Performance Analysis of Deep Learning Architectures in Classifying Fake and Real Images

Arya Faisal Akbar; Putu Desiana Wulaning Ayu; Dandy Pramana Hostiadi

doi:10.30595/juita.v13i2.25790

Authors

Arya Faisal Akbar Institut Teknologi dan Bisnis STIKOM Bali
Putu Desiana Wulaning Ayu Institut Teknologi dan Bisnis STIKOM Bali
Dandy Pramana Hostiadi Institut Teknologi dan Bisnis STIKOM Bali

DOI:

https://doi.org/10.30595/juita.v13i2.25790

Keywords:

CIFAKE Dataset, Real and Fake Face Detection Dataset, Image Classification, Dynamic Dropout, Deep Learning Architectures

Abstract

The advancements in artificial intelligence (AI) have significantly enhanced image manipulation capabilities, yet they also raise concerns regarding the proliferation of synthetic images. This study investigates the impact of Dynamic Dropout in optimizing deep learning models, including ResNet-101, DenseNet-201, VGG-19, and AlexNet, for classifying real and synthetic images using the CIFAKE and Real and Fake Face datasets. Dynamic Dropout was applied with a progressively increasing rate from 20 percent to 50 percent to enhance training stability and generalization. The results indicate that the optimal configuration consisting of 15 epochs, the Adam optimizer, and Dynamic Dropout consistently outperformed Static Dropout across all models. DenseNet-201 with Dynamic Dropout achieved the highest accuracy of 97.42%, with a precision of 97.33%, recall of 97.58%, and an F1-score of 97.45%. ResNet-101 and VGG-19 exhibited enhanced training stability, while AlexNet proved efficient for lightweight datasets. The Adam optimizer outperformed Nadam, offering greater stability in deeper architectures. Additionally, the 15th epoch was identified as the optimal training duration, balancing accuracy and overfitting mitigation. These findings underscore the importance of selecting optimal training configurations to enhance deep learning performance. Future research should explore adaptive dropout strategies, assess scalability on diverse datasets, and validate these techniques in real-world applications such as digital forensics and AI-generated content detection

References

[1] R. Angelika Septi Rahayu and H. Santoso, “Analysis of Fake Face Images: Detecting the Authenticity of Manipulated Images Using Variational Autoencoder Methods and Deep Neural Network Forensics,” Sibatik J. | Vol., vol. 2, no. 9, pp. 2701–2726, 2023, [Online]. Available: https://publish.ojs-indonesia.com/index.php/SIBATIK.

[2] S. C. Hanebuth, D. Kalokitis, and A. Cedrone, “Structured Methodology for the Investigation of Contact Voltages,” IEEE Power Energy Technol. Syst. J., vol. 1, no. November 2014, pp. 1–11, 2015, doi: 10.1109/jpets.2014.2363403.

[3] L. Alzubaidi et al., Review of deep learning: concepts, CNN architectures, challenges, applications, future directions, vol. 8, no. 1. Springer International Publishing, 2021.

[4] S. Tammina, “Transfer learning using VGG-16 with Deep Convolutional Neural Network for Classifying Images,” Int. J. Sci. Res. Publ., vol. 9, no. 10, p. p9420, 2019, doi: 10.29322/ijsrp.9.10.2019.p9420.

[5] J. Liang, “Image classification based on RESNET,” J. Phys. Conf. Ser., vol. 1634, no. 1, 2020, doi: 10.1088/1742-6596/1634/1/012110.

[6] M. Tripathi, “Analysis of Convolutional Neural Network based Image Classification Techniques,” J. Innov. Image Process., vol. 3, no. 2, pp. 100–117, 2021, doi: 10.36548/jiip.2021.2.003.

[7] J. J. Bird and A. Lotfi, “CIFAKE: Image Classification and Explainable Identification of AI-Generated Synthetic Images,” IEEE Access, vol. 12, no. December 2023, pp. 15642–15650, 2024, doi: 10.1109/ACCESS.2024.3356122.

[8] Y. Wang, Y. Hao, and A. X. Cong, “Harnessing Machine Learning for Discerning AI-Generated Synthetic Images,” 2024, [Online]. Available: http://arxiv.org/abs/2401.07358.

[9] P. D. W. Ayu, G. A. Pradipta, I. M. D. Susila, D. P. Hostiadi, and M. Liandana, “Deep Learning Based Detection and Classification of Amniotic Fluid Echogenicity Type for Enhanced Prenatal Diagnosis,” Int. J. Intell. Eng. Syst., vol. 18, no. 1, pp. 246–267, 2025, doi: 10.22266/ijies2025.0229.18.

