Loading article...
Loading article...
Stream/Area:
-
Subject:
-
Paper ID:
IJSREM68155
Volume:
10 |
10 |
Oct 2026
Pages:
1-5
Publication Frequency:
Monthly (12 issues per year)
License:International Journal of Scientific Research in Engineering and Management follows the Creative Commons Attribution 4.0 (CC BY 4.0) License, allowing free use and distribution with proper attribution.
Computer science and engineering
Don Bosco Institute of Technology
Computer science and engineering
Don Bosco Institute of Technology
Computer science and engineering
Don Bosco Institute of Technology
Computer science and engineering
Don Bosco Institute of Technology
Computer science and engineering
Don Bosco Institute of Technology
The rapid improvement of generative models has made synthetic facial images increasingly difficult to distin-guish from authentic photographs. Existing deepfake de-tection research has explored different manipulation types, datasets, convolutional architectures, and learning strate-gies, with explainability becoming increasingly important for practical use [1]. This paper presents DeepEye, an image-based deepfake detection system that combines a pretrained DenseNet201 feature extractor with a Random Forest classi-fier and an explainable AI layer. The proposed pipeline ac-cepts a facial image resized to 224 × 224 pixels, extracts high-level visual representations using DenseNet201 pretrained on ImageNet, applies global average pooling to obtain 1920 features, standardizes the feature vectors using training-set statistics, and performs binary classification with Random Forest. DeepEye also uses Grad-CAM and SHAP to provide visual and feature-level explanations. The model was evalu-ated on the 140K Real and Fake Faces dataset using accuracy, precision, recall, F1-score, and ROC-AUC. The reported eval-uation produced 81.06% accuracy, 82.08% precision, 79.46% recall, 80.75% F1-score, and 89.59% ROC-AUC. The results indicate that the fixed feature extractor and learnable clas-sifier provide a practical baseline for image-level deepfake screening, while the explanation modules make the output easier to inspect.
KEYWORDS
This paper presented DeepEye, an explainable image-based deep-fake detection system that combines a fixed DenseNet201 feature extractor with a learnable Random Forest classifier. The pipeline converts 224 × 224 facial images into 1920-dimensional pooled representations, standardizes the features using training-set statis-tics, and performs binary real/fake classification. The reported results of 81.06% accuracy, 82.08% precision, 79.46% recall, 80.75% F1-score, and 89.59% ROC-AUC demonstrate that the approach can provide a useful baseline for detecting synthetic facial imagery in the selected dataset. The addition of Grad-CAM and SHAP gives the system an interpretability layer that is absent from a simple class-label pre- dictor. These explanations can help inspect whether the model relies on meaningful facial evidence and can support debugging when predictions appear uncertain. At the same time, the limita-tions of dataset dependence and changing generation techniques mean that the system should be treated as a research prototype rather than a universal authenticity verifier. Broader evaluation, fine-tuning, and cross-dataset testing are the main directions for strengthening DeepEye.
[1] M. Aleem, M. Umair, M. Zubair, R. Ibrahim, M. T. Naseem, M. M. Raza, M. N. Ali, and B.-S. Kim, “Seeing Through the Fake: Explainable AI With Multiple CNNs for Deepfake Detection,” IEEE Access, vol. 14, 2026, doi: 10.1109/AC-CESS.2025.3649128. [2] R. Tolosana, R. Vera-Rodriguez, J. Fierrez, A. Morales, and J. Ortega-Garcia, “DeepFakes and beyond: A survey of face manipulation and fake detection,” Information Fusion, vol. 64, pp. 131–148, 2020. [3] D. Afchar, V. Nozick, J. Yamagishi, and I. Echizen, “MesoNet: A compact facial video forgery detection network,” in Proc. IEEE WIFS, 2018. [4] A. Ro¨ ssler, D. Cozzolino, L. Verdoliva, C. Riess, J. Thies, and M. Nießner, “FaceForensics: A large-scale video dataset for forgery detection in human faces,” arXiv:1803.09179, 2018. [5] A. Ro¨ ssler, D. Cozzolino, L. Verdoliva, C. Riess, J. Thies, and M. Nießner, “FaceForensics++: Learning to detect manipulated facial images,” in Proc. IEEE/CVF ICCV, 2019, pp. 1–11. [6] B. Dolhansky, J. Bitton, B. Pflaum, J. Lu, R. Howes, M. Wang, and C. C. Ferrer, “The DeepFake Detection Challenge (DFDC) dataset,” arXiv:2006.07397, 2020. [7] Y. Li, X. Yang, P. Sun, H. Qi, and S. Lyu, “Celeb-DF: A new dataset for DeepFake forensics,” in Proc. IEEE/CVF CVPR, 2020. [8] Y. Li and S. Lyu, “Exposing DeepFake videos by detecting face warping artifacts,” in Proc. IEEE/CVF CVPR Workshops, 2019. [9] G. Huang, Z. Liu, L. Van Der Maaten, and K. Q. Weinberger, “Densely connected convolutional networks,” in Proc. IEEE CVPR, 2017, pp. 4700–4708. [10] J. Bayar and M. C. Stamm, “A deep learning approach to universal image manipulation detection using a new convolutional layer,” in Proc. ACM IH&MMSec, 2016.
Publication Date
10/21/2026
Peer Review Type
Double-Blind Peer Review
Review Rounds: 2
Documents
Article Views
1
Total views since publication
PDF Downloads
7
Total downloads since publication
Geographic Reach
5
Digital Preservation: This article is permanently archived on Zenodo and complies with all data preservation standards across major indexing platforms.
Plagiarism Screening
This manuscript has been checked for plagiarism using TurnitinPlagiarismChecker software.
Conflict of Interest
The authors declare no conflict of interest related to this research work.
Funding Information
No funding information available for this research.
Related research in the same subject area
Loading similar articles...