Enhancing Phishing Email Detection Using Hybrid Ensemble Learning

Email phishing represents a serious threat in the digital era, capable of compromising personal information. So, we need more effective technology using machine learning models to filter the email. This study aims to enhance the quality of phishing email detection by the application of ensemble learning to hybrid features, which are a combination of text-based and content-based features. In this methodology, Support Vector Machine will be used for content-based features processing and Multilayer Perceptron will be used for text-based features processing. To increase the accuracy, Ensemble Learning with Stacking Classifier and Soft Voting Classifier will be used to test the hybrid features. The results show that Stacking Classifier Ensemble Learning produced the greatest accuracy, increasing by 0.008 from MLP and 0.120 from SVM. However, the increase is not very visible due to imbalance in the number of features between content-based features and text-based features when combined in the Ensemble Learning. However, this hybrid model can still be improved by replacing the Machine Learning or Deep Learning models or changing the dataset.