DeepFake Detection Using Convolutional Neural Networks

Abstract

DeepFakes, or facial modifications produced by artificial intelligence, pose serious threats to the integrity of digital content and public trust. This paper proposes a method founded on Convolutional Neural Networks (CNNs) for binary facial image classification, separating real from fake. A properly processed and prebalanced dataset is utilized, and multiple data augmentation techniques are applied to ensure the model\'s maximum generalization capability. The proposed CNN structure, although relatively simple, is a sequence of several convolutional and pooling layers followed by dense layers with dropout regularization applied. Model training is performed with binary cross-entropy loss and optimized using the Adam optimizer. The experiment outcomes indicate that the model performs effectively on unseen test samples, thus demonstrating its capability to detect manipulated facial images and thus contributing to the overall goal of media integrity assurance.

Introduction

The rapid advancement of generative models—particularly autoencoders, GANs, and encoder-decoder networks—has led to the widespread creation of DeepFakes, highly realistic synthetic facial media that pose serious risks in digital forensics, misinformation, political manipulation, and cybercrime. As DeepFakes become harder to distinguish from real content, reliable and efficient detection systems are increasingly essential.

Most current DeepFake detection systems rely on heavy pretrained models such as XceptionNet and EfficientNet. While accurate, these approaches demand high computational resources, limiting their practical deployment. To address this limitation, the study proposes a lightweight, interpretable CNN model built from scratch using Keras, trained on a cleaned and augmented dataset of real and manipulated face images. Data augmentation plays a key role in improving generalization and robustness, enabling the model to detect DeepFake artifacts effectively in static images. The system is designed for potential use in social media moderation, digital authentication, and content verification.

A comprehensive literature review highlights diverse detection approaches, including CNNs, hybrid CNN-LSTM models, Vision Transformers (ViTs), and multimodal (audio-visual) systems. Many studies report high accuracy—often above 95%—yet most depend on complex architectures, large pretrained networks, or video-based temporal models. Additionally, audio deepfake detection, though improving, remains challenging. Traditional ML methods such as SVM and Random Forest also show competitive accuracy with lower computational demands.

The research gap identified is the absence of lightweight, from-scratch CNN models tailored for resource-constrained environments. Existing models, though accurate, are computationally intensive and difficult to interpret. This study addresses that gap by developing a simple, efficient CNN trained with extensive data augmentation to achieve strong performance without transfer learning.

The methodology includes dataset preparation (with balanced real and fake classes), preprocessing (resizing and normalization), and a robust augmentation pipeline using rotations, shifts, shearing, zooming, and flips. These steps expand dataset variability, reduce overfitting, and enhance model stability and performance on unseen data.

Conclusion

This paper presents an approach for detecting DeepFake images based on a Convolutional Neural Network (CNN) which was developed and trained from scratch, independent of pretrained networks. Empirical tests verified the model\'s effectiveness in detecting inaccuracies characteristic of synthetic media like the addition of artifacts and facial abnormalities. The paper points out that even simple CNN architectures, provided they are properly trained, can potentially provide competitive efficiency in the prevention of actual-world DeepFakes, particularly in resource-constrained environments. While the present study is primarily centered on static images, the approach can be adapted for video-based DeepFake detection by adding temporal analysis. the increasing threat of synthetic media.

