The pervasive use of social media has led to an alarming rise in cyberbullying, particularly among younger users, posing significant threats to mental and emotional well-being. Traditional approaches to cyberbullying detection have predominantly focused on textual analysis, which often fails to capture the multi-modal nature of bullying content, including images, videos, and contextual metadata. To address this limitation, we propose a novel multi-modal cyberbullying detection framework that integrates textual, visual, and contextual information to identify bullying behavior more effectively. Our approach leverages advanced deep learning techniques, including Hierarchical Attention Networks (HAN) and Bidirectional Long Short-Term Memory (BiLSTM) networks, to model the complex interactions between different modalities. The framework processes user-generated posts, combining text and image data, along with metadata such as timestamps and user interactions, to predict whether a post constitutes cyberbullying. This research provides a robust, scalable solution for identifying and mitigating harmful content on social networks.