Abstract:
The exponential growth of spam and phishing emails significantly threatens users' privacy, security, and productivity.
This research shows a deep learning model that demonstrated improved performance compared to similar state-of-the-art studies. Deep learning neural networks, with properly fine-tuned settings, can accurately sort emails into three categories: normal (ham), unwanted (spam), and deceptive (phishing). We use two popular techniques, Principal Component Analysis (PCA) and Particle Swarm Optimization (PSO), to select the essential features for email classification. Test results show that the PSO method outperforms the others, achieving an impressive 99.35% accuracy rate. The findings highlight the potential of deep learning algorithms for effective email filtering, helping to detect and reduce spam and phishing. Additionally, the study emphasizes the importance of feature selection in enhancing the performance of deep learning models for sorting emails as normal (ham), spam, or phishing.
