Deep-Learning-Based Sequential Phishing Detection

Abstract

In this paper, we propose a deep-learning-based sequential phishing detection to improve the security and speed of the phishing detection. In our proposed method, phishing websites are detected in three phases: the URL, domain, and HTML analysis phases. In these phases, URLs, DNS records, and HTML contents are input to CNN-BiLSTMs (Convolutional Neural Network-Bidirectional Long Short Term Memory), respectively. Through experiments, we show that our proposed method is faster than the existing detection method, in which URLs and HTML contents are input to a CNN-BiLSTM simultaneously.