Svm-based Spam Filtering

This paper mainly introduces a design model and implementation method of spam-filting system in Windows. This spam-filting system can identify , judge and filt some typical spam emails. It can also classify other emails. By statics, from 2001 the speed of spam emails increases very quickly. Spam emails that internet users received are twice of non-spam emails in quantities. Aim at fast increasing speed of spam emails these years, an effective method of preventing spam emails is requested urgently. Filtering with content is one of methods. So spam-filtering system this paper introduces in Windows specially researches the contents of emails. It has good value.The purpose of developing this system is mainly to be acquainted with current developing actuality of spam-filtering and to study filtering technology. It can prevent spams effectively. According to study and practice , some problems met in developing process of spam-filtering can be discovered. Some attitudes and opinions can also be presented by combining with my research works.This paper firstly introduces current developing actuality of spams and non-spams technology, some elemental conceptions and theories of spam-filtering. And this paper also introduces current research of information filtering and documents categorization . Then theory of svm,influence of paraments and method applying into this system is mainly discussed. And the method of building into outlook and concretely implement of this system are mainly discussed. Finally some obstacles of the current anti-spams are summarized. Some solutions of these problems and trends of development are discussed. All of these will direct for the later investigations.The purpose of this paper is to be an automated spam-filtering system that is built into outlook2000 with SVM. Several sets of sample messages were colleceted to build dictionaries of words found in e-mail communications, which were processed by the SVM to create a classification model for spam or nonspam messages. This system is divided into service and custom systems. The filtering with content is mainly on the custom and this system is applied on outlook. Spams are filtered by classification method of SVM. Experiments show that system applying on spam-filtering has itsacceptable accuracy.