Watermarking: Back to Square One?

In December, after months of conducting listening tests with audio professionals in the US, the Secure Digital Music Initiative (SDMI) choose a watermarking technology from Verance Corporation for DVD-Audio copyright protection. Test results had indicated that Verance's system was the least detectable of the contenders under consideration.

The tests were administered by Sony engineering vice president Malcolm Davidson, and conducted in studios in New York, Los Angeles, and Nashville using ABX switching of time-aligned high-quality playback equipment. (I participated in one of the Nashville tests, in which the test music was a rock recording from Dire Straits. In multiple trials I was unable to identify the watermarked version.) The SDMI's decision to go forward with watermarking was made despite protests from engineers and others in the industry, who variously objected to the test methodology, the limited type of music used, and the exclusion of professionals and ordinary music fans from outside the US. On June 5 of this year, Verance Corporation chairman David E. Leibowitz offered a strong defense to such criticism in a www.stereophile.com "Soapbox," reproduced in full in the "Manufacturers' Comments" section of the August 2000 Stereophile.

One of the most vociferous opponents of the rush to adopt Verance's watermarking scheme has been classical mastering engineer Tony Faulkner, of London's Green Room Productions, who presented some compelling arguments that watermarks seemingly undetectable on rock recordings might not be so on more revealing material, such as solo vocals or string quartets. He claimed that the signature imposed on recordings would become more apparent as the level of resolution increased.

Faulkner appears to have been vindicated by tests conducted in July at Whitfield Street Studios in London. In a trial that was extended to accommodate a larger-than-expected number of participants, British engineers, record producers, and journalists attempted to identify the marked recordings under blind conditions—with a high degree of success, according to the July 22 edition of New Scientist. As Faulkner had predicted, a statistically significant number of test subjects repeatedly identified the watermarked music, to the dismay of SDMI officials. "We are starting all over again," an SDMI spokesperson stated after the tests, which were marred by the early-morning theft of a server computer belonging to Davidson.

The audible alteration of recorded sound violates one of the SDMI's primary guidelines for watermarking: that it be undetectable on music played back in a high-resolution format. This requirement would seem to be at odds with another guideline: that the watermark be robust enough to survive multiple generations of low-resolution digital (MP3) or analog copying. Whether the Whitfield Street results will ultimately scuttle SDMI's adaptation of Verance watermarking wasn't clear by July 30, but the New Scientist report appeared at almost the same time as announcements of other watermarking efforts.

From Tokyo came a statement from Japanese authors' society JASRAC that it is joining forces with the international authors and composers organization CISAC and mechanical rights organization BIEM to evaluate digital watermark technologies in a project—to be known as "STEP 2000"—at Tokyo's Nomura Research Institute. Participating in the project will be representatives of Australia's Australasian Performing Rights Association; ASCAP and BMI from the US; BUMA/STEMRA from the Netherlands; Hong Kong's CASH; Germany's GEMA; France's SACEM; Spain's SGAE; Italy's SIAE; and Finland's TEOSTO.

"This project is aimed at furnishing enterprises involved in digital music distribution with major options for selection of digital watermark technology," the JASRAC announcement stated, with a disclaimer that its efforts are not intended as competition with SDMI. "We're sure that this will have a positive effect on the record industry," said JASRAC director Mamoru Kato. "This initiative supports SDMI."

Meanwhile, Minneapolis, Minnesota–based Cognicity, a watermarking technology developer, has signed a partnership agreement with the InterDeposit Digital Number (IDDN) organization, "an international federation made up of organizations concerned for the protection of intellectual property rights over digital works," in the words of a company press release. Terms of the agreement have Cognicity providing an " 'AudioKey' watermarking solution to allow lnterDeposit members to embed IDDN digital identifiers into their audio content, therefore being able to identify their music files by the IDDN system."

The insertion of digital ID numbers in recordings is intended to help track the origins of pirated recordings, for the possible prosecution of copyright violators. "Our members are very concerned about protecting their intellectual property," said Daniel Duthil, president of InterDeposit, "and the Cognicity AudioKey solution provides a truly world-class means of meeting this urgent need." The AudioKey tracking system is said to be "truly inaudible, yet persistent" and is applicable to all "popular audio formats."

As has been pointed out before, the inclusion of watermarks or any other identifiers is completely voluntary on the part of recording companies. The specialty labels that serve the audiophile market are unlikely to use any of them, but the major labels that control 85% of the global market in recorded music can be counted on to do so in a big way. Audiophiles in general support the concept of the protection of intellectual property, but object strenuously to efforts that might diminish the potential of the most transparent and revealing recording and playback technologies ever devised. In the forthcoming September issue of Stereophile, editor John Atkinson could generate a shockwave with his bold stance on the controversy. Watermarking is certain to be the hottest topic at the Audio Engineering Society convention in Los Angeles later that month.