Abstract
In recent times e-mail is proved to be an eradicator application on internet. Spam has traditionally been most visible threat and fashioning the major problem in the e-commerce society. Survey reports showed that the 60% of all mail messages are Spam [1]. Identifying and removal of spam from the email delivery systems allows end users to regain a useful means of communication. Spam filters offer a way to curb the problem. In this paper, we present learning based personalized anti spam filter. We discuss the architecture and present a comparative study of the proposed anti spam filtering technique. We use public and private email corpora from wide community of internet users collected over several months for comparative study. Simulation results show that appropriate training of a personalized filter gives more adequate results than the schemas previously proposed.