Improving the accuracy of cybersecurity spam email detection using ensemble techniques: A stacking approach Machine...

Area of Science:

• Computer Science
• Cybersecurity
• Machine Learning

Background:

• The proliferation of unsolicited email (spam) poses significant cybersecurity threats and degrades user experience.
• Effective spam detection mechanisms are essential for modern digital security infrastructure.
• Machine learning (ML) offers promising approaches for combating spam.

Purpose of the Study:

• To evaluate the performance of machine learning models for spam detection.
• To propose and validate an optimized stacking ensemble framework for enhanced spam identification.
• To demonstrate the superiority of ensemble methods over individual models.

Main Methods:

• Empirical analysis of machine learning models on public spam datasets.
• Development of a stacking ensemble framework integrating Naive Bayes (NBC), k-Nearest Neighbors (k-NN), Logistic Regression (LR), and XGBoost (XGBoost) base models.
• Grid search cross-validation with hyperparameter optimization for model tuning.

Main Results:

• Algorithmic ensemble methods demonstrated superior detection accuracy compared to individual models.
• The optimized stacking ensemble achieved a high accuracy of 99.79%.
• Statistically significant improvements were observed over baseline models and existing literature solutions.

Conclusions:

• Optimized ensemble methods represent a highly effective strategy for advanced spam detection.
• The proposed stacking ensemble framework offers a robust solution for enhancing cybersecurity against email-borne threats.
• Further research into ensemble techniques can bolster digital security and user experience.