Ensemble Learning Approach for Phishing Website Detection Using an Optimal Greedy Stacking Model

In our daily life, Phishing attack causes a substantial menace among different types of cybercrimes. Due to the COVID—19 pandemic, the growth of internet users in the last two years has increased enormously. At the same time, phishing attacks have also increased. Many internet users are suffering financially and their personal information is being misused for different crime purposes. Using a fake URL, the attacker attempts to access the user’s credential information like bank details, details of credit cards, individual details etc. In the modern digitization era, the phishing website detection algorithm is a challenging task to safeguard internet users from these attacks. In this study, a new technique has been formulated to detect phishing websites using an ensemble learning approach. Initially, important features are collected from different website URL addresses using the Random Forest Regressor model. Then different supervised Machine Learning (ML) algorithms, namely Naïve Bayes (NB), K-Nearest Neighbour (K-NN), Support Vector Machine (SVM), Random Forest (RF), Bagging (BG), Logistic Regression (LR) and Artificial Neural Network (ANN) are used to detect phishing website depending on collected features. Finally, to enhance the precision of detecting phishing websites, an optimal stacking model with a greedy approach has been employed. To assess the effectiveness of the proposed model, the widely used dataset from the UCI Machine Learning Repository (ML Repository) has been exploited. In identifying a phishing website, the suggested method achieves a level of 96.73% accuracy.