Phishing Website Detection Using Machine Learning

Phishing websites are fake websites that are created to look like trusted websites so that users enter their personal information, login details, banking details and other sensitive information.Traditional methods such as blacklists and fixed rules can identify many already known phishing websites, but they may have difficulty detecting newly created or modified phishing URLs.This research presents a machine learning based framework for phishing website detection using URL, domain and website-related features.The main idea is to use measurable website characteristics and train machine learning models to classify websites as either legitimate or phishing.Seven machine learning algorithms are considered in the framework: Logistic Regression, Decision Tree, Random Forest, Support Vector Machine, K-Nearest Neighbours, Naive Bayes and Gradient Boosting.The models are compared using accuracy, precision, recall, F1-score, false-positive rate, confusion matrix and ROC-AUC.A reported Random Forest benchmark on the same UCI dataset achieved 96.13% accuracy, 95.51% precision, 96.00% recall and 95.75% F1-score on 2,144 held-out records.The reported ROC-AUC was 0.9939.These results provide a useful benchmark for understanding the performance of machine learning based phishing detection.