Peer-to-peer networks are arousing optimism in the fight against spam but according to the Australian Coalition against Unsolicited Bulk Email they may actually contribute to the problem.
Napster co-founder Jordan Ritter is backing a new peer-to-peer spam filtering technology code-named Folsom and it's creating headlines in New Scientist. It combines heuristic scanning with distributed e-mail monitoring.
According to testing carried out by Folsom's creators, the system can eliminate nearly all spam on e-mail streams carrying between 40 and 60 percent unsolicited messages.
However, the Australian anti-spam group questions the effectiveness of spam-reduction systems like Folsom and the value of filtering at all. The organisation believes that while they reduce visible spam they could actually be part of the problem, not the solution.
For Troy Rollo, spokesperson for the Coalition Against Unsolicited Bulk Email (CAUBE), the concept behind Folsom is familiar. He said that it bares similarities with Distributed Checksum Clearinghouse (DCC); an idea for spam filtering born in the mid-90s.
DCC clients detect bulk e-mails through data sharing made possible by peer-to-peer networks. They determine how frequently an e-mail occurs across members to delineate legitimate e-mail traffic from unsolicited mass electronic mailing.
Rollo believes that the peer-to-peer e-mail-filtering concept is flawed.
"You have to delay your actual receipt of the e-mail to allow the peer-to-peer network to detect it," Rollo told ZDNet Australia. "The other thing is that they can legitimately detect 'opt-in' e-mail because all they can detect is 'bulkiness'".
However, Folsom's creators have added an heuristic component that calculates the probability that an e-mail's content is unsolicited advertising. If a message is classified as spam the program generates a signature for the e-mail which is automatically distributed to clients on the peer-to-peer network for filtering purposes.
Rollo has doubts that machine intelligence can be deployed in the fight against spam without accepting that legitimate e-mails will find their way in to the trash can.
"The problem with heuristic filters is that you're trying to put all your intelligence into programs but you're doing so by creating rules," he said. "They [machines] can mislabel things that aren't spam as if they were, and mislabel things that are spam as if they're not -- so they can get it wrong in both directions".
Nevertheless, Folsom's makers claim that its heuristic engine produces very few false positives.
According to CAUBE, even if peer-to-peer filters can be made to work effectively they would not eliminate the major problems caused by spam.
"The fundamental problems with all forms of filtering is that they tend to mask the problem at the user end thereby reducing the vocal complaint and the disincentive for people to engage in spam," Rollo said.
"That increases the volume of spam and the cost in terms of bandwidth, which is actually a significant cost."
