardware components to computers to create greater security for encryption, storage and software.
In 1999, a collection of hardware and software companies formed the Trusted Computing Platform Alliance ("TCPA") with the goal of transforming trusted computing research into a workable architecture for the PC. The TCPA was created with the express goal of providing a single, common platform for trusted computing.
Despite some early contentions, TCPA does not actually control the execution of DRM programs. TCPA simply protects the integrity of digital files and cryptographic keys, and guarantees that the hardware and low level operating state are in a known, "trusted" state. Nonetheless, the TCPA trusted framework could enable developers to write very secure DRM programs. A TCPA system could verify that DRM programs remain unaltered. DRM developers could also utilize TCPA to seal and encrypt content files, ensuring access only by authorized means.
TCPA provides a framework in which a computer starts up securely, verifies its individual components, and can verify this trusted state to third parties. On boot up, a device attached to a PC's motherboard known as a Trusted Platform Module ("TPM") examines the PC's configuration. This chip checks that all of the hardware is TCPA-compliant, and if so, the chip allows the computer to run in a "trusted" mode. A TCPA system then checks out the entire computer, providing a report called an "attestation," which verifies the "current configuration of the platform." The "[k]nowledge and confirmation of the current software running on a system" provided by attestation may be relayed to remote third parties. DRM producers can use this attestation to guarantee that an end user's computer is a secure DRM environment. The TPM module also creates a unique "identity" for the TCPA system. This identity is sent to "Certification Authority" (CA's), also known as a "Trusted Third Party" (TPP), which generates a "certificate" for various uses by the end user. Thus, the TCPA scheme enables third parties to identify and attest to the DRM security of end user's computers.
In April of 2003, several key members of the TCPA pulled out of the organization in order to form a new entity, the Trusted Computing Group ("TCG"). The TCPA's organizational goal was to develop a common technical specification for trusted computing systems. By contrast, TCG is focused on becoming "a more formal group with licensing policy, a marketing budget, and a mission to push the trusted computing technology into a variety of devices."
Upon its creation, TCG adopted the TCPA's present technical specifications as its own. All further specification development will be pursued by the TCG. Consequently, the TCPA organization acknowledges that "TCPA has recognized TCG as the industry standard organization that will work on future trusted computing specifications."
The TCG architecture controls internal aspects of a computer that an average computer user would never see - functions like boot up, memory access and storage, and cryptographic functions. TCG offers secure computing functions to the operating system or software programs to use as they see fit. A TCG subsystem is agnostic as to what operating system runs on top of its TCG components. The first specification, TCPA version 1.1b (now TCG version 1.1b), only details the subsystem support for a trusted computer. The next version of the TCG specification, 1.2, will include guidelines for trusted operating systems. At that point, fully integrated TCG-enabled computers will become a reality. Technology companies are already forging ahead with the 1.1b specification, as well as preparing for the 1.2 specification. To date, IBM has released "driver" software to utilize TCPA components for the open source operating system, GNU/Linux ("Linux"), and several TCPA vendors are working on Linux TCPA Application Program Interfaces ("API's").
Despite belonging to both TCPA and TCG, Microsoft has its own vision for trusted systems. Microsoft is currently developing its own trusted computing system, the Next-Generation Secure Computing Base ("NGSCB") (formerly known as "Palladium"). NGSCB is not an actual implementation of a TCPA or TCG specification - it's more. NGSCB attempts create both a TCG-like hardware subsystem, as well as a trusted operating environment. NGSCB creates a "secure, parallel" operating system, known as the "Nexus," which runs alongside the Windows operating system. When an application requires "trusted" features, NGSCB verifies the state of the computer and performs trusted computing functions. Technically, NGSCB implements four main security features:
Strong process isolation. Users can wall off and hide pages of main memory so that each nexus-aware application can be assured that it is not modified or observed by any other application or even the operating system.
Sealed storage. Information can be stored in such a way that only the application from which data is saved (or a trusted designated application or entity) can open it. With sealed storage, a nexus-aware application or module can mandate that the information be accessible only to itself or to a set of other trusted components that can be identified in a cryptographically secure manner.
Secure path to and from the user. Secure channels allow data to move safely from the keyboard/mouse to nexus-aware applications, and for data to move from nexus-aware applications to a region of the screen.
Attestation. Users have the ability to authenticate software or a combination of software and hardware. With attestation, a piece of code can digitally sign or otherwise attest to a piece of data and thus assure the recipient that the data was constructed by an unforgeable, cryptographically identified trusted software stack.
The bulk of the NGSCB hardware design is "remarkably similar" to the original TCPA specification, utilizing a separate chip to control the trusted system, new encryption functions, and sealed memory. NGSCB hardware module is called a "Security Support Component" ("SSC"), which comprises the same basic functionality as TCPA's TPM. Beyond the current TCG specification, NGSCB introduces the idea of "Notarized Computing Agents" ("NCA's"). The NGSCB Nexus runs all of the NCA's for a given computer. NCA's are essentially subsets of larger software applications that are responsible for attesting that a computer is "safe" for a program and accessing secure memory storage.
The NGSCB system is developing rapidly. Microsoft demonstrated the preliminary system and technical details in May, 2003 at the Windows Hardware Engineering Conference. It is speculated that NGSCB may be released as a part of the Windows operating system in 2005, with other Windows DRM operating system components being released this year.
