Deepfake technology, using sophisticated generative models like GANs and diffusion networks, has introduced the
world to super-realistic manipulated videos and images. Although these artificial media carry substantial risks to politics,
social networks, and cyber security, the current deep fake detectors are also often observed as black-box models and limits
their forensic applicability such as in legal cases where transparency is required. In this paper, we propose a deepfake detection
framework with explainability using off-the-shelf detection backbones, Grad-CAM, SHAP and attention-based heatmaps to
identify the manipulated areas of the face. We perform experiments on two challenging benchmark data sets: the Celeb-DF v2
(with high-quality face swaps) and Wild Deepfake (cross-pose, cross-scene, etc.) detection tasks. Experimental results show
that the proposed framework brings competitive detection performance and provides understandable outputs, which localize
manipulation artifacts. Experimental results on comparing with non-explainable baselines demonstrate that the detectors
enhanced with XAI can well suit for improving analyst confidence and forensic reliability. In addition to simulated data, we
illustrate several case studies of both successful detections and failure cases for insights into model improvement. The results
demonstrate that explainability contributes not only to the trustworthiness of a detection, but also to the admissibility in court
of AI-generated evidence, opening the door for practical and analyst-friendly as well as legally solid investigations systems
covering deepfake.
[1] N. Mansoor and A. I. Iliev, “Explainable AI for DeepFake Detection,” Applied Sciences, vol. 15, no. 2, p. 725, Jan. 2025.
doi: 10.3390/app15020725.
[2] P. Yu, J. Fei, H. Gao, X. Feng, Z. Xia, and C. H. Chang, “Unlocking the Capabilities of Vision-Language Models for
Generalizable and Explainable Deepfake Detection,” arXiv preprint arXiv:2503.14853, Mar. 2025.
[3] R. Kundu, A. Balachandran, and A. K. Roy-Chowdhury, “TruthLens: Explainable DeepFake Detection for Face
Manipulated and Fully Synthetic Data,” arXiv preprint arXiv:2503.15867, 2025.
[4] K. Tsigos, E. Apostolidis, S. Baxevanakis, S. Papadopoulos, and V. Mezaris, “Towards Quantitative Evaluation of
Explainable AI Methods for Deepfake Detection,” arXiv preprint arXiv:2404.18649, Apr. 2024.
[5] M. S. Momin, S. Das, A. Gupta, and P. Kumar, “Explainable Deepfake Detection Across Different Modalities,” Digital
Investigation, vol. 46, p. 301005, 2025. doi: 10.1016/j.diin.2025.301005.
[6] A. Alharbi, R. Ahmad, and N. Dey, “A Systematic Survey on Explainable AI Applied to Fake News Detection,” Computers
in Human Behavior Reports, vol. 12, p. 100303, 2023. doi: 10.1016/j.chbr.2023.100303.
[7] S. Venkateswarulu and A. Srinagesh, “DeepExplain: Enhancing DeepFake Detection through Transparent and Explainable
AI Model,” Informatica, vol. 47, no. 3, pp. 321–332, 2023. doi: 10.31449/inf.v47i3.5792.
[8] S. Mathews, R. Ranjan, A. Bansal, and S. Tiwari, “An Explainable Deepfake Detection Framework on a Novel Large-Scale
Deepfake Image Dataset,” Journal of Intelligent & Fuzzy Systems, vol. 44, no. 2, pp. 2157–2170, 2023. doi: 10.3233/JIFS
223456.
[9] M. M. Alwateer, “Explainable Deep Fake Framework for Images Creation and Classification,” Journal of Computer and
Communications, vol. 12, no. 5, pp. 86–101, May 2024. doi: 10.4236/jcc.2024.125006.
[10] A. Gupta, K. Mehta, and P. Jain, “A Survey on Multimedia-Enabled Deepfake Detection: State-of-the-Art, Challenges and
Future Directions,” Education and Information Technologies, vol. 30, no. 2, pp. 1123–1148, 2025. doi: 10.1007/s10639-025
12345-6.
[11] A. Verdoliva, “Deepfake Detection: A Comprehensive Survey from the Reliability Perspective,” arXiv preprint
arXiv:2211.10881, revised 2022.
[12] H. Lee, J. Kim, and S. Park, “A Comprehensive Survey with Critical Analysis for Deepfake Speech Detection and
Interpretability,” arXiv preprint arXiv:2409.15180, Sept. 2024.
[13] J. Wu, M. Zhang, and Y. Zhao, “XAI-Based Detection of Adversarial Attacks on Deepfake Detectors,” arXiv preprint
arXiv:2403.02955, Mar. 2024.
[14] T. Karras, M. Aittala, S. Laine, E. Härkönen, J. Hellsten, J. Lehtinen, and T. Aila, “Alias-Free Generative Adversarial
Networks,” in Proc. NeurIPS, vol. 34, pp. 852–863, 2021. doi: 10.48550/arXiv.2106.12423.
[15] J. Ho, A. Jain, and P. Abbeel, “Denoising Diffusion Probabilistic Models,” in Proc. NeurIPS, vol. 33, pp. 6840–6851, 2020
(widely cited through 2021–2025). doi: 10.48550/arXiv.2006.11239.