References

[1] Chitale, Madhura, Aakanksha Dhawale, Manushree Dubey, and Sunil Ghane. \"A Hybrid CNN-LSTM Approach for Deepfake Audio Detection.\" In 2024 3rd International Conference on Artificial Intelligence For Internet of Things (AIIoT), pp. 1-6. IEEE, 2024. [2] Tariq, S., S. Lee, and S. S. Woo. \"A convolutional lstm based residual network for deepfake video detection. arXiv 2020.\" arXiv preprint arXiv:2009.07480 (2020). [3] Al-Dulaimi, Omar Alfarouk Hadi Hasan, and Sefer Kurnaz. \"A hybrid CNN-LSTM approach for precision deepfake image detection based on transfer learning.\" Electronics 13, no. 9 (2024): 1662. [4] Tipper, Sarah, Hany F. Atlam, and Harjinder Singh Lallie. \"An Investigation into the Utilisation of CNN with LSTM for Video Deepfake Detection.\" Applied Sciences 14, no. 21 (2024): 9754. [5] Ali, Raza, Munir Kashif, and Almutairi Mubarak. \"A novel deep learning approach for deepfake image detection.\" Applied Sciences 12, no. 19 (2022): 9820. [6] Almutairi, Zaynab, and Hebah Elgibreen. \"A review of modern audio deepfake detection methods: challenges and future directions.\" Algorithms 15, no. 5 (2022): 155. [7] Guo, Yinlin, Haofan Huang, Xi Chen, He Zhao, and Yuehai Wang. \"Audio deepfake detection with self-supervised wavlm and multi-fusion attentive classifier.\" In ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 12702-12706. IEEE, 2024. [8] Parikh, Aman, Kristen Pereira, Pranav Kumar, and Kailas Devadkar. \"Audio-Visual Deepfake Detection System Using Multimodal Deep Learning.\" In 2023 3rd International Conference on Intelligent Technologies (CONIT), pp. 1-6. IEEE, 2023. [9] Doan, Thien-Phuc, Long Nguyen-Vu, Souhwan Jung, and Kihun Hong. \"BTS-E: Audio deepfake detection using breathing-talking-silence encoder.\" In ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 1-5. IEEE, 2023. [10] Taeb, Maryam, and Hongmei Chi. \"Comparison of deepfake detection techniques through deep learning.\" Journal of Cybersecurity and Privacy 2, no. 1 (2022): 89-106. [11] Soudy, Ahmed Hatem, Omnia Sayed, Hala Tag-Elser, Rewaa Ragab, Sohaila Mohsen, Tarek Mostafa, Amr A. Abohany, and Salwa O. Slim. \"Deepfake detection using convolutional vision transformers and convolutional neural networks.\" Neural Computing and Applications 36, no. 31 (2024): 19759-19775. [12] Yadav, Priti, Ishani Jaswal, Jaiprakash Maravi, Vibhash Choudhary, and Gargi Khanna. \"DeepFake detection using InceptionResNetV2 and LSTM.\" In International Conference on Emerging Technologies: AI, IoT, and CPS for Science Technology Applications. 2021. [13] Sudharson, S., Priyanka Kokil, D. Sasikala, and Penta Aravinda Swamy. \"Deepfake detection system using a hybrid model.\" In 2023 14th International Conference on Computing Communication and Networking Technologies (ICCCNT), pp. 1-6. IEEE, 2023. [14] Saini, Madan Lal, Arnav Patnaik, Dayal Chandra Sati, and Ratish Kumar. \"Deepfake Detection System Using Deep Neural Networks.\" In 2024 2nd International Conference on Computer, Communication and Control (IC4), pp. 1-5. IEEE, 2024. [15] Rana, Md Shohel, Mohammad Nur Nobi, Beddhu Murali, and Andrew H. Sung. \"Deepfake detection: A systematic literature review.\" IEEE access 10 (2022): 25494-25513. [16] Rajalaxmi, R. R., P. P. Sudharsana, A. M. Rithani, S. Preethika, P. Dhivakar, and E. Gothai. \"Deepfake detection using inception-resnet-v2 network.\" In 2023 7th International Conference on Computing Methodologies and Communication (ICCMC), pp. 580-586. IEEE, 2023. [17] Rana, Md Shohel, Beddhu Murali, and Andrew H. Sung. \"Deepfake detection using machine learning algorithms.\" In 2021 10th International Congress on Advanced Applied Informatics (IIAI-AAI), pp. 458-463. IEEE, 2021 [18] Agarwal, Harsh, and Ankur Singh. \"Deepfake detection using svm.\" In 2021 Second International Conference on Electronics and Sustainable Communication Systems (ICESC), pp. 1245-1249. IEEE, 2021. [19] Ashok, V., and Preetha Theresa Joy. \"Deepfake Detection Using XceptionNet.\" In 2023 IEEE International Conference on Recent Advances in Systems Science and Engineering (RASSE), pp. 1-5. IEEE, 2023. [20] Singh, Richa, K. Ashwini, B. Chandu Priya, and K. Pavan Kumar. \"Deepfake Face Extraction and Detection Using MTCNN-Vision Transformers.\" In 2024 Third International Conference on Distributed Computing and Electrical Circuits and Electronics (ICDCECE), pp. 01-08. IEEE, 2024. [21] Wang, Yufei, and Guangjun Liao. \"Deepfake Video Detection Based on Image Source Anomaly.\" In 2024 IEEE 2nd International Conference on Image Processing and Computer Applications (ICIPCA), pp. 397-401. IEEE, 2024.

Copyright

Copyright © 2025 Swati Shilaskar, Sanika Jadhav, Yuvraj Jadhav, Jai Ughade. 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.