Intelligent Phishing Website Detection Model Powered by Deep Learning Techniques

Abstract

Phishing websites or URLs differ from software flaws as they exploit human vulnerabilities rather than technical weaknesses. Various methods exist to undermine the security of an internet user, but the most prevalent approach is phishing. This sort of assault aims to acquire or exploit a user's personal data, including passwords, credit card details, identity, and account information. Phishers gather user information by pretending to be authentic websites that are visually indistinguishable. Users' confidential data can be potentially retrieved, exposing them to the possibility of financial detriment or identity fraud. Consequently, there is a pressing requirement to develop a system that efficiently identifies phishing websites. This research presents three discrete deep learning methodologies for identifying phishing websites, which involve the use of long short-term memory (LSTM) and convolutional neural network (CNN) for comparison, and ultimately an LSTM-CNN-based methodology. The experimental results confirm the precision of the proposed methods, specifically 99.2%, 97.6%, and 96.8% for CNN, LSTM–CNN, and LSTM, respectively. The CNN-based technology displayed a superior phishing detection mechanism.