Work

Optimization Accuracy of Deepfake Face Detection using DenseNet121 Technique

Year: 2026
Type: article
Abstract: The rapid advancement of generative artificial intelligence has enabled the creation of highly realistic deepfake facial images and videos, creating significant challenges for digital media authentication, privacy protection, and cybersecurity. Accurate and robust deepfake detection is therefore essential for identifying manipulated facial content. This study proposes an optimization-based deepfake face detection framework using the DenseNet121 deep learning architecture. The proposed approach utilizes DenseNet121 to automatically extract discriminative hierarchical features from facial images and classify them into authentic and deepfake categories. The methodology includes dataset collection, image preprocessing, face-region extraction, data augmentation, feature learning, model training, and performance evaluation. DenseNet121 is selected because its dense connectivity facilitates effective feature reuse and enables the propagation of information and gradients across network layers, which can improve feature representation while reducing redundant parameters. To optimize detection performance, appropriate training parameters and classification settings are systematically adjusted, and the resulting model is evaluated using accuracy, precision, recall, F1-score, and confusion matrix analysis. The proposed framework is intended to improve the reliable identification of subtle facial manipulation artifacts while maintaining efficient deep feature extraction. The study also considers practical challenges such as image quality variation, compression, illumination changes, and differences between manipulation techniques. The expected outcome is a robust DenseNet121-based detection model capable of achieving improved classification accuracy and better generalization for deepfake face detection. The proposed approach can support applications in digital forensics, social-media security, identity protection, and trustworthy multimedia authentication.
Language: English
Open Access status: hybrid
OpenAlex ID: W7217216638
FWCI:
Cites: 14
Cited by:
Related to: 10