Marked Music

Watermarking may hold key to sharing copyrighted material

To many, the fight between the record companies and the Napster-loving, music-swapping public looks like a battle to the death. While these two sides fight it out in court, another group of technologists is developing a compromise known as the digital watermark. If it works as planned, rights holders will be able to collect their royalties and consumers will be able to move their music relatively freely from machine to machine.

The aim is to imitate the way a watermark in paper can be used to add information about the origin and authenticity of the document. In this case, practitioners hope to inscribe a hard-to-spot copyright notice in every music file that says something like, "This performance copyright by Running Dog Records." Recording devices would be able to pick up this notice and inform users about the copyright, no matter how many times the music had been moved around the Internet.

Scott Moskowitz, chief executive of watermark technology company Blue Spike, explains that the goal is to design networks and music players to look for watermarks and determine whether copies are legitimate.

"You can say to Napster: Check this keyed watermark at the door. If it doesnt check out, you should reject it, " he says.

Digital watermarks work by introducing subtle changes to a music file in a predictable way. They might be as simple as a tone of a particular frequency repeated every so often, or as complicated as one that "bends" music in subtle ways. One solution would digitally simulate the acoustics of the location where the music was recorded and change them subtly in a regular pattern.

Watermarks can include varying amounts of information. Some may simply encode a message conveying that the file is copyrighted. More sophisticated versions can include contact information for the copyright holder. And it may be possible to customize the watermark to track down the original owner of a pirated file.

Blue Spike, for example, removes a few select tones in a very narrow band. Verance adds signals that are just out of the human perceptual range. Others try to bend the sound by changing the frequency slightly.

The art is in trying to make these changes as small and imperceptible as possible. Record companies and artists can be notoriously picky about the solution, and the products are tested for distortion.

"If you listen to Abbas music, you can hear people talk in the background," says Paul Kocher, founder and president of Cryptography Research. "But a lot of musicians, like Pink Floyd, have very specific ideas about which equipment produces the best sound. The artists are going to dislike technologies that introduce artifacts or the perception of artifacts."

War of the Techies

Watermark developers are fighting a battle against compression programmers, who try to shrink music files by excluding unwanted information. Compression software is essential for shipping music and other data across the Web, but this feature could turn it into an unintended tool for stripping out watermarks. The extra tones are often stripped out because the algorithm recognizes that theyre not likely to be heard by humans.

If this happens, it may be possible to turn an official recording sealed with the record companys watermark into a watermark-free home recording with no protection.

The Secure Digital Music Initiative (SDMI) aims to set a common standard acceptable to computer, software, entertainment and electronic goods companies. Its first-generation watermark simply tries to identify copyrighted music to computers and MP3 players.

"We do not carry that much information in the watermark," says Leonardo Chiariglione, the SDMIs executive director. "At the moment, the phase-one screening watermark carries how many copies you are allowed to make. Theres no indication of the author, the record company, etc."

Much of the debate has raged over whether it is possible to build watermarks that can withstand either casual compression by home users or dedicated assaults by the most competent programmers.

One solution would embed a watermark that indicated the song was protected by the latest tools. It would be, in essence, a one-bit message. Added to it would be a second watermark — a more fragile version that would be destroyed by any compression algorithm. If the first watermark were found without the second, then the file would be considered pirated.

Recently, the SDMI sponsored a hacking challenge, inviting the public to poke and prod some of the latest technology. The contest was controversial because it required entrants to sign away the intellectual property rights to their solutions and swear secrecy if they wanted to win part of the $10,000 prize.

Some participants were able to remove some of the watermarks. Two of the watermarking systems, one from Blue Spike and the other from Verance, were not considered broken, although there is debate about whether the measurement is subjective. A system was only considered broken if the watermark on a file could be repeatedly removed without ruining the sound quality of the file.

Secrets, Lies and Audiotape

The secrecy of the contest is still controversial. But David Leibowitz, chairman at Verance and president of its Digital Commerce Group, defends the practice. "With our technology, as with most security-type systems, you want to maintain its operational details in confidence in order to minimize the possibility of security breaches," Leibowitz says.

But others wonder whether a technology can really succeed if it needs to rely upon legal threats to corral secret information. Many feel that the industry is pinning its hopes on the controversial Digital Millennium Copyright Act (DMCA), which criminalizes circumventing copyright protection mechanisms.

However, academics and free-speech advocates are challenging that law as a violation of the First Amendment because it prevents security researchers from discussing the state of the art. The law and the stiff rules of the contest led many researchers to openly call for a contest boycott. Now many wonder if the best thinkers stayed away.

Despite these controversies, many in the industry still see great promise for watermarks. Some imagine sound files that can be played a few times as a free trial, but must be purchased if theyre going to be played repeatedly. Others imagine music that will get cheaper with time, perhaps becoming free after a few years. People wanting to hear the newest music would have to pay a premium, while those willing to wait would get a discount.

Leibowitz feels that the technology being developed at Verance will be able to make it easier for people to pay for music. "There are ways to create new applications to . . . convert SDMI compliance into a more positive consumer experience," he says. "For example, rather than having an SDMI-compliant device reject receipt of an unauthorized music track, the system could be enhanced to enable the consumer to acquire rights to receive and enjoy that music track."

It was this freedom to lightly control information that inspired some companies. Blue Spikes Moskowitz testified about the new DMCA: "When Thomas Jefferson said, Information wants to be free, he meant freely accessible, available to the eyes and ears of people who wait to be enriched by new knowledge and experience. . . . The threat to all of these advances by lock-and-key systems for securing copyrighted works is something that gravely concerns us. . . . Access-restriction technologies threaten the viability of a robust and fluid market for creative works."

Peter Wayner is the author of Compression Algorithms for Real Programmers and Digital Copyright Protection (Morgan Kaufmann Publishers).