ItemArticle

Hybrid CMNV2: DeepFake faces classification and recognition using deep learning methods

Title
Hybrid CMNV2: DeepFake faces classification and recognition using deep learning methods
Creator
Kumar, B. Anil; Misra, Neeraj Kumar; Pathak, Nirupma; Ahmadpour, Seyed-Sajad; Krishnamoorthy, Murugaperumal; Shukla, Dhirendra Kumar; Patidar, Mukesh; Hakimi, Musawer
Description
Deepfake detection has become a critical component of digital forensics and security, as manipulated images and videos increasingly threaten trust in visual media. However, existing methods often struggle with robustness under post-processing operations such as JPEG compression, Gaussian blur, scaling, and filtering, and with the growing diversity and realism of face image modification (FIM) forgeries. This work proposes CMNV2, a hybrid architecture that integrates MobileNetV2 with a custom CAFFE block to enhance feature extraction and classification accuracy. By adding five additional layers to a pre-trained structure, the model demonstrates superior resilience against complex real-world conditions and achieves 99.10% accuracy across multiple datasets, outperforming 13 baseline CNN models. The study trained and tested CMNV2 on 5,000 images (real and deepfake faces), using a combination of deep neural networks (DNNs), transfer learning (TL), and deep learning (DL) techniques. Compared to 13 CNN-based architectures, the proposed model achieved superior performance across some important evaluation metrics, including accuracy, precision, recall, F1-score, error rate, and computational efficiency. These results highlight hybrid CMNV2 as a robust and efficient solution for deepfake face detection and classification, with potential applications in security, healthcare, and education. 2025
Source
Results in Engineering;Volume;28;Issue;;Article No.;107513;
Date
01-01-2025
Publisher
Elsevier B.V.
Subject
Classification; Deep learning; Deep neural network; Detection; Hybrid CMNV2 model; Transfer learning
Coverage
Kumar B.A., Department of Computer Science and Engineering, Vignan's Foundation for Science, Technology and Research (VFSTR), (Deemed to be University), Off-campus Hyderabad, Telangana, India; Misra N.K., School of Electronics Engineering, VIT-AP University, Andhra Pradesh, Amaravathi, 522237, India; Pathak N., Department of Computer Science and Engineering, Koneru Lakshmaiah Education Foundation, Green Fields, Guntur District, Andhra Pradesh, Vaddeswaram, 522502, India; Ahmadpour S.-S., Department of Computer Engineering, Faculty of Engineering and Natural Science, Istanbul Atlas University, Istanbul, Turkey; Krishnamoorthy M., Department of Electrical and Electronics Engineering, Vardhaman College of Engineering, Hyderabad, 501218, India; Shukla D.K., School of Computer Science, Galgotias University, Greater Noida, India; Patidar M., Department of Computer Science & Engineering, Parul Institute of Engineering & Technology (Parul University) Vadodara, India; Hakimi M., Department of Computer Science, Samangan University, Samangan Province, Northeast Aybak, Afghanistan
Rights
All Open Access; Gold Open Access; Green Open Access
Relation
ISSN: 25901230;
Format
online
Language
English
Type
Article

Citation

Kumar, B. Anil; Misra, Neeraj Kumar; Pathak, Nirupma; Ahmadpour, Seyed-Sajad; Krishnamoorthy, Murugaperumal; Shukla, Dhirendra Kumar; Patidar, Mukesh; Hakimi, Musawer, “Hybrid CMNV2: DeepFake faces classification and recognition using deep learning methods,” CHRIST (Deemed To Be University) Institutional Repository, accessed June 18, 2026, https://archives.christuniversity.in/items/show/22443.

Output Formats