Machine learning algorithm for detecting suspicious email messages using Natural Language Processing NLP

Abstract

Email-based cyber threats continue to evolve, presenting challenges to traditional security measures that struggle to keep pace with sophisticated attack methods. This study introduces a machine learning-driven framework that integrates Natural Language Processing (NLP) with a Support Vector Classifier (SVC) to enhance email threat detection. The proposed model utilizes feature extraction techniques, including one-hot encoding, TF-IDF vectorization, and BERT embeddings, to capture linguistic patterns indicative of malicious emails. The model was trained and evaluated on a benchmark dataset, achieving an accuracy of 98.65%, demonstrating superior performance over conventional spam detection systems. The results show a significant reduction in false positives and false negatives, improving email security and reliability. This research contributes to the field by presenting an optimized, scalable solution that can adapt to emerging cyber threats and enhance automated email filtering mechanisms.