A New Approach for Collaborative Spam Filtering
In response to the growing volume and variety of spam, spam filters have moved away from monolithic repositories situated on central servers towards dynamic knowledge bases located on local servers. The state of the art in spam filtering sees content-based filters tending towards collaborative filters, whereby email is filtered at the mail server using content-based techniques, with users feeding information back about false positives and false negatives. This feedback enables the spam filter to track concept drift in spam and to be retrained in the case of false positives. While these filters can achieve statistically impressive accuracy rates, they remain prone to false positives, i.e., the erroneous classification of a legitimate email as spam.