In today’s digital ecosystem, fake news has evolved into a potent cyber threat vector, frequently exploited in psychological operations and information warfare. This paper presents a novel intent-based approach to fake news classification, framing it as an adversarial cyber activity. Unlike conventional binary detection systems, we introduce a multi-class taxonomy—comprising misinformation, disinformation, propaganda, and bot-driven content—to categorize fake news based on underlying malicious intent. Although graph-based modeling using Graph Neural Networks (GNNs) is explored as a future direction, the current implementation focuses on fine-tuning RoBERTa, a transformer-based language model, for title-level classification. The dataset was curated and labeled to reflect real-world threat scenarios, and the model was trained using standard NLP preprocessing and stratified evaluation. Experimental results demonstrate the model’s effectiveness, achieving a test accuracy of [insert actual