Improving DeepFake Video Detection Performance with a Noval Deep Learning Approach

With the rise in both the quantity and sophistication of deepfake videos, the need for robust detection systems to identify potentially misleading content on social media and the internet has become paramount. However, current automated face forgery detection systems still face limitations, often demonstrating bias towards the training dataset. This research paper addresses this issue by proposing a novel approach for detecting deepfake media. We introduce a custom Visual Geometry Group (VGG16) deepfake detection method that leverages convolutional neural network architectures. To evaluate the effectiveness of our approach, we utilize the deepfake detection challenge (DFDC) dataset on Kaggle to build network models and compare the performance of our custom VGG16 method against the standard VGG16. Additionally, we investigate the impact of data augmentation techniques on the performance of Convolutional Neural Network (CNN)-based deepfake detectors, examining their effect on both VGG16 and our custom VGG16 approach using the DFDC dataset. Our results demonstrate a high level of accuracy, with precision, recall, and f1-score values of 0.983, 0.975, and 0.979, respectively, and an overall accuracy of 0.986 for deepfake detection. This study presents a promising approach to enhance the accuracy of deepfake video detection, representing a crucial step towards mitigating the potential negative impacts of deepfake technology.