TCG provides the basic hardware framework for a trusted computer. However, trusted computing advocates "admit that without operating system support, they can only ensure a trusted state through boot-up." For DRM to utilize the full security of trusted computing, a secure framework must be available the entire time a computer is operating. Thus, regardless of whether NGSCB catches public favor, or if an operating system born out of TCG specification 1.2 becomes prevalent, trusted operating systems will become the final piece in implementing a secure environment for DRM.
NGSCB demonstrates that trusted systems are a forerunner to stronger DRM frameworks. Although Microsoft eschews that "NGSCB is not DRM," it admits that "DRM applications can be developed on systems that are built under the NGSCB architecture." In December, 2001, Microsoft received patents for a "Digital Rights Management Operating System," and for "Loading and Identifying a Digital Rights Management Operating System." It is unlikely that NGSCB presently will implement the DRM Operating System envisioned in Microsoft's patents, but like TCG, the NGSCB infrastructure makes DRM technologies more robust against cracking. Microsoft concludes that:
The operating system and hardware changes introduced by NGSCB offer a way to isolate applications (to avoid snooping and modification by other software) and store secrets for them while ensuring that only software trusted by the person granting access to the content or service has access to the enabling secrets. A DRM system can take advantage of this environment to help ensure that content is obtained and used only in accordance with a mutually understood set of rules.
Using a trusted computing environment, DRM developers can create programs that implement content restrictions on digital files. It is no surprise then that DRM developers are watching closely the development of TCG and trusted systems. Brian LaMacchia of Microsoft contends that the most critical needs of DRM systems are "trustworthy computing devices, robust trust management engines and a general-purpose rights expression/authorization language." Developers must then create systems (or "engines") which can determine which rights to grant users for digital files and which uses are restricted by the content owner. LaMacchia poses that the technology community should agree on a common language which programs will understand, so that content owners can designate DRM rights and restrictions once for every DRM system and trusted platform, rather than specifying authorization for every different program that could use a single digital file. Presently, the eXtensible rights Markup Language or ("XrML") is the forerunner with the Organization for the Advancement of Structured Information Standards ("OASIS"), the association that coordinates standards for XML languages.
Joe's simplified experience with a trusted computer running a DRM program is as follows: When Joe's TCG or NGSCB computer starts up, the trusted boot up device performs checks the hardware and software in Joe's computer. If Joe has changed nothing in his computer, then the trusted system will validate any "certificate" he has received from a CA (like Disney) and allow him to use his DRM software in the same manner as the last time he used his computer. If Joe has changed either the software or hardware, then the trusted system will detect this when it tries to access the certificate, and any certificates that relied on that particular component (hardware or software) will no longer work. If Joe replaced his TCG / NGSCB-compliant sound or video cards with ones that allowed him to record the sound or video output in digital form in an unrestricted manner, his trusted computer would attest to third parties that the sound or video cards had been changed. If Joe attempted to hack the code of his Disney-approved video viewing program, the changes to his viewing program would be detected by the trusted system, and Disney could invoke protections that prevented the program and digital files from any further use. Thus, if Joe's computer changes in any way that might threaten digital rights protection, Disney is empowered to stop Joe from viewing Snow White. Additionally, because the Snow White digital file was protected using trusted system encryption, Joe cannot use any other video program not explicitly authorized by Disney to view Snow White.
Now, if Joe wishes to view Snow White with his changed computer system, he must renegotiate with a CA (maybe Disney directly, or Microsoft, the company that produced the DRM program viewer) for a new certificate. If Joe's new system configuration checks out, the CA gives a new certificate and Joe may happily watch Snow and her companions. If Joe's new system is not trusted by the CA (at Disney's or Microsoft's behest), then Joe cannot watch Snow White with his new system configuration. Joe's only option is to change all of the hardware / software back to its original state in which it was initially approved. Thus, Disney has complete control over Joe's viewing because Disney can verify that Joe's hardware configuration and for software, Joe must use a Disney-authorized video viewing program - one that respects Disney's DRM commands for Joe's file.
The trusted computing initiative provides several critical components missing from DRM systems. The TCG and NGSCB architectures provide a secure computing base that takes care of the ever-pressing encryption and verification needs of DRM. Trusted computing engines provide content owners with a funnel to force end users to access DRM files only through authorized DRM programs. And common rights expression languages enable content owners to have an easy and efficient means of communicating DRM rights and restrictions to software programs.
Although trusted computing offers a giant leap forward for security for PC's, it will not be the final piece in the DRM-content puzzle. While this trusted computing enables far stronger DRM than currently available, the model is not likely to be secure against all attackers. In fact, it is likely that completely effective DRM is technologically impossible, both now and in the future.
Darko Kirovski, a security researcher at Microsoft, observes that "[e]very single device has to be secure ... [i]f one device is not secure, then [DRM] doesn't work." As discussed supra, in virtually every computer-based DRM system trotted out to the present, cracking one single digital file essentially renders the entire protection scheme insecure under Professor Felten's Napsterization threat model. If Joe finally gets fed up with Disney and successfully cracks his Snow White digital file, then he has individually broken the DRM system. If Joe can change the format of Snow White to something that plays on anyone's computer, then Disney's entire DRM protection has been compromised. A DRM system must therefore be BOBE-resistant (break-once, break-everywhere) to meet the Napsterization threat model. If a DRM system is unable to prevent BOBE-type attacks, then protected content will always be insecure. At this point, the best a DRM system can hope for is to look for a possible solution under the casual-copying model. However, the difficulty with using a near-perfect DRM model with existing technology is that popular digital content has the proclivity to spread rapidly worldwide.
