Spam Detection Using Statistical Theorem

摘要

Bayesian filtering is an anti-spam algorithm which is designed to tackle spam dealing with probabilities. It is named after Bayes' Theorem of statistics, which is used to calculate the probability whether a message is a spam or not. This filter works efficiently by comparing e-mail content (phrases or tokens) against stored databases. This paper proposes Binomial Distribution and Poisson Distribution to be implemented in Bayesian spam filter. This approach is beneficial for calculating the probability of a mail being spam containing words that are not stored in database (i.e., encountered by the spam filter for the first time) or rare words (less frequent words). The proposed distribution for spam filters reduces and controls the false positives.