Sea of Noise

Mon, 12 Jul 2004

Can Spammers Abuse Bayesian Filtering to Block Legitimate Messages?

Edward Felten made an interesting observation: Bayesian anti-spam filters are, in essence, trained by the bad guys. Could spammers exploit this situation to poison the filters and cause certain legitimate email to be blocked ("like a Google Bomb for spam filters", as Brian Carnell put it)? Those who commented on Felten's post believe not. I agree that it would be a difficult proposition, as spammers only train the filters with bad email, not good email. Nevertheless, it's a question worth considering. Hardcore spammers appear to be a fairly small group, and collective action on their part might not be difficult to arrange. Even if this particular approach wouldn't succeed, we should ask whether there are others ways they could cooperate to wreak havoc. (As if the continuous deluge of spam weren't enough...)

[/internet/spam] permanent link

Syndicate Me via RSS!

Creative Commons License
This work is licensed under a Creative Commons Attribution-ShareAlike 2.5 License.

Powered by Blosxom!