Enhancing Spam Classification with a Transformer-Infused CNN Approach

Spam messages have become a persistent challenge in digital communication, necessitating advanced detection techniques. This paper investigates a deep learning hybrid model that includes Convolutional Neural Networks (CNNs) and Transformer-based architectures to improve spam detection performance. Based on datasets and strategies drawn from recent studies, the model shows considerable advances in classification performance. The study also compares various other existing techniques and highlights the advantage of deep learning models compared to conventional methods regarding accuracy, efficiency, and practical application. Additionally, we analyze the adversarial weaknesses in spam detection and suggest countermeasures for improving robustness. Keywords—Spam detection, Deep Learning, CNN, Transformers, Natural Language Processing (NLP), Machine Learning, Email Filtering, Cybersecurity, Adversarial Learning, Feature Engineering