[10] C. Jiang, C. Jiang, D. Chen, and F. Hu, “Densely Connected Neural Networks for Nonlinear Regression,” Entropy, vol. 24, no. 7, pp. 0–9, 2022, doi: 10.3390/e24070876.

[11] I. W. Jepriana, I. G. Bintang, A. Budaya, G. A. Pradipta, P. Desiana, and W. Ayu, “Combining Oversampling and Pretrained Feature Extractor For Classification Diabetic Foot Uclear Thermogram Images,” vol. 12, no. 2, pp. 267–277, 2024.

[12] A. Krizhevsky, I. Sutskever, and G. E. Hinton, “ImageNet Classification with Deep Convolutional Neural Networks Alex,” Encycl. Mov. Disord. Three-Volume Set, pp. V2-257-V2-259, 2010, doi: 10.1016/B978-0-12-374105-9.00493-7.

[13] B. Huang et al., “Implicit Identity Driven Deepfake Face Swapping Detection,” pp. 4490–4499, 2023, doi: 10.1109/cvpr52729.2023.00436.

[14] E. al. Vismay Vora, “A Multimodal Approach for Detecting AI Generated Content using BERT and CNN,” Int. J. Recent Innov. Trends Comput. Commun., vol. 11, no. 9, pp. 691–701, 2023, doi: 10.17762/ijritcc.v11i9.8861.

[15] G. E. Bartos, Ó. Egyetem, S. Ö. Akyol, and S. Akyol, “Deep Learning for Image Authentication: A Comparative Study on Real and AI-Generated Image Classification,” no. November, 2023, [Online]. Available: https://www.researchgate.net/publication/375952278.

[16] J. Mallet, L. Pryor, R. Dave, and M. Vanamala, “Deepfake Detection Analyzing Hybrid Dataset Utilizing CNN and SVM,” ACM Int. Conf. Proceeding Ser., pp. 7–11, 2023, doi: 10.1145/3596947.3596954.

[17] J. Gallagher and W. Pugsley, “Development of a Dual-Input Neural Model for Detecting AI-Generated Imagery,” 2024, [Online]. Available: http://arxiv.org/abs/2406.13688.

[18] K. Simonyan and A. Zisserman, “Very deep convolutional networks for large-scale image recognition,” 3rd Int. Conf. Learn. Represent. ICLR 2015 - Conf. Track Proc., pp. 1–14, 2015.

[19] K. He, X. Zhang, S. Ren, and J. Sun, “Deep Residual Learning for Image Recognition Kaiming,” pp. 770–778, 2015, doi: 10.14456/mijet.2021.8.

[20] G. Huang, Z. Liu, and L. van der Maaten, “Densely Connected Convolutional Networks,” vol. 39, no. 9, pp. 1442–1446, 2016, doi: https://doi.org/10.48550/arXiv.1608.06993.

[21] S. Li, L. Wang, J. Li, and Y. Yao, “Image Classification Algorithm Based on Improved AlexNet,” J. Phys. Conf. Ser., vol. 1813, no. 1, 2021, doi: 10.1088/1742-6596/1813/1/012051.

[22] Y. Gal, “Dropout as a Bayesian Approximation : Insights and Applications,” 2015.

[23] A. Coates, H. Lee, and A. Y. Ng, “An analysis of single-layer networks in unsupervised feature learning,” J. Mach. Learn. Res., vol. 15, pp. 215–223, 2011.

[24] G. E. Hinton, N. Srivastava, A. Krizhevsky, I. Sutskever, and R. R. Salakhutdinov, “Improving neural networks by preventing co-adaptation of feature detectors,” pp. 1–18, 2012, [Online]. Available: http://arxiv.org/abs/1207.0580.

[25] S. Wager, S. Wang, and P. Liang, “Dropout training as adaptive regularization,” Adv. Neural Inf. Process. Syst., pp. 1–11, 2013.

[26] Ž. Vujović, “Classification Model Evaluation Metrics,” Int. J. Adv. Comput. Sci. Appl., vol. 12, no. 6, pp. 599–606, 2021, doi: 10.14569/IJACSA.2021.0120670.

[27] J. Erbani, P. É. Portier, E. Egyed-Zsigmond, and D. Nurbakova, “Confusion Matrices: A Unified Theory,” IEEE Access, vol. 12, no. October, 2024, doi: 10.1109/ACCESS.2024.3507199.