Trusted computing is an incomplete answer to an issue that requires a complete solution. Content that has been encrypted and stored securely by a trusted computer can still be hacked. Microsoft NGSCB general manager John Manferdelli has acknowledged that the NGSCB does not secure content against "sophisticated hardware attacks." The TCG has similarly noted that "[i]t is not a goal of the TCG to enable or embed digital rights management (DRM) technology in computing platforms." David Safford of IBM's Global Security Analysis Lab, concludes that :
[T]he TCPA chip is not well suited to DRM tasks, and IBM's implementation of the chip was neither designed not [sic] evaluated for the necessary tamper resistance needed to provide effective copy protection (Personally, I do not believe it is possible do [sic] provide effective copy protection at all, but that's another paper).
Additionally, both TCG and NGSCB, as presently specified, will continue to allow the use of unprotected content. TCPA/TCG Specification 1.1 does not even provide an operating system. Because the TCG functions can only be called by a program or operating system, TCG itself cannot restrict unauthorized or unprotected content from use on an end user's computer. Microsoft contends untrusted digital files (like MP3's) are safe because NGSCB does not interact in any way with untrusted programs. Thus, TCG and Palladium will play hacked content, even if formerly protected by trusted computing. Consequently, trusted computing cannot guarantee effective DRM under a Napsterization threat model.
Moreover, technologies tend to bring unintended consequences and applications not previously conceived. In a strange twist of irony, a group of Harvard researchers have determined that trusted computing could be used to create a secure Peer-to-Peer file sharing system. This new secure system could be used by digital pirates to further circumvent copyright restrictions on digital works and protect their activities.
The security experts opine that these past and predicted future DRM failures are a systematic indication. By the time that end users compromise trusted computers to a large degree, the content industry will likely move on to the next, more secure technology base for DRM systems. However, as long as the public can play unprotected content, only one break is necessary to have a worldwide DRM failure. And, securing against this last, single break is difficult - in fact, many experts say it is fundamentally impossible as a technical matter. Cryptography expert Bruce Schneier of Counterpane Internet Security contends that:
Abstractly, [secure DRM] is an impossible task. All entertainment media on the Internet (like everything else on the Internet) is just bits: ones and zeros. Bits are inherently copyable, easily and repeatedly. If you have a digital file -- text, music, video, or whatever -- you can make as many copies of that file as you want, do whatever you want with the copies. This is a natural law of the digital world, and makes copying on the Internet different from copying Rolex watches or Louis Vuitton luggage.
And therein lies the problem. Whatever Disney does to prevent Joe from hacking Snow White doesn't really matter if Joe has the bits, the actual digital content, on his computer. Eventually, Joe (or someone) will crack those bits, because they are out of Disney's control. Even if encrypted. Even if in hardware. And when Joe eventually releases the unprotected, hacked version of Snow White on a file sharing network, the game is up. Snow White is unprotected and available for any member of the public to use without Disney's control. If present DRM cannot, as a technological matter, ever be secure then the content industry cannot rely on the Napsterization threat model to solve their digital content woes. Consequently, the content industry must look elsewhere to fill the technological gaps that cannot be solved or change their ambitions regarding copyrighted digital content.
Although the pure technical question of DRM may be intractable, the content industry also relies on market, political and legal forces to address the shortcomings of DRM. Through a combination of DRM technology and these forces, the content industry may be able to address the difficulty of the Napsterization threat model, or possibly find a viable scenario for profiting under a casual-copying model.
The content industry can hope that sheer market force of DRM products can force out older non-DRM technologies over time or that trusted computing implementations will one day restrict all unprotected content. Presently TCG and NGSCB allow an end user to run in "untrusted" mode and access protected files. But, if new, DRM-protected content formats become ubiquitous, then the fact that a few stragglers can access unprotected content wouldn't matter as much. Bruce Perens, an open source software advocate, predicts that once DRM technologies cross a certain threshold of popularity, the public will all fall in line because without the new, DRM-friendly technology "you are an island .. [y]ou can't communicate with others. Everyone will be using this DRM, and you can't view Web pages." The actual threshold for such a displacement is still up in the air. The public's love of unprotected digital content and file sharing networks raises doubts about changing this model. But, the dominance of large technology producers, like Microsoft, which could arguably force a DRM system on consumers, may be able to reign in a large amount of the public. Either way, the content and technology industries are walking a delicate line - effective DRM products must restrict unauthorized use, but DRM with too many restrictions will likely never gain enough popularity to displace unprotected digital technologies.
From the legislative perspective, Sen. Holling's CBDTPA would secure DRM systems against even the Napsterization threat model, at least on paper. Mandating built-in copyright protection mechanisms still wouldn't prevent the cracking of individual digital files. But, if all computers contained strict digital copyright controls, a single cracked file couldn't be effectively played by the vast majority of the world. And, given the CBDTPA's severe civil and criminal sanctions, it is unlikely that average computer user would tinker with their hardware to access unprotected digital content. However, the expansive scope of the bill that would solve the content industry's DRM problems necessarily creates such a burden on the technology industry and public that any serious push for the bill would be met with enormous dissent. Even without Holling's CBDTPA or any of the other current pro-DRM bills, the content industry has several existing legal resources from which it could possibly bolster developing DRM technologies against attempts to circumvent digital content protections and controls.
