Desenvolvimento de um sistema Anti-Spam de Código Aberto

Abstract

In this work, we present the development of an Open Source Anti-Spam System (SASCA) in Java. Unlike commercial anti-spam systems, SASCA does not make use of block lists (black / white), but rather of m machine learning models for email classification. Several experiments were carried out on a basis of real e-mails collected at the Federal University of Itajubá. In the experiments carried out, it was observed that the SASCA performed very close to the CanIt commercial anti-spam system in terms of e-mail classification, but with a much better performance in terms of the time required for classification.