Abstract: In recent years, a drastic increase has occurred in the number of phishing attacks that led individuals to lose personal information, valuable assets, financial data, etc., to unauthorized hackers. Phishing is a type of malicious attack in which the cyber-attacker develops a mischievous website and baits the victims by making the website pose as a legitimate one, to steal their valuable information on personal and financial fronts. So, it is the need of the hour to develop an efficient and accurate method to find out the legitimacy of the websites. In literature, various methods have been developed in line with Machine Learning (ML), Deep Learning (DL), and other such approaches for the detection of malicious URLs. The current manuscript provides state-of-the-art URL phishing attack detection research based on deep learning models. A comprehensive comparison between these models is discussed and compared. The comparison covers the three phases used in URL phishing detection. These phases are data preprocessing, predictive model design, and performance evaluation. The study concludes that models based on CNN and Bi-LSTM outperform other models in terms of attack detection.

Keywords: phishing, url attack, machine learning, deep learning, cnn, bi-lstm

Pages: 558-573