Integrating AI and Semantic Web Technologies for Robust Phishing Detection in Virtual Realities

Abstract

Phishing detection is a critical challenge in virtual realities, where malicious activities can compromise user security. This paper presents a novel approach integrating AI and Semantic Web Technologies for robust phishing detection. The proposed model preprocesses text data and leverages a reduced six-layer BERT encoder to extract contextual embeddings. Outputs from BERT, including classifier, attention, and encoder layers, are combined with features derived from Semantic Web Technologies and a custom deep learning layer to form a unified representation. The concatenated features are passed to a linear layer for classification. Experiments demonstrate superior performance, achieving 95\% accuracy, 96\% F1-score, and a 0.99 ROC-AUC, outperforming standard machine learning models. This framework provides a reliable phishing detection solution for virtual environments.