FusionNet: A Comprehensive Framework for Automated Deepfake Detection
DOI:
https://doi.org/10.64296/vijir.v1i2.01Keywords:
Deepfake detection, Image classification, Deep Learning, Hybrid feature extraction, Convolutional Neural Network (CNN)Abstract
Advancement in deep generative models have led to the spread of deepfake media, which poses a serious danger to information authenticity, privacy, and trust in digital media forensics. In order to meet the increasing demand for reliable deepfake picture identification, this study uses the publicly accessible Kaggle Deepfake Dataset to investigate hybrid deep learning approaches. Initial experiments with standalone Keras models, including DenseNet121, RestNet50, Xception, NASNet-Mobile, VGG16, and InceptionV3, achieved accuracies ranging from 89% to 90% over 20 epochs. To push the boundaries of performance, we developed a novel hybrid model that concatenates feature maps extracted from DenseNet121 and DenseNet169. This method greatly improved classification performance, obtaining a 91.22% accuracy rate with high precision 90.82%, 94.37% recall, and 92.56% F1-score with error rate 8.78%. Confusion matrices, classification reports, and ROC curves were used to thoroughly assess the model, which showed that it was effective at differentiating real photos from ones that had been altered, with an AUC score of 0.98. Our findings underscore the effectiveness of combining pre-trained CNN architectures for deepfake detection and contribute to the advancement of scalable and reliable solutions for safeguarding digital media integrity.
References
[1] A. Kaur, A. N. Hoshyar, V. Saikrishna, S. Firmin, and F. Xia, “Deepfake video detection: challenges and opportunities,” Artificial Intelligence Review, vol. 57, no. 6, pp. 1–47, 2024.
[2] C. Barnes and T. Barraclough, “Deepfakes and synthetic media,” in Emerging technologies and international security, pp. 206–220, Routledge, 2020.
[3] Q. Zhao, X. Wang, W. Jiang, and Y. Xiong, “Deep learning for deepfake detection: A comprehensive review,” ACMTOMM, vol. 17, no. 2, pp. 1–24, 2021.
[4] D. Afchar, V. Nozick, J. Yamagishi, and I. Echizen, “Mesonet: a compact facial video forgery detection network,” in 2018 IEEE WIFS, pp. 1–7.
[5] S. R. A. Ahmed and E. Sonuç, “Retracted article: Deepfake detection using rationale-augmented convolutional neural network,” Applied Nanoscience, vol. 13, no. 2, pp. 1485–1493, 2023.
[6] A. Raza, K. Munir, and M. Almutairi, “A novel deep learning approach for deepfake image detection,” Applied Sciences, vol. 12, no. 19, p. 9820, 2022.
[7] A. Rössler, D. Cozzolino, L. Verdoliva, C. Riess, J. Thies, and M. Nießner, “Faceforensics++: Learning to detect manipulated facial images,” in ICCV, pp. 1–11, 2019.
[8] H. Zhang, J. Wu, S. Liu, and S. Han, “A pre-trained multi-representation fusion network for molecular property prediction,” Information Fusion, vol. 103, p. 102092, 2024.
[9] A. Basaran, “Kaggle dataset.” [https://www.kaggle.com/datasets/abdurahimbaaran/deepfake](https://www.kaggle.com/datasets/abdurahimbaaran/deepfake) dataset. Accessed: Dec. 31, 2024.
[10] R. Dwivedi, D. Dave, H. Naik, et al., “Explainable AI (XAI): Core ideas, techniques, and solutions,” ACM Computing Surveys, vol. 55, no. 9, pp. 1–33, 2023.