Copyright law strives to balance public access to works with creating incentives to produce by giving content owners a limited monopoly on copyrighted works. Presently, copyright law is at the center of a fierce debate between content owners, technology producers and the public over emerging digital technologies. Professor Jessica Litman observes that the "pressures put by new technology on the current copyright statute have sparked disputes over whether the current copyright statute can adjust to the climate of rapid technological change." Digital rights management technologies are inherently tied to copyright principles, as a technical means of enforcing legal rights of content owners.
The United States guarantees several exclusive rights to copyright owners including the rights of reproduction, preparation of derivative works, distribution, performance and display. The copyright owner subsequently has legal rights against parties who infringe on any of these rights, whether directly or indirectly.
However, copyright owners lack absolute control over their content. There are several exceptions to the basic exclusive rights of copyright law. The first sale doctrine allows a member of the public who has legally obtained a copyrighted work to resell the work without the copyright owner's authorization. The fair use doctrine, which is called by some observers as "the single most important set of legal principles" in copyright law, allows a person to disregard copyrights for "fair use" situations, like study and criticism. Fair use situations are often complicated due to the overwhelming disagreement on what constitutes a fair use. The codification of the fair use doctrine, 17 U.S.C. §107, purports to include such uses as "criticism, comment, news reporting, teaching ... , scholarship, or research" while leaving the ultimate fair use determination up to a number of factors.
In 1998, the U.S. Congress passed the Digital Millennium Copyright Act ("DMCA"), to bring copyright law up to date with modern digital technology. The DMCA created conditional immunity for internet service providers, added protections for "copyright management information," and addressed several other digital issues. Most importantly, the DMCA created an entirely new area of liability for the "circumvention of copyright protection systems". Section 1201(a)(1)(A), the individual access prohibition, prevents anyone from circumventing a technology that "controls access" to a copyrighted work. Section 1201(a)(2), the access circumvention tool prohibition, forbids the manufacture or distribution of any technology "primarily designed" to defeat the access controls to digital copyrighted content. Section 1201(b)(1), the copy circumvention tool prohibition, similarly forbids distribution of any technology "primarily designed" to defeat "copyright" controls to a digital copyrighted work. Essentially, DMCA anti-circumvention prevents any cracking of DRM technology that controls access to a work (individual access prohibition and access circumvention tool prohibition), and prohibits only the distribution of tools that crack DRM copy abilities. The only absent liability is that DMCA anti-circumvention permits individual cracks of DRM technology that controls copy abilities. Unfortunately, the line between what is an access protection and what is a copy protection is quite unclear at this time. The DMCA backs this broad new set of liabilities with traditional copyright remedies of damages, injunctive relief, special damages, and, in some cases, criminal penalties.
The DMCA contains some narrow anti-circumvention exceptions. Circumvention is permissible for reverse engineering where "program for the sole purpose of identifying and analyzing those elements of the program that are necessary to achieve interoperability" with other programs. Circumvention is also allowed in certain, specific situations for encryption research and security testing. However, professor Pamela Samuelson notes that these exemptions "are very narrowly drawn and fail to recognize many legitimate reasons for circumventing technical measures[.]"
The DMCA's anti-circumvention provisions have only been tested in only a handful of cases. The most significant anti-circumvention case to date arose over the encryption in DVD's. DVD's, which typically contain digital movies, are encrypted with a DRM technology known as Content Scramble System ("CSS"). CSS ostensibly prevents DVD's from being unencrypted on unauthorized DVD players. In September, 1999, Jon Johansen cracked the CSS system and released a program, DeCSS, which enabled unfettered access and decryption of DVD's. Johansen released DeCSS to the public, which quickly redistributed the program over the Internet. In November, 1999, Eric Corley posted the DeCSS program and links to other DeCSS copies on the website of the hacker magazine, 2600.
The eight major motion picture studios promptly sued Corley and 2600 in New York district court in Universal City Studios, Inc. v. Reimerdes ("Universal I"). The motion picture studios claimed that the distribution ran afoul of the DMCA's anti-circumvention provisions for distribution of access and copy control circumvention device. In addition to several constitutional claims, the defendants argued that DeCSS was exempted under the 17 U.S.C. §1201 statutory exceptions and fair use. The district court was not persuaded. Initially, the defendants claimed that DeCSS was protected reverse engineering, necessary to play DVD's on the Linux operating system (which, at that time, did not have a DVD player). The court denied the exception, noting that the defendants had failed to prove that "the 'sole' purpose of DeCSS was to create a Linux DVD player[]" because "DeCSS concededly was developed on and runs under Windows--a far more widely used operating system." The court similarly made short work of the defendants' encryption research and security testing exemption claims, holding the defendants' bad faith in posting the code forfeited both defenses. Finally, the defendants claimed that DeCSS was necessary for the public to exercise traditional fair use because the CSS encryption scheme uniformly prohibited all unauthorized conduct. However, the court responded that:
Access control measures such as CSS do involve some risk of preventing lawful as well as unlawful uses of copyrighted material. Congress, however, clearly faced up to and dealt with this question in enacting the DMCA.
