A CNN-Based SIA Screenshot Method to Visually Identify Phishing Websites

Abstract

Phishing evolves rapidly nowadays, causing much damage to finance, brand reputation, and privacy. Various phishing detection methods have been proposed along with the rise of phishing, but there are still research issues. Phishing websites mainly steal users’ information through visual deception and deep learning methods have been proved very effective in computer vision applications but there is a lack in the research on visual analysis using deep learning algorithms. Moreover, most research use balanced datasets, which is not the case in a real Web environment. Therefore, this paper proposes a security indicator area (SIA) which contains most security indicators that are designed to help users identify phishing sites. The proposed method then takes screenshots of SIA and uses a convolutional neural network (CNN) as a classifier. To prove the efficiency of the proposed method, this paper carries out several comparative experiments on an unbalanced dataset with much fewer phishing sites, which increases detection difficulty but also makes the detection closer to reality. The results show that the proposed method achieves the highest F1-score among the compared methods, while providing advantages on detection efficiency and data expansibility in phishing detection.