Identification of SPAM Emails Using Fuzzy Logic Based Categorization for an Intelligent Email Response System

Abstract

Email is the simplest and most consistent way of communication. Emails are used for fast and reliable communication at both personal and organizational levels, including academic institutions. Some organizations have deployed auto-email responders to deal with a heavy volume of emails to respond automatically to relevant routine mails while filtering out spams. Spammers send spam emails for hacking, phishing, denial of service or marketing purposes. The proposed system is built for the Stratford University as part of a project to develop an Intelligent Auto Email Response to help teaching staff to generate responses to the students’ emails. We propose a fuzzy logic based intelligent spam filtering technique in order to generate as human like response as possible. Spam dictionary is created with ranked spam words, phrases, hyperlinks, and likely attachments. It identifies spams at various levels, filtering at sender’s address, subject, email body, hyperlinks and attachments. Fuzzy rules are applied to categorize emails into spams and hams. The rules are flexible and can give classification results like human beings. The level of threat is identified with the help of spam data dictionary. The results are categorized and help in further decision making according to the level of threat. The model has been trained and tested with two data sets - CSDMC2010_SPAM (publicly available) and Strafford256 (a set of 256 real emails provided by the Stratford University for this research). The measured accuracy of the proposed system is about 98% with public dataset and 96% with our own validation dataset. The results revealed that our proposed system has much better accuracy than other algorithms used for this purpose.