Website Phishing Detection of Machine Learning Approach using SMOTE method

One of the most serious cyber-attacks is phishing, and experts are working to discover a cure. Evaluate how the accuracy of predictions varies in an unbalanced dataset in this study. The dataset is balanced using the Synthetic Minority Over-Sampling Technique (SMOTE). The SMOTE-sample approach is suggested for identifying genuine websites from phishing sites. The suggested model outperforms popular binary-classification techniques like CatBoost, Random Forest, and XGBoost, according to experimental results using data from the Phishtank website's dataset for the identification of phishing websites. People who lack the domain knowledge for machine learning's modelling and parameter tuning approaches might benefit from the method of creating phishing website detection models. Phishing attempts and early detection of suspicious features can both be helpful. According to the findings, CatBoost has a detection accuracy of up to 97%, making it a far superior classifier than Random Forest and XGBoost. Finally, the results of the experiment showed that five-fold cross-validation outperformed the best accuracy of 97.7%.