An In-Depth Review of ML and DL Approaches for Phishing Email Identification and Mitigation

Abstract

Businesses lose money, and consumers become annoyed by phishing emails, which are a major problem on the Internet these days. A form of artificial intelligence called machine learning has been shown to be a useful tool for spotting email threats; yet, there is currently no comprehensive answer to the issue of phishing email filtering. This study fills this gap by providing an extensive analysis of the most recent approaches to phishing detection, ranging from conventional ML methods to advance DL frameworks.  The expanding use of artificial intelligence, namely ML and DL, in phishing email detection and mitigation is examined in this review. ML models with efficient feature selection and classification, such as Random Forest, Decision Tree, and SVM, exhibit high accuracy, while DL architectures like DNNs, CNNs, RNNs, and LSTMs excel in automated feature extraction and sequential data analysis, offering scalability and adaptability for real-time detection. When Natural Language Processing (NLP) and hybrid models are used, DL techniques demonstrate encouraging accuracy and resilience, lowering false positives and improving security. The study emphasizes the effectiveness of DL and ensemble models in thwarting advanced phishing efforts as well as the significance of AI-driven layered defenses in stopping the phishing lifecycle.