A Malicious URL Detection Method Based on CNN

Abstract

In recent years, phishing events occur frequently, and the detection of phishing URL has become a common concern in the field of network security. In previous studies, researchers distinguish phishing URLs from normal URLs by the string characteristics. However, this method is difficult to achieve efficient detection accuracy in the case of attacker's forgery of URL features. In this paper, we propose a phishing URL detection method based on URL content. Firstly, the simple features are used for preliminary screening, and then, the URLs that cannot be distinguished by simple features are input into the simulation environment to obtain the page content. Finally, the method based on CNN is used to detect malicious URLs based on the page content. Experiments show that our method can achieve satisfying detection accuracy.