31-10-2009, 03:13 PM
Abstract
A lot of research has been done in the area of spam _ltering and several sophisticated methods using arti_cial intelligence have been proposed. However, most of the open source spam _lters available to- day do not provide consistent accuracy levels. Spamassassin, which is at present said to be best open source spam _lter, aims at providing a single solution that can satiate the needs of individuals as well as large organizations. It _lters spam at various levels using di_erent methods with the idea that spammers will be blocked atleast at one of the levels. However, none of them take concept drift[1] into consideration which is a very dominant factor in achieving consistent accuracy levels. In this project we intend to address this issue by developing a collaborative spam _lter that uses a centralized incrementally learning spam rules database to detect spam. The _lter is aimed for use at an intranet or organization level. The main idea behind using this technique for an organization is that a collaborative _lter along with centralized incre- mental learning[3] will greatly help in detecting concept drift at the earliest and signi_cantly reduce the chances of a given spam spreading too much in an organization.
[attachment=416]