Home / Current Issue / Paper 1710909
Deepfake Detection System Using Deep Learning
Subject area: Science,Engineering and Technology · Area of research: Artificial Intelligence, Deep Learning
Abstract
Deepfakes are realistic synthetic images and videos generated by advanced generative models such as GANs and neural rendering pipelines. They present serious threats to privacy, trust, and public discourse by enabling impersonation, misinformation, and malicious content creation. This paper proposes a robust deepfake detection system that integrates spatial, temporal, and frequency-domain analyses using deep learning. The pipeline includes face detection and alignment, frame-level CNN feature extraction, frequency residual analysis, and temporal modeling with recurrent layers. An ensemble fusion strategy combines complementary cues to improve detection under compression and post-processing. Experimental evaluation on public benchmarks demonstrates strong accuracy, recall, and AUC-ROC, highlighting the system?s potential for deployment in content moderation workflows.
Keywords
Deepfake Detection, Deep Learning, CNN, GAN, FaceForensics++, Frequency Analysis, Temporal Modeling
References
[1] I. Goodfellow et al., Generative Adversarial Nets,¨ NIPS, 2014.¨
[2] A. Rossler, D. Cozzolino, L. Verdoliva, C. Riess, J. Thies, M. Nie,¨ FaceForensics++: Learning tDetecto Manipulated Facial Images,¨ ¨ICCV, 2019.
[3] F. Chollet, Xception: Deep Learning with Depthwise Separable Convo-¨ lutions,CVPR, 2017.¨
[4] A. Afchar, V. Nozick, J. Yamagishi, I. Echizen, MesoNet: a Compact¨ Facial Video Forgery Detection Network,WIFS, 2018.¨
[5] H. T. Nguyen, J. Yamagishi, I. Echizen, Use of a Capsule Network to¨ Detect Fake Images and Videos,¨ICASSP, 2019.
[6] A. Li and S. Lyu, Exposing DeepFake Videos By Detecting Face¨ Warping Artifacts,CVPRW, 2019.¨
How to cite this paper
@article{1710909,
author = {Adittya Mondal, Bivash Mazumder, Ranganath, Bhagyashri Wakde},
title = {Deepfake Detection System Using Deep Learning},
journal = {Iconic Research And Engineering Journals},
year = {2025},
volume = {9},
number = {3},
pages = {1361-1363},
issn = {2456-8880},
url = {https://www.irejournals.com/formatedpaper/1710909.pdf},
abstract = {Deepfakes are realistic synthetic images and videos generated by advanced generative models such as GANs and neural rendering pipelines. They present serious threats to privacy, trust, and public discourse by enabling impersonation, misinformation, and malicious content creation. This paper proposes a robust deepfake detection system that integrates spatial, temporal, and frequency-domain analyses using deep learning. The pipeline includes face detection and alignment, frame-level CNN feature extraction, frequency residual analysis, and temporal modeling with recurrent layers. An ensemble fusion strategy combines complementary cues to improve detection under compression and post-processing. Experimental evaluation on public benchmarks demonstrates strong accuracy, recall, and AUC-ROC, highlighting the system?s potential for deployment in content moderation workflows.},
keywords = {Deepfake Detection, Deep Learning, CNN, GAN, FaceForensics++, Frequency Analysis, Temporal Modeling},
month = {September},
}