Filtering spam at e-mail server level with improved CRM114

Security managers and network engineers are increasingly required to implant corporative spam-filtering services. End-users don't want to interact with spam-classify applications, so network engineers usually have to implement and manage the spam-filtering system at the e-mail server. Due to the processing speeds needed to put these solutions into work at the server level, the options at hand are reduced to applications of the black-list/white-list type. This is the reason behind the fact that most applications based on AI techniques run only on the client side, particularly those based in the Naïve Bayes scheme, which has proved to be one of the most successful approaches to fight against spam, but nowadays is not as fast as other techniques and still not able to process the high amount of email traffic expected at a mail server. However, spam mutates and the spamies techniques have quickly evolved to easily pass the traditional black/white list applications, so there is a compelling need for the use of more advanced techniques at the server level, notably those based in the Naïve Bayes algorithm. This article explores this possibility and concludes that, simple improvements to a well-known Naïve-Bayes technique (CRM114[2]), following some ideas suggested in [8], could turn this algorithm into a much faster and significantly better one that, due to these improvements in speed, could be used at the server level.