Phishing Website Detection Using Ensemble Algorithm Convolutional Neural Network and Bidirectional LSTM

Abstract

This study focuses on phishing website detection by leveraging an ensemble of Convolutional Neural Network (CNN) and Bidirectional Long Short-Term Memory (BiLSTM) models. Phishing is a type of cyber attack where attackers disguise themselves as legitimate entities to trick individuals into providing sensitive information such as usernames, passwords, and credit card details. Given the escalating threat of phishing attacks and the limitations of traditional detection methods, this research explores the potential of machine learning techniques to enhance detection accuracy and robustness. By integrating CNN and BiLSTM models within an ensemble framework, the study demonstrates improved performance in identifying phishing websites through real-time analysis. The ensemble model benefits from the strengths of both CNN and BiLSTM architectures. CNNs are effective in feature extraction from input data, and capturing spatial hierarchies, while BiLSTMs excel at understanding sequential dependencies. The model achieves an accuracy of 89.5%, with training and validation accuracies converging to high values, and exhibits a consistent decrease in both training and validation losses, indicating robust performance without overfitting.