Deep Learning and Ensemble Methods for Robust Financial Email Phishing Detection: A Performance Benchmark
Fardeen Mansoori, Khushi Chauhan, Sandeep Kumar · 2025
Email spam and phishing attacks, such as those rising from the increased reliance on digital communication, pose considerable threats to the security of financial data. These attacks usually attempt to trick recipients into providing sensitive information that can be used for significant financial loss or the disruption of the integrity of financial systems. Thus, with such sophisticated tricks being discovered by these modern-day spammers, there is a growing need for more effective, adaptive detection mechanisms beyond traditional spam filters. This paper focuses on detecting spam emails that target financial fraud using the power of machine learning (ML). Four very commonly used models, namely Naive Bayes (NB), Support Vector Machine (SVM), Random Forest Classifier (RF), and Deep Neural Network (DNN), are probed for their effectiveness in detecting spam emails from a Kaggle labeled email dataset.Our comparative analysis shows that, of all the tested models, the SVM classifier outperformed and achieved a great accuracy of as much as 98%. This is an indication that SVM is well-suited for detecting complex patterns that may be hidden in the email content, indicating phishing or spam. Also, the study provides an intense examination of evaluation metrics such as accuracy, precision, recall, F1-score, and the Area Under the Receiver Operating Characteristic Curve AUC-ROC. These metrics help one understand the strengths and weaknesses of each model, more so in dealing with imbalanced datasets where phishing email occurrences are relatively few compared to the regular ones.