Fake News Detection Using Deep Learning

Abstract

In today’s digital world, social media has become the main source of information. But at the same time, fake news is spreading very fast and creating confusion among people. It can influence opinions, create panic, and even affect important decisions. So, detecting fake news has become very important. This project focuses on detecting fake news using Deep Learning techniques. The system collects data from different sources like news articles, social media posts, images, videos, and audio. It supports multiple languages, so it can identify fake content from different regions. Different models are used for different types of data. CNN models are used for images and videos, while RNN, LSTM, and Transformer models are used for text and audio analysis. Before processing, the data is cleaned and pre-processed using techniques like tokenization and feature extraction. The system is trained using standard datasets like Fake News Dataset and LIAR Dataset. After training, it can classify news as real or fake with high accuracy. To improve performance, the collected data undergoes preprocessing steps such as tokenization, stop-word removal, image resizing, frame extraction, and spectrogram generation for audio signals. Features extracted from different media types are combined through a fusion mechanism to generate accurate predictions.The results show that the system works effectively and helps in reducing misinformation. This project improves the reliability of information and supports safe and trustworthy communication in the digital environment.

Introduction

The text explains a deep learning–based system designed to detect fake news across multiple media formats, including text, images, videos, and audio.

It begins by highlighting how social media platforms have made fake news widespread and highly impactful in areas such as politics, healthcare, and education. Traditional manual fact-checking methods are slow and cannot handle the massive flow of online data, which has led to the adoption of AI and deep learning solutions for automated fake news detection.

The literature review shows that early machine learning methods like SVM and Naive Bayes relied on manual feature extraction and had limited accuracy. Modern approaches using CNNs, RNNs, LSTMs, and Transformer models like BERT perform better by capturing deeper contextual and semantic information. However, challenges such as high computational cost, dataset requirements, and difficulty detecting advanced deepfakes still remain.

The study aims to develop an intelligent system that can accurately classify content as real or fake, reduce misinformation, support multiple data types and languages, and handle real-time large-scale data. It also focuses on detecting advanced manipulations such as deepfake videos and synthetic audio.

The proposed system uses a multimodal deep learning framework. It collects data from various sources, preprocesses it (cleaning text, resizing images, extracting video frames, and processing audio), extracts features using NLP and CNN-based methods, and then classifies content using models like CNN, RNN/LSTM, and BERT. A fusion mechanism combines outputs from different media types to improve accuracy. The system then provides a final prediction (real, fake, or manipulated) along with confidence scores and explanations.

The architecture consists of input, preprocessing, feature extraction, deep learning processing, fusion, classification, and output layers. This makes the system scalable and suitable for real-time applications in media monitoring and fact-checking.

Finally, the research methodology describes data collection from datasets like LIAR and Fake News Net, followed by preprocessing, feature extraction, model training, testing, and evaluation to build an effective fake news detection system aimed at reducing misinformation.

Conclusion

This project presents a smart and efficient Fake News Detection Using Deep Learning system has been proposed to address the growing problem of misinformation in digital platforms. With the rapid increase of social media usage, fake news spreads quickly and influences public opinion, social harmony, and decision making processes. Traditional manual verification methods are slow and unable to manage the large volume of online content. Therefore, an automated deep learning-based solution is highly necessary. The proposed system uses a multimodal framework capable of analysing different forms of content such as text, images, videos, and audio. Advanced deep learning models including CNN, RNN, LSTM, and Transformer networks are used to identify false information, manipulated images, deepfake videos, and synthetic voice clips. By combining outputs from multiple models, the system achieves better prediction accuracy than single-source detection methods. Experimental analysis shows that the proposed model provides high accuracy, precision, recall, and reliable classification performance. The image and video models performed strongly in detecting manipulated media, while text and audio models effectively identified misleading language and fake speech patterns. This proves that the system is suitable for handling modern fake news challenges across multiple media formats. The system also includes strong security features such as data protection, secure access control, model safety, and defence against adversarial attacks. These features improve trustworthiness and ensure safe deployment in real-world environments. The architecture is scalable and can be integrated with websites, social media platforms, mobile applications, and fact-checking tools.Overall, the proposed fake news detection system provides an accurate, secure, and intelligent solution for reducing misinformation spread in society. It can help media organizations, government agencies, educational institutions, and online platforms maintain the credibility of digital information. In future, the system can be enhanced with realtime monitoring, multilingual expansion, explainable AI, and blockchain-based verification methods to build a safer and more trustworthy digital communication environment.

References

[1] W. Y. Wang, “LIAR: A Benchmark Dataset for Fake News Detection,” Proceedings of ACL, 2017. [2] J. Devlin, M. Chang, K. Lee, and K. Toutanova, “BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding,” NAACL, 2018. [3] Kaggle, “Fake News Dataset,”Available: https://www.kaggle.com Google AI,“DeepFake Detection Challenge Dataset,” 2020. [4] L. Hu, S. Wei, Z. Zhao, and B. Wu, “Deep learning for fake news detection: A comprehensive survey,” AI Open, vol. 3, pp. 133–155, 2022. [5] S. Singhania, N. Fernandez, and S. Rao, “3HAN: A Deep Neural Network for Fake News Detection,” arXiv preprint arXiv:2306.12014, 2023. [6] [1] K. Shu, A. Sliva, S. Wang, J. Tang, and H. Liu, “Fake news detection on social media: A data mining perspective,” ACM SIGKDD Explorations Newsletter, vol. 19, no. 1, pp. 22–36, 2017. [7] Y. Guo, S. Ji, X. Fang, D. K. W. Chiu, and H. Leung, “An Unsupervised Fake News Detection Framework Based on Structural Contrastive Learning,” Cybersecurity, vol. 8, no. 41, 2025.

Copyright

Copyright © 2026 Mani Kumari Veeramallu , Swapna Veernala, Raj Kiran Kopuri, Yesubabu Siramdasu, Akanksha Bulla. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.