Towards an Ontology Driven Spam Filter

Email´s popularity has led to the increase in unsolicited mails. Currently spam filters use the structure and syntax of email body along with training methods to classify email as spam or ham. These include techniques such as word statistics and Bayesian filters. In this paper we propose to extend spam filters to use the semantics of an email as an additional parameter for classification. We suggest a system that uses ontologies to discover relationships between tokens in an email. Using semantics presents challenges such as: building the ontology, relationship discovery, relevancy scoring and so on. We discuss these challenges in detail and propose possible solutions to them.