Title: Detection and enhanced security against cyberbullying on social media using a hybrid deep learning framework

Authors: Moushmee Milind Kuri; Ganesh R. Pathak

Addresses: Department of Computer Science and Engineering, MIT-ADT University, Pune, Maharashtra, 412201, India ' Department of Computer Science and Engineering, MIT-ADT University, Pune, Maharashtra, 412201, India

Abstract: Cyberbullying on social media is a widespread issue, prompting the need for an optimised hybrid deep learning (DL) framework for effective detection and prevention. Due to its complex and subjective nature, cyberbullying is difficult to label for training machine learning models, and manual annotation can be time-consuming and biased. This research develops a novel DL framework for detecting cyberbullying by pre-processing textual data through cleaning, tokenisation, and leveraging advanced techniques like BERT and Roberta for feature extraction. Roberta, with its refined training approach, outperforms BERT in sentiment analysis tasks. Findings reveal that non-bullying tweets have 98% mention levels while bullying tweets show consistently elevated levels. By combining BERT/Roberta for sentiment analysis, BIGRU networks for subtle cues, and CNNs for keyword identification, the framework enhances cyberbullying detection. Future applications include real-time monitoring, preventive measures, and further advancements in natural language understanding for social media interactions, promoting safer online environments.

Keywords: hybrid deep learning framework; cyberbullying; social media platform; robustly optimised BERT pertaining approach; convolutional neural networks; CNNs; natural language processing; NLP.

DOI: 10.1504/IJESDF.2026.156142

International Journal of Electronic Security and Digital Forensics, 2026 Vol.18 No.5, pp.609 - 628

Received: 06 Sep 2024
Accepted: 31 Oct 2024
Published online: 07 Sep 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article