Napster Co-Developer Offers Anti-Spam ToolJordan Ritter, one of the developers of online music sharing service Napster, and software developer Vipul Ved Prakash plan to introduce anti-spam technology next month that reportedly goes further than other such technologies by making judgments about e-mail and learning patterns of spam.
The developers claim tests of the technology, called Folsom, on e-mail networks containing 40 percent to 60 percent spam reduced unwanted messages to almost zero. They also said it blocks fewer legitimate e-mails than other technological solutions.
Folsom, which is to be released in May, will be marketed through start-up company Cloudmark, San Mateo, CA. The company's www.cloudmark.com Web site has a page stating the site will be up soon and provides general information and investor relations e-mail addresses. There was no word on what, if any, cost there would be for the product.
Based on software developed by Prakash called Razor, Folsom lets e-mail recipients mark a particular message as spam. The program then automatically assigns the message a small signature based on its contents and forwards the message to a number of servers on the Internet. The signatures are automatically downloaded by other computers equipped with Folsom, enabling users to block other copies of that message.
The learning aspect of the technology lets Folsom identify new spam by looking at the words and phrases in previously marked messages and making a statistical judgment about the new message.
The claim that Folsom differs from other anti-spam technologies is mostly hype, said Anne Mitchell, director of legal and public affairs at anti-spam group Mail Abuse Prevention System LLC, Redwood City, CA.
"It's interesting that it attempts to distinguish other systems as relying on human input, because so does this system," she said. "It relies on people reporting that a message is spam to the central server. So, in that way it's really no different than any other system out there."
Mitchell said that Folsom appears to identify spam based on content, something MAPS does not do.
"True, it has a 'machine learning' element," she said. "However, it appears that is content based. One of the things MAPS does not do in any way is judge something based on content. It's all about permission."
MAPS reportedly is working on an anti-spam method that uses pattern recognition, but Mitchell would not provide details about the system or how far along it is in the development cycle.
No one from Cloudmark was available to comment.
Ray Everett-Church, chief privacy officer at ePrivacy Group, said Folsom sounds like an interesting approach to fighting spam. However, he said it may not solve the issue of the cost incurred by ISPs handling spam.
"The challenge is still dealing with volumes of spam and the complexity of ISP mail systems," he said. "[E]nd-user spam filtering has a fundamental problem in that by the time spam hits a filter, the costs of transporting the spam have already been incurred."