Empirical Evaluation of Phishing Attack Identification Over Websites by Using Intelligent Deep Learning Principle
G Anitha, S K Mouleeswaran, Dickson Irudayaraj., G. Santhi, Rajkumar Chadge, Hani Attar
Abstract
Phishing schemes keep on dominating the cyber threat landscape hence requiring enhancement of detection methods. This research sets out a new and novel idea for the detection of phishing website approach incorporating the XGBoost with ANN to boost the identification performance. The proposed model was compared with nine machine learning and deep learning models which includes Logistic Regression (LR), Decision Trees (DT), Random Forest (RF), SVM, KNN, Naïve Bayes, XG Boost, Artificial Neural Networks (ANN), CNN. It is apparent that the Ensembled XGBoost-ANN model was able to attain a higher accuracy rate of 9,463% and ranks on top of all the other models. It also showed great accuracy to prove precision, recall and F1-score metrics of the model. The findings reported herein support the suitability of the ensemble technique in modeling the intricate relationships characteristic of phishing websites, thereby holding the promise of being a practical solution in real-world environments. The reason why this methodology is promising for deployment in cybersecurity frameworks targeting the mitigation of phishing attacks is its capability of achieving high detection rates without making many false-positive detections.