Intelligent Learning Architecture with Hybrid Features for Phishing Detection

This study proposes a novel machine learning architecture that uses deep learning technology to extract features from the structure of a web page and construct a model for phishing detection. Hackers can commit crimes through a variety of Internet technologies. In recent years, phishing incidents have become more frequent, and the rapid development of information technology has enabled hackers to develop more advanced phishing attacks. Furthermore, the release of phishing toolkits, which are collections of software tools, make it easier for people with minimal technical skills to launch their own phishing attacks. Therefore, more attention must be paid to the prevention of such attacks. Protection from phishing websites has various aspects, including user training, public awareness, technical security measures and others. In this research, we further improve the phishing detection on phishing kits. This research proposes to use the combination HTML structural feature with the features proposed by AI@ntiPhish1.0 to train the phishing detection model. Relevant experimental results demonstrate that the combination of AI@ntiPhish1.0 features with extracted HTML structural features is more effective on detecting the phishing kits, increasing the accuracy thereof from 82% to 87.2%.