Paper Contents
Abstract
This paper presents a deepfake detection system that effectively addresses the challenge of identifying video forgery by utilizing advanced deep learning techniques. The system leverages a convolutional neural network (CNN) built on a ResNet framework to achieve robust feature extraction, while maintaining spatial and temporal information from video frames. A sequence descriptor is utilized to generate strong representations of these extracted features, which are subsequently classified using a detection network comprised of fully connected layers. This classification method differentiates between authentic and altered videos with a high degree of precision. The proposed framework not only achieves accurate detection but is also adaptable to new deepfake techniques, ensuring long-term effectiveness. By integrating innovative architectures and temporal analysis, the system offers a dependable and efficient solution to uphold the authenticity of digital media in an era characterized by increasing deepfake manipulation. Experimental findings indicate that our approach significantly enhances detection performance, achieving an optimal balance between accuracy and computational efficiency across various contexts.Index Terms Deepfake Detection, Video Forgeries, Convolutional Neural Network (CNN), ResNet Feature Extraction, Fully Connected Layers, Digital Media Authenticity, Manipulated Videos, Emerging Deepfake Techniques, Facial Feature Representation, Detection Accuracy, Adaptive Model, Forgery Identification.
Copyright
Copyright © 2025 SHANMUGAPRIYA R. This is an open access article distributed under the Creative Commons Attribution License.
