,
Research Scholar, Department of Computer Science, Karpagam Academy of Higher Education , Coimbatore, Tamil Nadu , India
Professor and Head, Department of Computer Science, Karpagam Academy of Higher Education , Coimbatore, Tamil Nadu , India
The technologies of deepfakes are actively used on social media, creating fake video and audio by manipulating existing media content. The face-swapping technologies that are applied in the creation of deepfakes cause severe problems in society, such as identity theft and the spread of unsuitable content. Although various machine learning and deep learning techniques have been used to detect deepfakes, the generalization abilities and linguistic properties of deepfakes remain problematic. This paper presents the Deep Inception V7 Convolution Neural Network (CNN) model to identify deepfake content in different types of image synthesis in videos. The proposed model works on the principles of spatiotemporal specifics of video frames, detecting faces and body parts. The first step is converting videos into frames, and the second step is to segment them and isolate face and body features. These divided parts of the face are then fed into the deep inception network that records the spatiotemporal variation in convolution and max-pooling layers to produce maps of feature richness with contextual information. These feature maps are sent to a fully connected layer to differentiate between real and fake videos. The Adam optimizer is used to optimize the model, making it more robust and accurate. The DFDC benchmark dataset is used to perform training and validation. Through performance analysis, it has been identified that the proposed model has a training accuracy of 96.87% and a validation accuracy of 94.74%, which is better than the current methods of deepfake detection. The findings illustrate that the Deep Inception V7 CNN is effective in detecting deepfakes in different settings. These results have significant implications for enhancing information transparency in online content through the use of advanced technology.
This is an open access article distributed under the Creative Commons Attribution Non-Commercial License (CC BY-NC) License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
The statements, opinions and data contained in the journal are solely those of the individual authors and contributors and not of the publisher and the editor(s). We stay neutral with regard to jurisdictional claims in published maps and institutional affiliations.
