Email phishing remains a persistent cybersecurity threat, undermining communication systems and exploiting unsuspecting users. This study proposes a machine learning based approach to detect phishing emails using the CRISP-DM methodology. A dataset of 18,650 emails obtained from Kaggle was analyzed with multiple algorithms, including Naïve Bayes and Random Forest, to evaluate their effectiveness in distinguishing between legitimate and malicious messages. Performance was assessed using accuracy, precision, recall, and F1-score. The Random Forest model achieved the highest performance with 98.30% accuracy, 0.96 precision, 1.00 recall, and an F1-score of 0.90. In contrast, Naïve Bayes produced a lower accuracy of 53.80% but achieved perfect recall, highlighting the importance of comparing algorithms to uncover their respective strengths and limitations. Future work will focus on refining the proposed solution and extending its applicability to address region-specific challenges, particularly in Nigeria, where phishing scams such as “yahoo yahoo” are increasingly prevalent. This research contributes to the advancement of robust email security strategies against evolving phishing threats.
Orokpo et al. (Sat,) studied this question.