AI-Powered Fake News and Image Verification System Using Multimodal Deep Learning

Abstract

The rapid proliferation of digital misinformation, fueled by the accessibility of generative artificialintelligenceandsophisticatedmediamanipulationtools,posesacriticalthreattopublictrust and democratic discourse. Traditional fact-checking methods are manual, time-consuming, and struggletoscalewiththevelocityofonlinecontentsharing. Furthermore, modern misinformation is increasingly multimodal, combining doctored imagery (deepfakes)with deceptive, textual narratives.This study presents the design, implementation, and evaluation of an end-to-end AI-powered verification system that addresses these challenges using multimodal deep learning.The proposed system features a dual-engine architecture: a text verification pipeline that utilizes a transformer-based DistilRoBERTa model fine-tuned on the FEVER dataset, integrated with knowledgegraphreasoning,websearchcross-referencing,andtemporalconsistencychecks;and an image verification engine that detects AI-generated manipulations using convolutional neural networks, noise residue extraction, and metadata analysis. The system is unified by an Optical CharacterRecognition(OCR)module that extracts textual claims from news images to perform a jointtext-imageanalysis.ThebackendwasimplementedusingFastAPItodeliverhigh-performance asynchronous processing, whereas the frontend was constructed as a React dashboard providing real-timescanningfeedbackandaverdictconsole.Theexperimentalresultsindicatethatoursystem attains a high level of accuracy accuracy across a range of text and image benchmarks, thereby offering a practical and robust tool for real-world fact-checking applications.

Introduction

This paper presents a multimodal fake news and misinformation detection system that combines text verification, image authenticity analysis, and cross-modal reasoning to identify deceptive digital content. Traditional fake news detection methods focus on either text or images separately, but modern misinformation often combines misleading text with manipulated or AI-generated images, making single-modal approaches ineffective.

The proposed framework analyzes text claims, standalone images, and news articles containing both text and visuals. It uses OCR (Optical Character Recognition) to extract text from images, a DistilRoBERTa-based fact verification model trained on the FEVER dataset for claim validation, and an image verification engine that detects AI-generated or manipulated images through deep learning, sensor noise analysis, and metadata inspection. Results from text and image analysis are fused to generate a final credibility verdict.

The system addresses key challenges such as misinformation spread, deepfakes, and the limitations of manual fact-checking by providing an automated, real-time verification tool. Its modular architecture includes OCR extraction, text verification with knowledge graph and temporal reasoning, image authenticity analysis, and multimodal fusion.

The implementation uses FastAPI for the backend, React for the frontend, PyTorch and Hugging Face Transformers for machine learning, and EasyOCR for text extraction. Experimental results demonstrate successful real-time classification of fake and authentic text, images, and news articles, providing users with clear credibility scores and verdicts. Overall, the framework offers an efficient, scalable, and user-friendly solution for combating modern multimodal misinformation and fake news.

Conclusion

Wehavedesignedandimplementedanintegrated,multimodalfakenewsandimageverificationsystem. Bycombiningnaturallanguageprocessingmodels(DistilRoBERTafine-tunedonthe FEVER dataset), computer vision models (CNN classifiers and noise residues), and structured knowledge engines, the system provides a robust defense against digital misinformation. The FastAPI backend ensures fast, asynchronous analysis of complex pipelines, while the React dashboard offers a user-friendly, responsive interface. The modular design allows for the independent updating of individual classifiers as generative technologies evolve.Future research directions include the integration of larger, instruction- tuned LLMs for more nuanced fact-checking explanations, as well as the implementation of cryptographic watermarking parsers to detect emerging camera-signature standards.

References

[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. [2] J. Thorne, A. Vlachos, C. Christodoulopoulos and A. Mittal, \"FEVER: A Large-Scale Dataset for Fact Extraction and Verification,\" Proceedings of NAACL-HLT, pp. 809-819, 2018. [3] J. Devlin, M. W. Chang, K. Lee and K. Toutanova, \"BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding,\" Proceedings of NAACL-HLT, pp. 4171-4186, 2019. [4] P. Lewis, E. Perez, A. Piktus et al., \"Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks,\" Advances in Neural Information Processing Systems, vol. 33, pp. 9459-9474, 2020. [5] A. Rossler, D. Cozzolino, L. Verdoliva, C. Riess, J. Thies and M. Nießner, \"FaceForensics++: LearningtoDetectManipulatedFacialImages,\"ProceedingsoftheIEEEInternationalConference on Computer Vision (ICCV), pp. 1-11, 2019. [6] S. Y. Wang, O. Wang, R. Zhang, A. Owens and A. A. Efros, \"CNN-Generated Images Are Surprisingly Easy to Spot... for Now,\" Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 8695-8704, 2020. [7] T.Karras,S.LaineandT.Aila,\"AStyle-BasedGeneratorArchitectureforGenerativeAdversarial Networks,\" Proceedings of CVPR, pp. 4401-4410, 2019.

Copyright

Copyright © 2026 Prof. Poojitha N, Rajeshwari H C, Radhika , Hamsaveni P. 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.