The war on copying

The communications industry is ready for an infusion of data, such as digital video, to drive it to recovery, but music, video, and other digital-content owners continue to keep a tight rein on their growing mass of IP (intellectual property) while waiting for a secure DRM (digital-rights-management) scheme to materialize. The complexity of DRM, however, makes it a nontrivial addition to a system, especially a consumer device with a low cost threshold.

Several standards are under development to define how to encrypt material and distribute it over public networks. Rather than define the policies themselves, they define a foundation framework that can support a variety of DRM policies. For example, the ISMA (Internet Streaming Media Alliance) 1.0 Encryption and Authentication spec, scheduled for approval this month, describes how to apply the Advanced Encryption Standard (AES) to content and packetize content for distribution in such a way as to prevent a late packet from disrupting the stream.

Intentionally, these specs do not define what keys you should use or how to manage particular rights. One reason for this omission is that the industry is unclear on how you should manage keys or express rights. Key management and expression of rights, however, is where DRM gets the most complex.

The TCP (Trusted Computing Platform) from the TCG (Trusted Computing Group) addresses key management by specifying a trusted module that applications can use to protect content. The module is actually a processing subsystem; all encryption and decryption happens on the module, so keys are never in the clear. However, to decrypt a 2-Mbyte/sec video stream, the module needs significant processing ability. The $4.25 (1 million) AT97SC3201 TPM from Atmel, for example, performs a 2048-bit RSA sign in 500 msec.

Support for TCP should appear in Microsoft's next version of Windows (formerly known by the code name "Palladium"). However, Microsoft has made public that it intends to introduce changes that will make the operating system incompatible with chips that follow the current version of the TCP spec.

Several proprietary DRM schemes are under development. Sony and Philips, among a plethora of hopefuls, have their own architectures. However, the lack of a consistent model for how to pay for content and use it will easily confuse users. Different business models that may restrict use to a single person or device or a specific time period will constrain the use of similar content, such as music files. Such flexibility is great for content owners but requires users to interact differently with every piece of content. Confusion will arise when users think they are buying one form of license and getting another. Buying and using content needs to be as easy as saying "I want it" and clicking on a button. It shouldn't mean deciding how you want it and having to read pages of small print to understand what you are buying.

Most companies mistakenly believe that content protection is about protecting content. Consider that renting a new video release for a single night costs $4 to $5, but renting an old video for five days costs $5. Most content makes the majority of its revenue in the first few weeks of release. Thus, content protection is really about protecting the release window.

Many companies mistakenly focus on the technology when trying to understand DRM and fail to consider the real social issues that managing content involves. For example, DRM schemes that tie content to a single PC fail to address the needs of, say, a child of divorced parents who lives in two homes. Even more common is the person who wants to play music at home, at work, in the car, on a portable player, and at a friend's house. The killer app for digital content is the connected home, yet most DRM schemes undermine consumers' ability to easily move content between devices. Protection isn't just about security; you need to consider convenience, as well.

It is in these real-world individual issues that DRM will meet the most resistance. Content owners want strict definitions of stealing and honesty. However, schemes that are too difficult to work with or appear insensitive to individual circumstances may drive "honest" people to reconsider what honesty really means.

A popular protection mechanism is to tie content to a hardware "dongle," trusted module, or system component that cannot be copied. Traditional dongles hold all keys to content (or the master key that protects all other keys). You need the dongle to access content, and you can tie only one dongle to any particular content, a fairly limiting prospect. Consider that if you lose or misplace the dongle, you lose access to all the content it protects. Also, you can access content from only one location at a time; in other words, you can't listen to a song on your MP3 player while your spouse uses the stereo to listen to a different song tied to the same dongle.

A two-tiered scheme allows more flexibility. A master security dongle, perhaps on the motherboard of a PC, holds the keys to content. To access content, you use an individual dongle. The master can either pass keys to the individual dongle, which introduces vulnerabilities, or decrypt and then re-encrypt the content to tie it to the individual dongle. The advantage of this system is that several users can access the same content. However, you need to have mechanisms in place, such a limit on the number of dongles supported, to prevent users from re-encrypting content to a friend's dongle.

Flash modules are currently a popular method for transporting content. Several schemes tie content to individual flash modules. However, because an individual can have multiple modules, it's not unreasonable to consider that friends might share content by exchanging flash modules. Thus, tying content to the flash alone is insufficient.

Requiring users to occasionally "check in" over the Internet to authenticate access to content addresses the issue of content running on an approved list of registered devices. If a user wants to transfer rights to another device, the old device can be unregistered, and access to content will fail after the next scheduled check-in. Scheduling check-ins requires diplomacy, because users may perceive that you are tracking their usage. Alternatively, a device may not have access to the Internet each time a user wants to access content, such as when a user watches a movie on a plane. You need to be able to postpone check-in by a reasonable window.

You also need a plan to account for the failure of a tied component. Users may tie personal data to the component that they need to be able to recover. You can't have a backdoor in the scheme, because others can compromise it. Thus, you need some way to register the material a user owns or to back up keys. Backing up keys is not a preferred mechanism, however, because it requires you to take the keys off the secure dongle. A rogue entity's authenticating itself to the secure dongle compromises the keys. Additionally, you have to protect against the accidental destruction of keys. If it is possible to erase the module via software, then a virus could trigger an erasure and wreak havoc.

