A Comprehensive Machine Learning Approach to Phishing Detection Using Features Selection and Deep Learning Models

Abstract

Phishing attacks have become a key cybersecurity threat and have used the trust of users in order to harvest sensitive information. This research work is based on an advanced phishing detection model by combining feature selection tasks with machine learning and deep learning models. Using a labeled dataset, where the status field denotes legitimate or phishing websites, we do a performance evaluation and comparison of different models such as Graph Convolutional Network (GCN),Tab Transformer, Auto encoder, Feedforward Neural Network (FNN), and Deep Neural Network (DNN). By applying the optimum feature selection, we improve the performance of the models, lower the computational complexity, and improve the generalization. The system implementation is performed in Python programming language and deployed with a web interface for interaction (Flask web Service) according to the style of user interaction (html, css); Our results show that the synergy of integrating deep learning architectures with feature engineering results in good enhancement of phishing detection accuracy and robustness. This approach is scalable and efficient way to protect the users from phishing attacks in real world applications.