Improving Email Security with Fuzzy Rules

Phishing and other malicious email messages are increasingly serious security threats. An important tool for countering such email threats is the automated or semiautomated detection of malicious email. This paper reports work on using fuzzy rules to classify email for such purposes. The effectiveness of a fuzzy rule-based classifier is studied experimentally on a real dataset and compared with results for other classifiers, including those based on crisp rules and decision trees. The human-readability and editability of the classifiers produced by these methods is also studied.