Improving Online Security: A Deep Learning Model for Phishing URL Detection

Abstract Today, phishing attacks have become a sig-nificant threat to online security. These false fronts areused to mislead people and businesses into clicking onmalicious URLs, revealing their password, credit cardinformation, or personal data. Infiltrating fraudulentwebsites or fake apps: Cybercriminals are upping theirgame by creating more and more representative de-signs of what to the inexperienced user seems like a“trustworthy” site, where they can be misled to entertheir username and password. Various methods such asblacklist, whitelist, heuristic exchange features, and vi-sual similarity-based techniques are being proposed foranti-phishing. Modern browsers come to help and limitthe way users can be phished into a vicious agenda,but yet, users fall prey to these obvious attacks and re-veal their secret data. In earlier research, the authorsintroduced a machine-learning approach for detectingphishing websites, achieving an accuracy of 96.85% with87 features. In this paper, we present new phishingURL detection models based on a Deep Neural Net-work (DNN) using the same dataset and 87 featuresfrom the previous work. The proposed method achievesa higher accuracy of 99.43% with DNN. Additionally,cross-validation is employed to enhance robustness andimprove the speed of phishing detection