Many users don't back up their computers. Therefore, some mechanism must be in place to recover licenses after a system crash. One strategy dictates that the user has to contact the site from which they purchased content, but having to contact a multiplicity of vendors about a variety of content could be a herculean task. Consider also that when you tie content to a specific component, this component will not be present when the user upgrades equipment, forcing the user to download and rekey all content.

A robust DRM scheme needs to automate the transfer of rights between devices. For example, Turbo Tax 2002 offered a flexible license that allowed you to install the software on multiple machines, so that you could work on your taxes from any machine in the house, but only the designated primary machine supported the key functions of printing and electronic submission. If you wanted to designate another machine, you had to make a phone call. This process was inconvenient for users and ate into profits.

Some vendors have suggested a clearinghouse that manages all of the rights and licenses an individual has purchased. If the computer crashes or the user upgrades, the clearinghouse manages the rights-transfer process. Additionally, with a clearinghouse, you can use a portable component, such as a smart card, to exercise rights at a temporary site, such as a friend's computer.

A simple means of defeating rule-based protection is patching. A user who patches the application so that it ignores a flag or dongle defeats a do-not-copy flag or dongle. Tying the rules directly into the content is more effective but more complex (see sidebar "Forensics").

An improved defense is to execute critical-function code from a protected domain. You can execute code on a back-end server, which requires a network connection, or a trusted coprocessor, such as a smart card. For example, Sospita and Schlumberger have teamed up to allow you to execute critical code on a smart card. Note, you can't run high-performance code; response time for a first execution can take around 270 msec or as little as 8 msec if the function is cached on the card. Card readers might be pricey for some applications: They cost $15 to $25 each, depending upon volume.

Trying to prevent every instance of copying will probably require schemes so complex that users will be unwilling to use them (see sidebar "Protecting content"). Even if you can create a secure scheme, managing all the special cases, including lost tokens, legitimate transfer of keys, and others, may eat away more profit than those cases protect. Consider supporting a reasonable level of protection. For example, allowing users to install software twice lets a user to create a backup, transfer to a new computer, quickly handle the loss of a dongle, and so on. You can avoid most service calls with this one strategy. Certainly, two friends might copy the software, but compare that lost revenue with the cost of managing legitimate transfers.

Protected content yields a questionable benefit. You can't share content, access it anywhere you like or resell it after you are finished with it. Additionally, understanding DRM licenses and dealing with restrictive use are annoying issues. If no perceivable benefit exists, then users have no incentive for cooperating with DRM.

OEMs face the issue of having to allocate more development resources and equipment budget to DRM functions, resulting in a more expensive device that is more complex to use but provides no extra value to the consumer to offset the cost. Compound this expense with devices needing to support several DRM schemes.

In other words, the users and OEMs, not the content owners, bear the cost burden of DRM. Neither users nor OEMs gain any real benefit from cooperating to create secure DRM; in fact, it's in their best interests if such schemes never succeed.

Movie and record labels also seem to be counting on legal instead of technical protection. The DMCA (Digital Millennium Copyright Act) of 1998 criminalizes any action that breaks copy protection, regardless of the reason for it. For example, if you copy a movie to your laptop hard drive so that you don't have to run the DVD drive on your battery, you have committed a crime, even though there have been no real damages or lost revenue.

Recently, the US Supreme Court ruled on DVD protection, stating that publishing trade secrets circumvents protection schemes as covered under the DMCA. If they take this idea to the extreme, vendors could protect content using any "trade secret," even one as simple as shifting bits by one place. Traditionally, trade secrets were valuable only in the sense that they were secret. Now, the law protects them. And content owners are pushing for laws that allow prosecution of individuals. Effectively, the government and users will bear the burden of protecting content through the court system.

Content owners fail to see that it simply doesn't make financial sense to take someone to court for "stealing" $20 of content. Just imagine the RIAA (Recording Industry Association of America) asking district attorneys to fill courtrooms with 13-year-old "criminals." In the end, legal means alone will not protect content. If enough people ignore the legal barriers, the system will collapse under its own weight. The only effective protection scheme is one that actually prevents action.

The truth is that DRM is not for the benefit or protection of users, no matter what content owners or standards groups say. For DRM to succeed, content owners must take responsibility and bear the burden of its implementation. They must recognize that users already expect certain capabilities, such as the ability to use purchased content as and where they like. Owners must also recognize that people will accept a definition of "honest" only if it doesn't unreasonably inconvenience them.

Both sides—owners and users—must benefit in real ways for DRM to take hold. DRM must be seamless, nonintrusive, and painless if content owners expect users to cooperate. The main challenge for DRM is that an efficient system, at least from a user and an OEM perspective, is in place. DRM's big promise is to make listening to music or watching a movie a more complex endeavor.

It's easy to lose sight of the true goal of copy protection. It is not about preventing copying. In most cases, a perfect digital copy is unnecessary; many DVD-copying applications make good-enough copies—copies that users can't tell from the originals—from analog outputs. Effective copy protection, rather, is about making distribution of pirated material difficult enough that you can turn most nonpaying pirates into paying users.

Of course, you can always try charging a reasonable price and trusting people to be honest. Just think of all the money you'll save not having to implement DRM.

Author Information

You can reach Contributing Technical Editor Nicholas Cravotta at editor@nicholascravotta.com.