Phishing Webpage Classification Method based on Joint Features

Abstract

Phishing websites aiming at stealing users ‘ information by mimicking legitimate trustworthy websites are still a major threat for today ’ s Internet thread. This paper systematically describes a phishing webpage detection method based on joint features. The joint feature is introduced to carry out feature cross-expansion, and the expanded joint feature vector length ranges from 40 to 376. The paper compares four optimal parameter models based on basic features and joint features: random forest is the optimal model; neural network model has a small increase in classification ability after introducing joint features; logistic regression and Adaboost have classification ability after introducing joint features constant.