TrueNews: AI Powered Detection of Manipulated Text and Images

Abstract

The proliferation of fake news across digital platforms has become a critical issue, leading to widespread misinformation with significant s ocietal i mplications. T his s urvey p aper presents a comprehensive review of recent advancements in fake news detection, leveraging machine learning (ML), deep learning (DL), and natural language processing (NLP) techniques. The reviewed studies cover diverse approaches, ranging from content-based methods to the integration of social context, multimedia, and knowledge-enhanced models. Traditional machine learning algorithms such as Random Forest, Support Vector Machines, and logistic regression are commonly employed for binary classification tasks, using features derived from linguistic patterns, source credibility, and metadata. In addition, enhanced models such as bidirectional LSTM-RNN and hybrid CNN-LSTM architectures, coupled with FastText embeddings, demonstrate significant improvements in detecting fake news in real-time scenarios and across multimedia-rich datasets. The integration of social network features alongside textual content is a growing focus, where user behavior and social capital contribute to a more comprehensive fake news detection process. Transformer-based models, such as BERT, XLNet, and RoBERTa, show promising results in handling syntactic and semantic complexities, outperforming traditional RNN-based methods. Additionally, knowledge-augmented models utilizing large-scale open knowledge graphs offer a novel direction for multi-modal fake news verification by enhancing the model’s understanding of both textual and visual content. The survey also highlights the growing trend toward explainable AI (XAI) in fake news detection, providing transparency and interpretability in decision-making. By employing state-of-the-art models alongside regularization techniques and hyperparameter optimization, these studies collectively strive to address key challenges in fake news detection, including early identification, data scarcity, and model generalization. This survey concludes by emphasizing the need for continued innovation in scalable and robust fake news detection systems, integrating diverse data modalities and ensuring real-time detection capabilities across a range of online platforms.