A New Deep Learning-Based Methodology for Video Deepfake Detection Using XGBoost.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Cybersecurity

Background:

• The proliferation of realistic face-swapping deepfake videos poses significant threats to privacy and national security.
• Distinguishing between authentic and manipulated video content is a critical challenge in the digital age.

Purpose of the Study:

• To develop and evaluate a novel deepfake detection method combining You Only Look Once (YOLO), Convolutional Neural Network (CNN), and Extreme Gradient Boosting (XGBoost).
• To enhance the accuracy and reliability of deepfake detection systems.

Main Methods:

• Utilizing the YOLO face detector to isolate facial regions within video frames.
• Employing the InceptionResNetV2 CNN architecture for feature extraction from detected faces.
• Integrating XGBoost as a classifier on top of the CNN features for final deepfake recognition.

Main Results:

• The YOLO-CNN-XGBoost method achieved an Area Under the Receiver Operating Characteristic Curve (AUC) of 90.62%.
• Performance metrics include 90.73% accuracy, 93.53% specificity, 85.39% sensitivity, 85.39% recall, 87.36% precision, and 86.36% F1-measure on the CelebDF-FaceForencics++ (c23) dataset.
• The proposed method demonstrated superior performance compared to existing state-of-the-art deepfake detection techniques.

Conclusions:

• The YOLO-CNN-XGBoost method presents a robust and effective solution for deepfake video detection.
• The hybrid approach significantly improves the ability to identify manipulated media, addressing critical privacy and security concerns.