WillRo: A Deep Learning App for Potential Phishing Threat Detection

Abstract

This article presents an application called WillRo App, designed to detect potential phishing by analyzing website screenshots in real time. The system integrated Robotic Process Automation (RPA) to capture screenshots, and the YOLOv5 deep learning model in order to classify phishing and no-phishing content. The results demonstrated a precision of 85.80%, a recall of 93.00%, a mAP@0.5 of 66.60%, and a mAP@0.5 -0.95 of 32.70%. These values showed a reliable detection performance, making WillRo a possible model for phishing detection. Future work should focus on improving the model with additional features to increase its accuracy.