PhishBlocker: Phishing Detection System based on Uniform Resource Locator (URL) using Hybrid Approach

Abstract

Phishing attacks remain a significant threat, emphasizing the need for effective detection systems. This study presents PhishBlocker, a URL-based Phishing Detection System using an integrated approach combining Support Vector Machine (SVM) with rule-based scoring to distinguish phishing URLs from legitimate ones. The system uses a balanced dataset of 60,000 URLs that consists of 30,000 phishing, 30,000 legitimate URLs, cleaned and analyzed with 22 features like domain age, ssl validity, redirection depth and phishing-related keywords. Implementing Object-Oriented Analysis and Design (OOAD), PhishBlocker uses a consensus scoring approach, integrating SVM and rule-based scores. The dataset is split 80:20 for training and testing, with n-fold cross-validation for model evaluation. The system is developed in Python with Flask, Scikit-learn and Chart.js, the system achieved 93% accuracy from training and testing with the balanced dataset. Future work should integrate deep learning and real-time adaptive training to enhance detection, benefiting organizations, cybersecurity professionals, and end-users with improved phishing defenses and reduced false positives.