Application of a Recurrent Neural Network Model to Prevent Phishing Attacks: A Systematic Review, Challenges and Future Work

Oropeza C.; Daza A. · TEM Journal (2025) · DOI: 10.18421/tem144-07

Resumen

Phishing is the most popular form of attacks in cyberspace, accounting for 1,270,883 of attacks on organizations. The main objective of the study is to understand Recurrent Neural Networks (RNN) applications in a solution to prevent phishing attacks. A systematic review of the literature was carried out to identify aspects such as: existing research on the types of RNN, their performance against different datasets and algorithms that complement this solution. For this, 30 articles were analyzed. Among the results obtained, it stands out that the most used type of RNN deep learning algorithm was Long Short-Term Memory (LSTM), the most used dataset was PhishTank, while LSTM was the model that obtained the best accuracy with a range of 92.12% to 99.86%, and Convolutional Neural Network (CNN) was the most used complementary algorithm to prevent phishing attacks. This research provides scientific evidence on how Deep Learning techniques can improve the detection of phishing attacks, contributing to the field of cybersecurity, in the prevention, detection and management of these attacks. In addition, it provides information to make more accurate decisions in the protection of computer systems, identifying gaps in the literature related to the prevention of phishing attacks.

Instituciones: Universidad César Vallejo