A Deep Learning Approach to Detecting and Preventing Misinformation in Online Media

Abstract

The increasing trend of sharing information on the internet has resulted in an unprecedented increase in the spread of fake news on digital media platforms. The automatic detection of such fake news has become a challenge for researchers and policymakers. This paper proposes an AI-based framework for fake news detection using natural language processing (NLP) and deep learning techniques. By using transformer-based models like BERT and RoBERTa, along with linguistic and semantic feature analysis, our proposed model has shown robust classification performance on benchmark datasets such as LIAR and Fake- NewsNet. The experimental results have shown that the proposed model performs better than the traditional machine learning models, achieving an accuracy of 93.4 in binary classification tasks. Furthermore, the study explores the interpretability of deep learning models through attention visualization, providing insights into how AI systems can make explainable judgments about news credibility. The findings contribute to the develop- ment of scalable and explainable AI solutions for combating misinformation in digital ecosystems.