Napster said Friday that it has licensed acoustic fingerprinting technology from Alexandria, Va.-based Relatable that identifies songs using the wavelength patterns produced by their sounds. It was unclear, however, when the file-swapping company planned to implement filters based on such a system or whether such an approach would satisfy its adversaries.
The deal follows renewed threats this month of a court-ordered shutdown of the wildly popular file-swapping network. The company is battling a massive copyright infringement suit against the record industry and has already begun to block some song files from its network. But those efforts, which rely on text-based filters that identify songs by title and artist, have been widely judged a failure.
"For a long time, Napster was being lazy and inadequately blocking tracks from bands like Metallica weeks after" the court order, said Aram Sinnreich, an analyst at Jupiter Research.
Napster must tread carefully in suddenly blocking access to wide swaths of popular music. Millions of people now use the service primarily as a way to create free MP3 collections, and they could easily abandon it if they can't find the songs they want.
Nevertheless, the company's legal setbacks have put it on the hot seat to implement effective filters quickly.
Earlier this month, U.S. District Judge Marilyn Hall Patel criticized Napster's blocking efforts as "disgraceful" and threatened to shut the service down. Stopping short of requiring Napster to change its filtering practices, Patel has sought an opinion from an appointed technical expert.
Whether Relatable's technology will prove more effective than the current text-based system is unclear. Just weeks ago, Napster's own attorneys argued that fingerprint filters are not yet feasible.
Pat Breslin, the chief executive of Relatable, said it is still undetermined how Napster will specifically use its technology. But the thought that his company's software could be the gatekeeper for millions of Napster users has kept him busy.
"We do have technical challenges because never has there been a fingerprinting solution that would scale to Napster," Breslin said." It is daunting, but our design specifications address speed and scale." Relatable's fingerprint technology, called TRM, identifies each song uniquely by comparing acoustic patterns. Since no song sounds exactly alike, the software can assign an identity for every song that it encounters, although it requires a master list for comparison.
Theoretically, the technology works like this: If a Napster user tries to download "Hotel California" from another computer's hard drive, the filtering software will know what song it is and whether it can permit the transfer. If the song has been identified by the record industry as a copyrighted work, then it's blocked.
The system is considered potentially more effective than text-based filter systems, such as the one currently employed by Napster, which identify songs by file name.
Text-based filters are widely used in search engines, but they're not effective in peer-to-peer systems where dozens of file names may represent the same song. After Napster began text filtering, for example, people began offering tools to change the spellings of popular songs and encrypt titles, making it more difficult to police the network.
That's not to say Relatable's system is foolproof. There are still ways of tricking the fingerprinting system, but these methods are more sophisticated than simply misspelling the song's title.
Gene Hoffman, chief executive of EMusic, which also uses Relatable products, said someone savvy and malicious enough could trick the software. The compromiser could feed nonsense codes and numbers while the software searches for song identities.
"I can write a client that looks and smells like Napster but misbehaves," Hoffman said. "Instead of turning in a correct fingerprint, you could fake it so they all have the same fingerprint."
Another drawback, analysts said, is that for the software to know what songs are copyrighted in the first place, there needs to be a central database of identities for comparison.
"For fingerprints to be useful, there needs to be a master database with the official information about the rights owner of the song," said Greg Rohda, an analyst at online music research company Webnoize. "That's going to have to come from the recording industry." Napster began blocking songs early last month after the 9th U.S. Circuit Court of Appeals revised an early court injunction, letting the company continue the swapping service as long as it took all "reasonable" steps toward blocking copyrighted songs identified by the record labels.
Citing the appeals court's insistence that the industry point to specific files on the service, Napster has to this point been blocking songs only when given an artist name, song title, and file name. The company has blocked 311,000 individual works, although the Recording Industry Association of America says labels have identified more than 600,000 songs.
Napster hopes that fingerprinting technology can fill some of the gaps in its filtering strategy. If it doesn't fit the bill, however, there are other ways that online companies can control and identify content files, according to Webnoize.
One method, called MD 5 Hash, provides a mathematical equation that assigns a numerical value based on the data contained in the file. This system is sensitive to any changes in files, however. If the file is changed even slightly, it can become unrecognizable.
File-swapping services can also remove themselves from the legal and financial hassle of using other people's content by creating and distributing their own.
Regardless of questions over its effectiveness, however, Napster's initiative in seeking to incorporate a better way to filter copyrighted songs has so far won initial approval from the record industry.
"Napster's apparent interest in complying with the order is good news for creators and seems to be a step in the right direction," Cary Sherman, senior executive vice president and general counsel for the RIAA, said in a statement.