An Evaluation and Analysis of Static and Adaptive Bayesian Spam Filters

Abstract

Spams and spamming methods are increasing vastly and getting complicated due to the rapid growth in networks, communications and technologies. Therefore, spam filters need to be tested continuously to evaluate their capabilities and efficiency in detecting and preventing spams. This paper discusses spams filtering problem using Bayesian classifier. It shows how using a combination of black and white lists and a customized spam filter based on users’ feedback can enhance the performance of Bayesian classifier. The paper evaluates three models of spam filters which are Static Bayesian Spam Filter, Light Adaptive Bayesian Spam Filter, and Enhanced Adaptive Bayesian Spam Filter. The experiments demonstrate that Enhanced Adaptive Bayesian Spam Filter, which is the one that uses black/white lists and users’ feedback, has the highest performance.