Abstract—Cyberbullying and online threat detection are the new vulnerabilities of cybersecurity, and the conventional tools are unable to keep pace with the development of harmful content. The paper discusses HYBERT-X as a hybrid deep learning model that integrates Bidirectional Long Short-Term Memory (BiLSTM), Graph Convolutional Network (GCN) based learning, Bidirectional Encoder Representations from Transformers (BERT) and Extreme Gradient Boosting (XGBoost) to detect threats with high accuracy and interpretability. The proposed framework uses an ensemble classification layer to boost the accuracy and interpretability of detection based on semantic, temporal, and structural representation. The experimental assessment done on the typical benchmark datasets has shown that HYBERT-X is up to 95.53% accurate in comparison with the currently available baseline models both in generalization ability and explainability. The findings suggest that this work is computationally effective, scalable, and can be adapted to real-time cybersecurity applications and is applicable to next-generation cybersecurity systems against online threats and cyberbullying.
 
Keywords—cyberbullying detection, cybersecurity, hybrid deep learning, Bidirectional Encoder Representations from Transformers (BERT), Extreme Gradient Boosting (XGBoost)


Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).