PHISHING DETECTION IN CYBERSECURITY USING DEEP LEARNING: A SYSTEMATIC SURVEY OF METHODS AND APPLICATIONS

Abstract

Phishing is a sort of cyberattack that has gained a lot of attention recently. Every day, hundreds of consumers utilizing various online services are targeted through copycat websites. Phishers target naive users by sending them malicious emails, social media communications, or text messages with the goal of stealing sensitive information like login passwords. In order to trick their victims into divulging critical information, cybercriminals craft phishing URLs that mimic legitimate websites. Traditional detection methods are overwhelmed by the complex and dynamic nature of these threats, prompting a boom in research utilizing deep learning (DL) technologies. This paper showcases models according to their resistance and offers a comprehensive review of DL-based phishing detection approaches for the goal of recognizing phishing emails, URLs, and websites. Their analysis covers real-world applications on several platforms, including mobile, email gateways, and browsers. It tackles present issues such as model interpretability, real-time detection delay, data privacy, and evasion methods. Furthermore, this study identifies critical research gaps in dataset diversity and adversarial robustness, advocating for lightweight, interpretable, and adaptive DL models to enhance phishing detection systems. Their findings contribute to shaping future research directions toward more secure and resilient cyber defense mechanisms.