Phishing Website Detection using Machine Learning and Deep Learning Techniques

Abstract

Phishing attacks are a growing cybersecurity threat, exploiting deceptive websites to steal sensitive user information.Traditional detection techniques struggle to adapt to the evolving nature of phishing tactics, leading to reduced accuracy.This research presents an enhanced phishing web site detection framework using deep learning models, that is, GatedRecurrent Units (GRU) and Convolutional Neural Networks (CNN), in order to enhance classification precision and counter cyber threats efficiently. Initially, machine learning- based classifiers such as Decision Trees and Random Forests were employed to distinguish between phishing and legitimate websites. These models provided baseline insights into feature importance and classification effectiveness. Subsequently, deep learning approaches, including GRU and CNN, were integrated to enhance detection capabilities by capturing sequential and spatial patterns in URLs and website structures. Experimental results demonstrate that CNNs outperform GRUs in detecting phishing websites, highlighting their ability to recognize complex features within malicious URLs. The study also incorporates an optimized data preprocessing pipeline, including URL normalization and tokenization, ensuring robust feature extraction. The findings of this research contribute to strengthening online security by providing a scalable, automated phishing detection system. Future enhancements include integrating ensemble learning techniques and deploying the system as a cloud-based solution for real-time phishing prevention