An Explainable Deep Learning Framework for Robust Deepfake Detection in Video Streams

DOI

10.2991/978-94-6239-697-5_11How to use a DOI?

Keywords

Deepfake detection; explainable artificial Intelligence; CNN-LSTM; Grad-CAM; video forensics

Abstract

The creation of deepfakes with the help of state-of-the-art deep learning methods has become a significant challenge to the authenticity of digital media, cybersecurity and trust in the society. As the Generative Adversarial Networks (GANs) and diffusion models of synthesis continue to progress quickly, fake videos are becoming more and more difficult to be detected as fake. Even though detection methods based on deep learning have shown high accuracy, their non-transparency prevents their application in the real world in forensics and legal fields. The proposed paper offers an Explainable Deep Learning Framework of Robust Deepfake Detection in Video Streams that is based on integrating spatial-temporal learning with post-hoc explainability. The suggested model combines a Convolutional Neural Network (CNN) based on spatial artifact extraction and a Long Short-Term Memory (LSTM) network based on time inconsistency modelling on video frames. Gradient-weighted Class Activation Mapping (Grad-CAM) is also used to provide model decisions explanations visually to enhance interpretability. Four benchmark datasets, including FaceForensics++, Celeb-DF, DFDC and DeeperForensics-1.0, are experimented on with a variety of measures. The proposed solution is more accurate and robust and has a better generalization performance as compared to the baseline CNN and CNN-RNN. Moreover, the explainability analysis shows that the model always concentrates on manipulated regions of the face including eye boundaries, mouth contours, and facial boundaries. The findings support the fact that explainable AI should be incorporated into deepfake detection because it increases the level of trust and forensic utility.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Amitoj Kaur
AU  - Shivangi Sharma
PY  - 2026
DA  - 2026/06/04
TI  - An Explainable Deep Learning Framework for Robust Deepfake Detection in Video Streams
BT  - Proceedings of the Conference on Bridging Engineering Disciplines with AI and Machine Learning (BEDAIML 2026)
PB  - Atlantis Press
SP  - 117
EP  - 124
SN  - 1951-6851
UR  - https://doi.org/10.2991/978-94-6239-697-5_11
DO  - 10.2991/978-94-6239-697-5_11
ID  - Kaur2026
ER  -