The court concluded that although fair use may provide some defense for traditional infringement claims, Congress had clearly intended to entirely foreclose the fair use doctrine for 17 U.S.C. §1201(a) anti-circumvention claims. Accordingly, the district court found that the defendants had violated the DMCA and awarded the motion picture studios an injunction against Corley or 2600 posting the DeCSS program or links to it.
The defendants appealed to the Second Circuit in Universal City Studios, Inc. v. Corley ("Universal II"). Examining only free speech issues and a narrow fair use question, the Second Circuit affirmed all of the district court's findings. The Second Circuit upheld the constitutionality of the district court's fair use analysis, opining that the defendants had not made a credible fair use claim. The court concluded that "the Supreme Court has never held that fair use is constitutionally required" and moreover, that "[f]air use has never been held to be a guarantee of access to copyrighted material in order to copy it by the fair user's preferred technique or in the format of the original."
Much has been written on drastic implications of DRM for copyright law and privacy issues. Academics have theorized about what perfect DRM systems may entail and what they might imply for the public should they ever arrive. However, given the stunning failures of all previous wide-scale DRM attempts, the issue has been largely, well, academic. Until now. Trusted computing, although not complete, is real and coming soon. Within the next few years, trusted computing technology will make its way into the hardware of computers and electronic devices. Whether or not trusted computing was intended to promote DRM is simply a side issue for copyright law. Like other emerging technologies, trusted computing provides new capabilities that can be exploited by DRM producers and create new tensions within the legal system.
Trusted computing systems, as seen through the development of TCG and NGSCB, disaggregate the security and protection tools from the actual DRM products. Unlike CSS, where the DRM system uses specific encryption functionality to protect a DVD, trusted computing security schemes are largely implementation-agnostic. TCG will use the same encryption techniques and attestation for a web browser to keep a credit card secure for online e-commerce as Disney utilizes to keep Joe from unauthorized access of Snow White. There is no design or purpose in the encryption, sealing and attestation functions of trusted computing.
It is also important to consider how an end user might actually circumvent the security of a DRM program based on a trusted system. At the most basic level, a hacker could attempt to circumvent the trusted computing hardware subsystem. As described above, it is possible for experts to access individual digital files. However, an individual crack of a file on a trusted computer doesn't compromise the file on other computers, or other files on the original trusted computer. An attack on the whole system (like DeCSS) does not presently appear possible for trusted systems. And, a hack on the trusted subsystem would change some of the hardware or software values that would be used to attest the state of the computer to a third party.
Moving up a level, a hacker could attempt to write an interoperable program to use DRM-protected files without authorization or attack the DRM program. Jon Johansen's DeCSS successfully accessed the digital content of encrypted DVD digital files. However, with trusted computing, a DRM program and its files are protected by the trusted computing system, not the DRM program itself (like CSS). Thus, if the trusted computing scheme is effectively utilized by DRM producers, this method of attack ultimately defaults a hacker back to the initial (and quite difficult) option of hacking the trusted hardware system.
Of course, virtually anyone can still search for new, unprotected content on the Internet. Joe may not be able to access the TCG-sealed Snow White file on his computer, but he can easily use a Peer-to-Peer file sharing program to find a different copy of Snow White in an unprotected file, which plays without a Disney-approved media program. This scenario doesn't implicate DMCA §1201 because it deals with content not controlled by "a technological measure." Although a very important digital copyright issue, because trusted computing does not presently prohibit users from not using the trusted system or DRM programs, the copyright implications of this scenario will not be discussed further in this paper.
Disaggregation of trusted computing security from DRM restrictions poses challenges for DMCA §1201 liability. DMCA §1201(a)(1), individual anti-circumvention, likely applies to trusted systems in a similar fashion to present software-based DRM. A trusted computing security subsystem is plausibly a protected §1201(a)(1) "technological measure." And, a trusted system could restrict access to a copyrighted work, satisfying the remaining elements of a DMCA § 1201(a)(1) anti-circumvention claim. Whether Joe hacks his software-based DRM program for his Snow White DVD, or hacks his TCG-compliant hardware to get at a trusted version of the Snow White file, he will likely be liable under § 1201(a)(1).
However, defending against individual cracks with DMCA § 1201(a)(1) isn't the real concern of the content industry. Copyright owners are far more concerned with the widespread distribution of tools that enable all of a DRM protection scheme to be broken. The tools prohibited under DMCA §1201(a)(2) (distribution of access-circumvention tools) and DMCA § 1201(b)(1) (distribution of copy-circumvention tools), if distributed worldwide, pose a threat to the security of all digital rights schemes.
Although previous sections in this paper indicate that any attacks beyond individual file cracks are technologically infeasible, the legal implications of a system-wide attack are still worth exploring. Assuming a tool capable of defeating trusted computing could be created and distributed, the implications for DMCA liability under §1201(a)(2) and § 1201(b)(1) become more complicated. Both DMCA trafficking provisions create liability only where the device is primarily designed for circumvention, has limited commercial significance other than circumvention, or, is marketed as a circumvention device for cracking access or copy-protections. In Universal I, the district court found that CSS effectively controlled copyrighted work, and that DeCSS was primarily designed to circumvent this access control under §1201(a)(2)(A). The court felt that this fact was enough to prove limited commercial significance as well.
