Policing Hosts

If there are subverted hosts, then content will leak into the darknet.  If the darknet is efficient, then content will be rapidly propagated to all interested peers.  In the light of this, technologists are looking for alternative protection schemes.  In this section we will evaluate watermarking and fingerprinting technologies.

Watermarking

Watermarking embeds an “indelible” invisible mark in content.  A plethora of schemes exist for audio/video and still image content and computer programs. 

There are a variety of schemes for exploiting watermarks for content-protection.  Consider a rendering device that locates and interprets watermarks.  If a watermark is found then special action is taken.  Two common actions are:

1)            Restrict behavior:  For example, a bus-adapter may refuse to pass content that has the “copy once” and “already copied once” bits set.

2)            Require a license to play:  For example, if a watermark is found indicating that content is rights-restricted then the renderer may demand a license indicating that the user is authorized to play the content.

Such systems were proposed for audio content – for example the secure digital music initiative (SDMI), and are under consideration for video by the copy-protection technical working group (CPTWG) .

There are several reasons why it appears unlikely that such systems will ever become an effective anti-piracy technology.  From a commercial point of view, building a watermark detector into a device renders it strictly less useful for consumers than a competing product that does not. This argues that watermarking schemes are unlikely to be widely deployed, unless mandated by legislation. The recently proposed Hollings bill is a step along these lines.

We contrast watermark-based policing with classical DRM:  If a general-purpose device is equipped with a classical DRM-system, it can play all content acquired from the darknet, and have access to new content acquired through the DRM-channel.  This is in stark distinction to reduction of functionality inherent in watermark-based policing.

Even if watermarking systems were mandated, this approach is likely to fail due to a variety of technical inadequacies. The first inadequacy concerns the robustness of the embedding layer. We are not aware of systems for which simple data transformations cannot strip the mark or make it unreadable.  Marks can be made more robust, but in order to recover marks after adversarial manipulation, the reader must typically search a large phase space, and this quickly becomes untenable. In spite of the proliferation of proposed watermarking schemes, it remains doubtful whether robust embedding layers for the relevant content types can be found.

A second inadequacy lies in unrealistic assumptions about key management. Most watermarking schemes require widely deployed cryptographic keys. Standard watermarking schemes are based on the normal cryptographic principles of a public algorithm and secret keys. Most schemes use a shared-key between marker and detector. In practice, this means that all detectors need a private key, and, typically, share a single private key.  It would be naïve to assume that these keys will remain secret for long in an adversarial environment.  Once the key or keys are compromised, the darknet will propagate them efficiently, and the scheme collapses. There have been proposals for public-key watermarking systems. However, so far, this work does not seem practical and the corresponding schemes do not even begin to approach the robustness of the cryptographic systems whose name they borrow.

A final consideration bears on the location of mandatory watermark detectors in client devices. On open computing devices (e.g. personal computers), these detectors could, in principle, be placed in software or in hardware. Placing detectors in software would be largely meaningless, as circumvention of the detector would be as simple as replacing it by a different piece of software. This includes detectors placed in the operating system, all of whose components can be easily replaced, modified and propagated over the darknet.

Alternatively, the detectors could be placed in hardware (e.g. audio and video cards). In the presence of the problems described this would lead to untenable renewability problems --- the hardware would be ineffective within days of deployment. Consumers, on the other hand, expect the hardware to remain in use for many years.  Finally, consumers themselves are likely to rebel against “footing the bill” for these ineffective content protection systems.  It is virtually certain, that the darknet would be filled with a continuous supply of watermark removal tools, based on compromised keys and weaknesses in the embedding layer. Attempts to force the public to “update” their hardware would not only be intrusive, but impractical.

In summary, attempts to mandate content protection systems based on watermark detection at the consumer’s machine suffer from commercial drawbacks and severe technical deficiencies. These schemes, which aim to provide content protection beyond DRM by attacking the darknet, are rendered entirely ineffective by the presence of even a moderately functional darknet.

Fingerprinting

Fingerprint schemes are based on similar technologies and concepts to watermarking schemes.  However, whereas watermarking is designed to perform a-priori policing, fingerprinting is designed to provide a-posteriori forensics. 

In the simplest case, fingerprinting is used for individual-sale content (as opposed to super-distribution or broadcast – although it can be applied there with some additional assumptions).  When a client purchases an object, the supplier marks it with an individualized mark that identifies the purchaser.  The purchaser is free to use the content, but if it appears on a darknet, a policeman can identify the source of the content and the offender can be prosecuted.

Fingerprinting suffers from fewer technical problems than watermarking.  The main advantage is that no widespread key-distribution is needed – a publisher can use whatever secret or proprietary fingerprinting technology they choose, and is entirely responsible for the management of their own keys.

Fingerprinting has one problem that is not found in watermarking.  Since each fingerprinted copy of a piece of media is different, if a user can obtain several different copies, he can launch collusion attacks (e.g. averaging).  In general, such attacks are very damaging to the fingerprint payload.

It remains to be seen whether fingerprinting will act as a deterrent to theft.  There is currently no legal precedent for media fingerprints being evidence of crime, and this case will probably be hard to make – after all, detection is a statistical process with false positives, and plenty of opportunity for deniability.  However, we anticipate that there will be uneasiness in sharing a piece of content that may contain a person’s identity, and that ultimately leaves that person’s control. 

Note also that with widely distributed watermarking detectors, it is easy to see whether you have successfully removed a watermark.  There is no such assurance for determining whether a fingerprint has been successfully removed from an object because users are not necessarily knowledgeable about the fingerprint scheme or schemes in use.  However, if it turns out that the deterrence of fingerprinting is small (i.e. everyone shares their media regardless of the presence of marks), there is probably no reasonable legal response.  Finally, distribution schemes in which objects must be individualized will be expensive.