Email spam filtering using the combination of two improved versions of support vector domain description

Abstract

Email is fast and cheap message transfer way, it has become widely used form of communication, due to the popularization of Internet and the increasing use of smart devices. However, an unsolicited kind of email known as spam has been appeared and caused major problems of the today’s Internet, by bringing financial damage to companies and annoying individual users. In this paper we present a new method to classify automatically legitimate email from spam, based on the combination of two improved versions of Support Vector Domain Description (SVDD), the first one aims to adjust the volume of the minimal spheres, while the second replaces the standard decision function of SVDD by a new improved one. An experimental evaluation of the proposed method is carried out on a benchmark spam email dataset. The experimental results demonstrate that the proposed method achieves high recognition rate with good generalization ability.