A Multi-Modal Deepfake Detection Framework with Explainable AI

View at publisher Summarise with VIRAT
Authors: G. Smilarubavathy, Kaviya D, Jane Afra
Cited by: 0
References: 14

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

The rapid rise of AI-generated media has posed major challenges for digital security and trust, particularly with the spread of realistic deepfakes. This research introduces a multimodal system to detect fake news and deepfakes using text, audio, and video analysis. Video frames are analyzed for spatial and temporal features using a ResNeXt50-LSTM network, while SVM with TF-IDF features and CNNs handle text. Audio content is classified using deep learning to identify synthetic speech. Explainable AI tools like LIME and Grad-CAM help users understand model decisions. The web-based system allows real-time checks of multimedia authenticity, with validation accuracies of 95.6% for text, 93.7% for audio, and 90% for video. The modular design enables flexible usage across applications, and interpretable outputs enhance user trust and regulatory compliance. Overall, the system advances scalable and explainable deepfake detection for strengthening digital media integrity.

Fields of Study & Subjects

Digital Media Forensic Detection Misinformation and Its Impacts Spam and Phishing Detection

Publication details

Journal / source
International Conference on Communication, Computing and Electronics Systems (ICCCES)
Pages
2139–2146
Publication date
21 January 2026
Language
EN