Spam Detection: A Syntax and Semantic-based Approach

Spams are electronic junk mails that are costly for the individuals and organizations who receive them and Internet service providers who handle them. They also highly contribute to the lack of trust in email service by the users, intrusion of privacy, and fraud. Spam detectors with a wide range of effectiveness are exist in the market. To the best of our knowledge, none of the existing spam detectors uses a semantic-based approach in detection of spams. In this paper, we present Semantix-Plus which uses both semantic-based indications and syntax-based indications to convert a given e-mail into a 10-variable record. The detector then uses a backpropagation neural network for the classification of the e-mail. We have compared the behavior of the Semantix-Plus with the spam detectors used by Google, Yahoo, and AASU mail servers for a limited number of spams. The results reveal that the Semantix-Plus has a superior detection power to its counterparts.