International Journal for Recent Development in Science and Technology

Cybercrime, especially phishing, is one of the growing concerns today.First discovered in 1996, phishing has become one of the most dangerous forms of cybercrime, using email manipulation and fraudulent websites to steal sensitive information.Although many studies have been conducted on the prevention, detection, and solutions of phishing, there is no holistic approach that has been established.This paper presents an automated phishing detection framework using machine learning algorithms.The objective is that of distinguishing phishing from benign while keeping the false positive as low as possible.Therefore, the research focuses on a static phishing detector for windows operating systems using more than 11,000 genuine and phishing website URLs that have been preprocessed on which several machine learning algorithms are applied, including: Decision Tree (DT), Logistic Regression (LR), Random Forest (RF), Naive Bayes (NB), Gradient Boosting Classifier (GBM), K-Nearest Neighbor (KNN), and Support Vector Classifier (SVC).In addition, a hybrid LSD model combining LR, SVC, and DT with soft and hard voting mechanisms is proposed.The LSD model applies canopy feature selection, cross-fold validation, and grid search for hyperparameter optimization.The models are tested using metrics such as accuracy, precision, recall, F1-score, and specificity.The results show that the LSD model outperforms others in accuracy and efficiency.The model, although successful, does not inherently learn to change with the changing phishing techniques as attackers constantly modify their methods.This can be overcome by retraining the system using updated data.Although this study is on static analysis, dynamic analysis may be integrated in future work to enhance detection.Overall, the proposed framework provides an effective solution for phishing detection, although periodic updates are necessary to maintain effectiveness.