Analysis of Learning Techniques for Phishing Website Detection

Phishing attacks have emerged as a new type of cybercrime with the increasing popularity of the Internet in the early 1990s. It aims to obtain users' credentials by presenting fake web pages that appear to be legitimate sites. Phishing detection mechanisms are developed to protect users against such attacks after receiving the phishing email, which are grouped into heuristic-based, list-based, learning-based, and visual similarity-based technical categories. This study furnishes an elaborate investigation and analysis of both seven well-known machine learning algorithms and four prominent deep learning algorithms to detect phishing attempts. Experimental findings using the Phishing Websites dataset show that Random Forest (RF) outperforms the existing up-to-date machine learning approaches with an accuracy score of 0.991, and Deep Neural Networks (DNN) exhibits superior performance with an accuracy score of 0.988. Also, the long short-term memory (LSTM) models got a noteworthy increase in detection accuracy by 5 percent.