Spam filtering

Abstract

In this project, I study the problem of spam filtering. I have implemented a mail proxy server, ICF, Internet Content Filtering system, for addressing and testing various spam filtering algorithms. Spam filtering is different from other document classification. Email contains special characteristics which help to design an algorithm with high accurate rate for filtering. I will study a common spam filtering algorithm, Naive Bayesian classifier and a well known classification algorithm, support vector machine. We will also study the feasibility of cocktail approach, which can achieve filter sharing among group of user. So that suffering of a user will be the pleasure of others. In our experiment, I found that the accurate rate of filtering is as high as 98%.