Leveraging Memetic Algorithm and Machine Learning Methods for Email-Based Spam Detection

Email is the most frequently utilized method of communication between people nowadays. This is because most people in institutions, banks, universities, hospitals, and others depend mainly on e-mail to communicate with each other and transfer information and data effectively and quickly. However, it has been observed in recent years that fraudsters have become smarter in defrauding people, especially via email, and deceiving them by creating fake accounts and impersonation, which has led to an increase in cybercrimes such as phishing and impersonating others to steal their money and bank account data. In this paper, we use a machine learning (ML) based approach in email header analysis as a powerful tool for detecting phishing and spam emails. Specifically, the efficacy of three different machine learning algorithms has been trained and tested in the model, which are: Naive Bayes Classifier (NB Classifier), Multi-Layer Perceptron Classifier (MLP Classifier), and Random Forest. Consequently, our approach, which focuses on using genetic algorithms and simulated annealing for feature selection, effectively detects spam, ham, and phishing emails with high accuracy, precision, and recall. Also, it achieves a 99.69% accuracy for spam detection and a 99.12% accuracy for phishing detection. The findings are compared to some state-of-the-art models and indicate the supremacy of our model.