Microsoft will introduce a new spam-filtering technology across its e-mail platforms that identifies and blocks unwanted e-mail based on user-generated feedback, the company said yesterday.
The SmartScreen technology, developed by Microsoft's research unit, relies on feedback from a group of hundreds of thousands of Hotmail users to catalog spam characteristics. The Hotmail users mark spam messages, which are sent to Microsoft's anti-spam group. The machine-learning technology then takes the common characteristics of the spam messages and generates rules to apply to inbound e-mail, scoring each for the probability it is spam.
Messages arriving at a user's computer are sorted by their spam score based on SmartScreen's set of spam characteristics, with user-set thresholds relegating suspect messages to the junk-mail folder or deletion. SmartScreen is bypassed for senders on a recipient's safe-sender, safe-recipient or blocked-sender lists.
Microsoft chairman Bill Gates said in a speech Sunday night that fighting spam is a priority for the company, and that SmartScreen will play an important role in putting spammers out of business.
“We believe these new approaches will shift the tide, that between what we're doing with technology and what's being done on the legal front, it makes the business proposition for spammers no longer attractive,” he said in a keynote address at the COMDEX Las Vegas trade show. “And we've got to keep working until we achieve that. And I believe very strongly, that's an achievable goal.”
The sheer number of samples collected — more than 5 million so far — makes the criteria statistically sound and decreases the likelihood of false positives, Microsoft said.
Microsoft already has introduced SmartScreen to Hotmail, MSN 8 and Outlook 2003. It will form the backbone of the message filter in Microsoft Exchange Server, scheduled for release in the first half of 2004. Longtime Microsoft anti-spam provider Brightmail will complement SmartScreen for Exchange Server.
Along with identifying the bad, Microsoft also aims to identify legitimate e-mail. George Webb, a group manager for Microsoft's anti-spam group, said the company is considering a trusted-sender program for e-mail marketers.
“The approach of having some kind of legitimate-sender program makes a lot of sense,” he said. Webb thinks an industry-wide approach would work best.
The dual problems of the anonymity of e-mail and its low cost have long bedeviled efforts to crack down on spam. In response, Microsoft is in talks with Yahoo and AOL over instituting a program that would let marketers put up a bond to join a registry or qualify for an industry seal of approval.
Under a bonded-sender program, the bond would be debited for complaints. Commercial e-mailers would not worry about their messages getting caught in spam filters, and the extra cost to e-mailing would topple spammers' flimsy business models, according to proponents.
“The very benefit of e-mail, that it's inexpensive to send messages, is being exploited to try and find a very few set of people who actually want to respond to these messages,” Gates said. “But because the cost is so low, even if one in 10,000 responds it's economic for them to send out that e-mail.”
Microsoft, AOL and Yahoo, as part of the anti-spam alliance they formed in April, have explored options for bonded-sender programs including the E-mail Service Provider Coalition's Project Lumos. In October, the ESPC partnered with e-mail architecture company Ironport to use its bonded-sender program. TRUSTe and ePrivacy Group also have a trusted-sender program. Anti-spam company Cloudmark this fall introduced its own registry for commercial e-mailers.
“We have been working very actively on it,” Webb said. “We do think it is a promising approach. We hope to have some movement in the short term.”