Trusted computing, by contrast, is not solely designed to protect a copyrighted work, as CSS does. Under one possible interpretation, distribution of a tool for generally cracking a trusted system would not necessarily create liability under DMCA §1201(a)(2) or § 1201(b)(1), because there is no inherent primary design for such a tool. Cracking the one trusted computing scheme likely means cracking everything. Thus, a tool "solely" designed to crack Joe's Snow White DRM file, would be the same tool as one that allowed a forgetful end user to recover their own encrypted files or passwords on their computer. The DMCA makes a basic assumption that a technological protection measure will only have the purpose of protecting copyrighted works. When a technological protection measure like trusted computing has a more universal application (i.e., not just DRM), the liability model runs into ambiguous ground.
An alternate interpretation of the DMCA distribution provisions would argue that since a trusted subsystem "effectively controls access to a work" (or copying), any device which circumvents this protection scheme is liable under DMCA §1201(a)(2) and § 1201(b)(1). The argument would be that the trusted system effectively protects access and/or copy abilities for DRM programs, whether the protection is disaggregated from a DRM program or not. A general trusted system circumvention tool would thus be liable for cracking the entire trusted system security because the whole trusted system effectively controlled access / copy abilities for a DRM program. However, as described above, one has to hack the whole trusted system or none of it because the same protections are used, regardless of function. DMCA liability in these scenarios could thus swallow the primary design requirement under §1201(a)(2)(A) and § 1201(b)(1)(A) because the disaggregated trusted computing security model is not built with any primary design and any tool capable of circumventing trusted system-wide security is also capable of circumventing a DRM protection scheme relying on the same trusted system. This scenario ultimately raises a troubling prospect where DMCA anti-circumvention liability could be further extended to tools possessing no DRM circumvention purposes.
Additionally, the implementation or marketing of a specific trusted computing circumvention tool could still implicate DMCA liability. If a developer released a program capable of cracking the entire trusted computing scheme, but which only actually implemented a specific DRM-cracking purpose, this would probably satisfy the primary design element of DMCA §1201(a)(2)(A) or DMCA §1201(b)(1)(A). And, a trusted computing circumvention tool (even if general in nature) marketed as a DRM-defeating device, would likely fall under DMCA §1201(a)(2)(C) or DMCA §1201(b)(1)(C), which creates liability for marketing a tool as a copyright circumvention device.
The DMCA purports to respect traditional fair use in Section 1201(c)(1), declaring that "[n]othing in this section shall affect rights, remedies, limitations, or defenses to copyright infringement, including fair use, under this title." Nonetheless, both the Universal I and Universal II courts upheld the proposition that fair use is not a defense to anti-circumvention liability for cracking access controls under the DMCA. Although possibly beyond the scope of Congress' intent for the DMCA, these are the only relevant rulings thus far. Applied to a trusted computing DRM scenario, the fair use defense would be completely unavailable for cracking a trusted system, if a court found the trusted system a protected technological measure. Because the DeCSS courts allowed no fair use for cracking access controls, and cracking a trusted system for any purpose (even completely academic or traditionally fair use) requires cracking the whole system, any trusted system circumvention tool would forego fair use defenses despite the optimistic statement in Section 1201(c)(1).
The DMCA contains several statutory exceptions. Considering the narrow language of the provisions, none are likely to create significant immunities for the dissection of a trusted computing system. Nonetheless, these provisions are likely to come to the forefront of the DRM debate once trusted computing systems become more prevalent.
DMCA §1201(f) provides a circumvention exemption for reverse engineering. Section 1201(f)(3) requires that the exempted circumvention be "solely for the purpose of enabling interoperability" with other programs. In Universal I, the district court held that DeCSS did not qualify for a reverse engineering exception because DeCSS did not "solely" enable interoperability of DVD's with the Linux operating system - the program also worked on Windows.
However, this exemption could prove important to software developers working with trusted computing systems. Because trusted computing can encrypt and seal files from any program with ease, software developers could use this functionality to ensure that only their programs could use their proprietary file formats. A common worry is that Microsoft will utilize trusted computing to ensure that Word documents can only be opened with the Microsoft Word. Microsoft has already announced that digital rights management will be built into its Office 2003 software, and will possibly thwart compatibility with other word processing programs and previous versions of Microsoft products. When trusted computing becomes available, such a scheme will be even more effective.
Trusted computing threatens to change the whole arena of reverse engineering. The open source community relies extensively on such reverse engineering of proprietary protocols and document formats in general to produce open source software. Critics already pose that the DMCA's restriction of acceptable reverse engineering is overbroad and gives too much legal protection to content owners. Trusted computing could augment content owners' technological protection of proprietary formats. To successfully create an interoperable product, the whole system must be cracked. If Company X wants to create a legal alternative viewing program for Joe's Snow White digital file, it must rely on being able to play all digital files from Disney, not just Joe's. Thus, if trusted computing security is resistant to such a system-wide attack as promised, then the legal status of reverse engineering is a needless investigation - reverse engineering will be technologically impossible.
If, on the other hand, the entire trusted computing system could be compromised, then DMCA anti-circumvention liability is still a possibility. To invoke Section 1201(f) as a defense to anti-circumvention liability for their work, reverse engineers would have to be very careful. The security to the trusted computing subsystem is disaggregated, but Section 1201(f)(3) requires that the only purpose of reverse engineering be interoperability with another program. Thus, at least under the Universal I analysis, an interoperable program would have to crack the whole trusted computing system, and then implement the narrowest set of functionality that would only enable an end user to interoperate with the program. Essentially, if Company X made a Linux viewing program for Joe's Snow White video, in order to escape anti-circumvention liability with DMCA §1201(f) they would have to ensure that program only worked on Linux and that it did not disable any of Disney's DRM controls.
