Detection Of Phishing Emails Based On The Lexical Features

The massive Phishing attacks,which usually spread by emails, make users suffering from significant financial losses and suspecting the websites about E-bank or E-commerce. It impacts the development of E-commerce. Detecting Phishing emails, the origin of the attacks, could prevent those huge financial losses. For this reason, the detecting technology of Phishing Emails is meaningful. In this paper, the main work is:1. According to the development trends of phishing emails, several new features are proposed for detection. The attackers are developing the strategies, so that the mails can pass the mail filters. As main parts of the detecting method, the features of the phishing emails should be evolving so as to accommodate the altered strategies. Considering the evolution and detection of Phishing emails, the thesis presents several new lexical features. And the experiment results show that it has better performance with the new features.2. Taking the online-learning strategy, the classification model will update when errors come out. The model was stable after being built up. However, the fixed models can not meet the needs of multiple emails. They could possibly return errors while new mails came out. For these reasons, online-learning categorizer is necessary for the detection. By updating the models, the accuracy was improved and errors were reduced. The results show that, it has better performance if the online-learning strategy is applied.