Year: 2026
Type: book-chapter
Abstract: Deepfake technology represents a growing threat within digital media, not only because of the risks of misinforming the public but also of privacy violations and security breaches.This research works on the binary classification of facial features to increase deepfake detection.Taking advantage of the inherent differences between real and fake faces, we presented an effective method using deep learning to identify minute changes in texture, lighting, and facial movements.Our technique is centered around the extraction of facial regions most likely to be altered (eyes, mouth, skin, etc.) through automation of image segmentation.The next step is to use multiple CNNs and dense layers to obtain a high-level feature representation and to categorize it into real or fake.Besides, our method utilizes data augmentation along with ensemble techniques to make it more resistant to changes in quality and other environmental factors.The accomplishment of the advanced performance model with minimum computing power use has proven the effectiveness of the proposed framework on the Kaggle dataset.In conclusion, our study points out binary classification of facial features could be a key method for identifying deepfakes in a manner that is accurate, reliable, and scalable.Thus, we are hopeful that it will lead to the practical application of media forensics and digital security.
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7220846015