International Journal of Scientific Research and Engineering Development

The rapid growth of social media platforms has led to a significant increase in automated accounts, commonly known as bots, which are often used for misinformation, spam, political manipulation, and malicious campaigns. Traditional bot detection methods primarily rely on textual analysis or metadata-based features, which are increasingly ineffective against sophisticated bots that mimic human behavior using advanced artificial intelligence techniques. Moreover, modern bots leverage advanced language models, synthetic profile images, and coordinated interaction strategies to evade conventional detection mechanisms. The dynamic and evolving nature of these bots presents substantial challenges, as single-modality detection systems often fail to capture the complex behavioral and structural patterns exhibited across different data sources. This paper proposes a multimodal deep learning framework for social media bot detection that integrates textual content, visual information, user metadata, and network interaction features. Textual features are extracted using transformer-based language models, visual features are obtained through convolutional neural networks, and structural patterns are captured using graph neural networks. The learned modality-specific embeddings are fused using an attention-based mechanism to enhance classification performance. Experimental evaluation on benchmark social bot datasets demonstrates that the proposed multimodal approach significantly outperforms unimodal baselines in terms of accuracy, precision, recall, and F1-score. The results highlight the effectiveness.

Shyni M,"Identifying Bots in Social Media Using Multimodal Deep Learning" International Journal of Scientific Research and Engineering Development, V9(1): Page(962-967) Jan-Feb 2026. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.