Research in the International Journal of Information and Computer Security has looked at the potential for artificial intelligence (AI) systems to detect and mitigate cyberbullying on social media platforms. The work might be a useful tool for safeguarding users, especially children and adolescents.
Cyberbullying, defined as intentional, harmful behaviour conducted via digital means such as messages, images, or videos, has become a significant problem in online spaces. In its digital form, bullying can be persistent, ubiquitous, and beyond the reach of parents, teachers, or platform moderators. Its victims can experience anxiety and depression to suicidal thoughts. The challenge of addressing this kind of abuse at scale, across the billions of interactions generated daily on social media, represents an enormous technological challenge.
In this current work, the team has developed a hybrid deep-learning system that uses two distinct neural network architectures: a Convolutional Neural Network (CNN) and a Long Short-Term Memory (LSTM) network to meet the challenge. CNNs are typically used in image and pattern recognition tasks and in this case can identify specific features of text, such as bullying-related keywords or phrases. LSTMs, on the other hand, are designed to analyse sequences of data and interpret their context and the emotional content of the language used.
By combining the strengths of both CNNs and LSTMs, the team’s hybrid system can not only detect blatant insults but can also spot more subtle, context-dependent forms of harassment that might evade simpler detection systems.
The team has demonstrated that their system can handle vast streams of content in real time with a high degree of accuracy and so be used to automate moderation of content where manual moderation would fail.
While no algorithm can wholly replace human judgement or solve the more profound social roots of online abuse, this research represents a new tool in the technological toolkit for combat cyberbullying.
Geetha, R., Jebamalar, G.B., Vignesh, B.G.D., Kamalanaban, E. and Doss, S. (2025) ‘An efficient cyberbullying detection framework on social media platforms using a hybrid deep learning model’, Int. J. Information and Computer Security, Vol. 26, No. 3, pp.255–271.