Research on GAN-Based Training Data Generative Technique for Phishing Web Pages

In order to solve the problems that machine learning algorithms face in the field of Phishing web page recognition, a feature sample expansion model of Phishing web page training data MD-GAN based on generative adversarial network (GAN) is proposed to expand the sample data set. Based on the in-depth analysis of the characteristics of Phishing web pages, this paper uses regular expression and image transformation methods to design and extract 32-dimensional features including URL, HTML and JavaScript information as model inputs, and optimizes the original generated adversarial network model by introducing local discriminators to improve the quality of generated samples. Experimental results demonstrate that, in contrast to other sample expansion approaches in various classification algorithms, the performance of this method is further enhanced.