Special Reports

MORE than two-thirds of all email traffic is spam. But the problem could be reduced if our computers work together to control it.

Today's anti-spam software-filters block messages that have content such as advertising slogans or sexually explicit words that is similar to that of spam emails already received and identified. Therefore, they cannot pick up new spam messages that are unlike any received before.

But anti-spam programs would be vastly more powerful if they could pool information about spam, much as police in different places share tips on known criminals. So says computer scientist Vwani Roychowdhury of the University of California, Los Angeles, who together with Oscar Boykin of the University of Florida and other colleagues has now proposed a practical way of doing it.

The team suggests adding software to standard email programs that could orchestrate a behind-the-scenes collaboration. When you receive a new message, your anti-spam software would first check it against your own database of known spam. If it doesn't find a match, it would then forward the same query to a few randomly selected email addresses in your contacts book. Similar software on each computer that receives the query would then check the message against its own spam database, and so on, until a match is found, or the message is deemed original.

In this way, an entire social network of email users can pool its experience of spam messages, greatly increasing a spam filter's accuracy. In simulations, the researchers found that if the network contained many users - hundreds of thousands or millions - then it would detect almost all spam emails, while only rarely misclassifying legitimate messages.

"This is a really great idea," says computer scientist David Hales of the University of Bologna in Italy. "It turns the existing trusted social network into a kind of extended spam filter."

As Roychowdhury and his colleagues point out, the inherent trust within the social email network can also be used to foil spammers' attempts to sabotage the system. A spammer might try to wreck the system from within, posing as an ordinary user, but supplying false information: listing legitimate emails as "known spam" in their own email system, for example.

But the anti-spam software could be told to weight the responses it gets, lending more weight to those returning from its most trusted contacts - people to whom the software's owner frequently sends emails etc. Spammers give themselves away by their pattern of email usage because they send a lot of emails but don't receive many.

The researchers aim to make their software available soon, and hope that it will spread rapidly, as the system's success depends on it having a large number of users. "The main strength of the idea," says Boykin, "is that essentially everyone on the planet would be collaboratively filtering spam."