Watermarking and Fingerprinting

"Watermarking and fingerprinting are two forms of technology known generically as content identification. Watermarking works by embedding data into digital images, audio, or video in such a way that the data is very difficult to remove and the effect on a user's perception of the content is (usually) nonexistent. The data embedded in a watermark is often the identity of the content, though it could also include the identity of a user or device that downloaded it, or of a retailer that sold it.

Fingerprinting is a set of techniques for analyzing content, reducing its unique characteristics to a set of one or more numbers that serve as "fingerprints," and looking those fingerprints up in a database to determine the identity of the content.

Interest in both of these techniques has been growing rapidly in recent months. They are passive, meaning that their use in identifying and tracking content do not (by themselves) interfere with a user's ability to play, copy, or send it. They complement or substitute for active content control techniques such as the encryption that is used in typical DRM technologies.

Practical applications of digital watermarking for tracking content usage have been in existence for roughly a decade, but during the first Internet bubble, watermarking vendors oversold the content industry on the technology as a panacea for Internet piracy. This resulted in a backlash against a set of techniques that, in retrospect, were fairly basic. But watermarking techniques have become much more sophisticated and useful recently, and a wider variety of participants in the content value chain have gotten involved. Many vendors are involved in the watermarking arena, including Digimarc, Philips, Thomson, Cinea, Verimatrix, Activated Content, USVO, and Bitmunk.

Fingerprinting is a more recent technology; it came about in the 2000-2001 timeframe and was proposed as a way to make the original Napster P2P network copyright-compliant. Now there are a handful of music services (e.g., iMesh/BearShare) that use audio fingerprinting, most typically to block uploads of copyrighted music tracks to P2P networks. Such networks are often licensed by the major music companies, indicating their increasing comfort level with the technology -- although no one believes that it works one hundred percent.

Audible Magic and Gracenote are two of the leading audio fingerprinting technology vendors. SNOCAP uses fingerprinting (from Gracenote) to power services like its ad-driven model with imeem. Fingerprinting is capable of supporting wide ranges of innovative content business models; as with watermarking, the surface has barely been scratched, and we'll see some very interesting developments in the near future.

More recent fingerprinting solutions focus on video content, which is more technologically challenging than audio. As we saw last week, Google unveiled a video fingerprinting scheme, which is turning out to be less sophisticated than third-party technologies that are being developed by vendors such as Audible Magic, Philips, Vobile, Zeitera, and others. Attributor has a variation on this theme: a fingerprinting scheme for text content, which is being used by some of the major news wire services to track placement (both licensed and unlicensed) of their news content on various websites.

Watermarking and fingerprinting are distinct yet synergistic technologies. Their importance in the world of digital content rights is growing rapidly; in time, they may become more important than encryption-based DRM technology in certain media market segments."

By Bill Rosenblatt, DRM Watch