Intelligent Phishing Email Detection with Multi-Feature Analysis (IPED-MFA)

The rise in bulk email phishing attempts has emerged as a serious cyber-risk for businesses given the pervasive usage of email as a primary form of communication. Despite the existence of manual and technical countermeasures, phishing emails still pose a risk of deceiving employees. Machine learning has played a vital role in improving the detection of many types of cyberattacks, including email phishing. In this paper, the utilization of a machine learning-based IDS model to detect phishing emails before they reach inboxes is proposed. A review of phishing email detection techniques and algorithms is presented, and a publicly available dataset of ham and phishing emails is discussed and used. Multiple machine learning classifiers were deployed (RF, SVM, Adaboost, Logistic Regression, and KNN), and their evaluation scores were generated and analysed. Using recursive feature elimination and multi-feature analysis, Random Forest outperformed the other classifiers with an accuracy of 98.6%, closely followed by Adaboost with 98.1%.