Trusted computing systems must contain personally identifying information for attestation functions, so that third parties can determine which end user is requesting digital content. The use of digital personal information has already garnered a firestorm of controversy from privacy advocates. Keeping digital privacy in mind, DMCA §1201(i) allows individual end users to circumvent access controls solely to protect personally identifying information. However, the exemption is of very limited significance to cracking a trusted computing security scheme. The exception applies only to individual users and not to the distribution of circumvention tools. And, although individually cracking a trusted computing-protected file is possible, the end user is only allowed to remove personally identifying information under the DMCA §1201(i) exemption. Such a change to the file would almost certainly render the digital file untrusted by the system and it would fail in any remote attestation exchange. Joe could possibly crack his Snow White file and remove the identifier "Joe Smith" from the file, but doing so would make the file unplayable, and Disney would likely not re-approve the file for Joe's use without personally identifying him. Thus, although DMCA §1201(i) applies in theory, it would be of little practical use for privacy-minded members of the public.
The DMCA exceptions for security testing and encryption research give some leeway to researchers and security experts attempting to crack a trusted computing scheme. The DMCA's encryption research exemption, Section 1201(g), requires that the conduct be "necessary to conduct such encryption research" and for the purpose of disseminating information "to advance the state of knowledge" rather than providing a practical exploit. The security testing exemption, Section 1201(j), applies only to persons testing their own internal computer security and prohibits dissemination in any manner that facilitates copyright infringement. Essentially, these exemptions allow purely informational / academic dissection of a trusted computing security platform. But, the limitations on both exemptions prevent any workable circumvention of a trusted computing scheme to escape liability using Section 1201(g) or Section 1201(j).
The DMCA was Congress' answer to the demands placed upon copyright by the digital revolution. However, it appears that just years later, the rapid pace of technological development is straining even the newest legislation. Trusted computing security models challenge the DMCA because it wasn't designed with disaggregated security and protection measures in mind. To the further chagrin of DMCA critics, anti-circumvention liability could possibly expand under trusted computing-backed DRM schemes. Or, the general application of trusted computing could immunize some forms of circumventing conduct. Finally, differing possible interpretations of liability may be wholly unnecessary, for if trusted computing effectively locks out all system-level hacking, then the content owner's technology will trump the application of copyright law.
Evolving DRM schemes raise the specter of the "privatization" of copyright law by content owners, where technology and not the law governs copyrights. Although traditional copyright law has many exemptions and limitations, DRM with trusted computing offers the possibility of absolute control over the public's use of digital content. Content owners clearly have a vested interest in policing copyright infringement with technologies that they can control, in lieu of legal remedies that they often cannot.
Many critics argue that such a privatization by DRM threatens the public's copyright rights and freedoms. Robin Gross notes that
[b]y essentially privatizing copyright law, these DRM schemes maintain none of the balance that the public law of copyright enshrines. Ironically, society is embarking on a dangerous path of narrowing the public's access to creative expression at exactly the time that technological advances protect publishers' rights more effectively than ever before.
Professor Elizabeth Thornburg poses that absolute DRM schemes threaten to entirely control public copyright principles such as fair use, first sale rights, and making backup copies - each specific use would have to be licensed from the content owners. Thornburg contends that digital copyright controls are particularly problematic because there is no external, human check on the restrictions of DRM systems. Essentially, DRM systems create the equivalent of a permanent injunction against traditional copyright exemptions with significant procedural biases in favor of the content industry. Professor Lawrence Lessig questions the public's recourse under such a scenario and postulates that
the problems are worse when code displaces copyright law. Again -- where do we challenge code? When the software protects in a particular way without relying in the end on the state, where can we challenge the nature of the protection? Where can we demand balance when the code takes it away?
Because of the fallibility of existing software-based DRM schemes, the public has yet to grapple with any serious DRM impediments. Trusted computing may well take a large role in changing this. The low level security schemes inherent in trusted computing will present very effective protection mechanisms for DRM programs, at least for the near term. And, there are no technological mechanisms for fair use or other copyright limitations built into current trusted computing specifications. Content owners and DRM producers will likely be able to specify exactly what the public can or cannot do with their content. When trusted computers actually ship and DRM producers write programs on top of trusted platforms, technology will possibly hand over all control of digital copyright law from the legal system to the content industry.
The new level of DRM security possible with trusted computing is not likely to last over the long term. Trusted systems don't address the fundamental problem posed by security experts - one cannot ultimately control bits of information on the public's computers. As previously noted, trusted systems explicitly do not protect against specific hardware attacks. Thus, the Napsterization threat model remains open and trusted computing will fail as a security framework for DRM because of development of system-wide exploits or the distribution of single, cracked files over the Internet. The content industry will then have to look to a new technology that promises to finally, finally lock down all of the bits on a user's system. However, if Schneier and his compatriots are correct, even this next technology won't be completely secure because it is impossible to secure all computers, and the cycle will continue.
Most likely, the public will be left in an "arms race" scenario, where the content industry continually pursues new technologies for DRM systems that clever hackers will inevitably manage to break, preventing a completely secure system. Professors Pamela Samuelson and Suzanne Scotchmer note that such a "measures-and-countermeasures war" is likely to be expensive and wasteful. They also observe that such a "war" is driven in part by DMCA anti-circumvention liability, which gives significant advantages to the content industry. Nonetheless, despite the specter of DMCA liability and improved technology, hackers and reverse engineers appear determined to release circumvention tools.
