A Long Short-term Memory Network Model for Detecting Phishing Websites

Abstract

Cybercriminals employ a variety of strategies to trick victims into clicking on dangerous links or divulging personal information. Phishing is the most common form of psychological manipulation. It can lead to identity theft, ransomware attacks, credit card fraud, and other major financial losses for both individuals and businesses. Organizations all throughout the world are adopting clever solutions in response to threats that are becoming more sophisticated. Because deep learning and other technologies can recognize and infer patterns from big datasets, they have been shown to be useful in securing cybersecurity operations and computer operations (Cagatay, et al., 2022). In this paper, an LSTM network model is proposed for detecting phishing websites. Legitimate URLs and phishing URLs were collected from various sources and features that identify a website to be phishing were extracted. Features with a mean greater than 1.5 were selected using Random Forest, RFE Variance Threshold and Univariate model. The LSTM model was experimented using 50, 70 and 100 units in the dense layer. The LSTM model with 100 units demonstrated the best performance overall. It achieved the highest accuracy of 96.68%, balanced accuracy of 96.56%, precision of 97.17%, recall of 95.78%, F1 score of 96.47 and the lowest FPR 2.66%.