The public stands to lose both traditional copyright freedoms and improved technological functionality in such a conflict scenario. Most of the public will likely to be unable to circumvent trusted computing-based DRM. And, they will be less likely to circumvent the subsequent DRM systems. However, as the content industry and DRM developers tirelessly work on thwarting the hacker elite, resulting DRM systems will become more and more restrictive. Mark Cooper, research director for Consumer Federation of America observes that DRM "isn't going to stop serious hackers" - "[a]ll you end up with here is an inconvenience to the average consumer." Trusted systems will no longer allow the public access to their own digital files. Perhaps this is an acceptable cost, given the personal security benefits of trusted computing. But, a substantial issue remains - what will the next generation of DRM systems force the public to give up?
The content industry cannot solve the Napsterization threat model by solely relying on trusted computing to stop copyright infringement. Already, there are signs that perhaps the content industry is ready to embrace a casual-copying threat model, focusing only on keeping the majority of the public honest. Apple Computer's iTunes Music Store appears to be a step in this direction. Apple's downloadable music files are protected by DRM that allows end users wide latitude in the copying and use of digital files. The combination of reasonable pricing ($1 per song) and a modest DRM scheme appears popular - Apple has sold over 3 million songs since the Music Store's launch.
In the alternate, the most logical push by the content industry in the future will be to contain the damage of single file cracks by preventing their use on other machines. However, for that to happen, virtually all unprotected digital content must be regulated - an extraordinarily substantial task. This would most likely mean technological or legal mandates.
If trusted computing becomes technologically ubiquitous, a DRM system can be superimposed on the framework. Technology companies could then pursue DRM operating systems (like that in Microsoft's DRM OS patent) that freeze out any unprotected content. If enough people can be enticed to purchase computers that technologically restrict any unauthorized digital copyright use, then perhaps DRM does have a chance at solving the break-once, break-everywhere problem. Unfortunately, it remains rather difficult to convince the public to buy into a technology that substantially restricts functionality and freedom.
On the legal side, Senator Holling's CBDTPA would mandate that effective DRM be built in to every new computing device. Such a law would likely solve break-once, break-everywhere problem of DRM because each individual computer would only play authorized content. Each CBDTPA computer would need to be individually cracked to play unauthorized content. But requiring that every computer restrict a wide range of digital conduct is an enormous leap from the more modest fact of trusted computing offering tools to implement DRM functionality. Thus, for the time being, it appears that the technology industry, the civil libertarians and political opponents of DRM mandates will keep such legislation at bay.
Many parties have also offered solutions to sway the balance of power away from the content owners. The Boucher, Wyden and Lofgren bills all attempt to chisel away at the legal side of content industry's DRM campaign. Hackers, like Jon Johansen, chip away at the technology each new DRM protection scheme. From both ends, Deidre Mulligan and Aaron Burstein argue that perhaps copyright limitations can be built in to DRM languages and implementations. Dan Burk and Julie Cohen suggest that additional technology and/or regulation can regain traditional fair use grounds under developing rights management schemes.
And, perhaps an "arms race" in trusted computing or DRM development isn't such a bad thing after all. The astonishing rate of development of technologies is likely due in large part to the various parties, problems, incentives and pressures surrounding the changing technological arena. Adam Thierer, an analyst for the Cato Institute, poses that for DRM:
[t]he better alternative to federal mandates on either side of this debate is to instead just encourage a technological free-for-all in the marketplace[.] Let the industry do whatever it wants in an attempt to bottle up their content, but also let consumers continue to experiment with and use digital content in creative ways without fears of federal intervention at every turn... There's no reason for Congress to intervene in an attempt to solve each and every intellectual property dispute, as has seemingly becoming the case in recent years.
Conclusion
Trusted computing is a complex technical subject as well as a complicated policy issue for digital copyright law. Digital rights management will continually be pursued by the content industry in the attempt to plug the holes created by Napster and file sharing on the Internet. Trusted computing offers a new level of security for computers with both negative and positive implications for the public. Developers point out that trusted computing systems are agnostic towards DRM. But the DRM implementations built on trusted systems will not be. Trusted computing will provide a building block for a new generation of secure DRM technologies, but is not the ultimate problem nor final solution for the DRM debate. At least not yet.
For the near term, the new model of disaggregating security features from DRM functionality creates issues for the inevitable DMCA anti-circumvention debate once hackers set in to defeat trusted computing devices. Trusted computing will provide the means for another step towards that privatization of digital copyright law. And, the range of the public's traditional copyright freedoms will incrementally narrow, with expanding legal doctrines on one side, and stronger digital rights locks on the other.
Over time, DRM based on trusted computing cannot lock down all computers without the help from either legal or economic forces. The content industry will continue to push for stronger legal and technological DRM measures, which hackers will attempt to break, while civil libertarians voice their concerns. The technology community is still split over the content industry's demand for secure DRM technologies and the public's insatiable appetite for devices without such controls. And, predictably, the politicians are listening and responsive to all parties on the issue. But, from any perspective, trusted computing demonstrates that as the stakes are raised in the DRM controversy, the public risks being left in the middle of an ever-escalating war for copyright control on their computers.
