Protecting and Mitigating Against Cyber Threats

14Deepfake Face Detection Using Deep Convolutional Neural Networks: A Comparative Study

Krishna Prasanna Gottumukkala^*, Sirikonda Manasa, Komal Chakravarthy and Kolikipogu Ramakrishna

Department of Information Technology Chaitanya Bharathi Institute of Technology (A), Hyderabad, Telangana, India

Abstract

Concerns over the possible abuse of altered photos have grown significantly as a result of the deepfake technology’s quick development. In order to detect deepfake images, this research paper introduces a new and enhanced Deep Convolutional Neural Network (D-CNN) architecture which offers decent accuracy and great generalizability. Observing the issue’s seriousness and depth, the study is conducted on different models for deepfake image detection including D-CNN such as DenseNet, Customized CNN, and Mesonet of CNN architectures on a sizable dataset that included both real and deep fake images from seven different sources. This research paper provides a thorough study and examination of various existing CNN-based methods for detecting deepfake images, highlighting their advantages and potential limitations. The solid foundation for detecting deepfake image content is then outlined in the D-CNN-based architecture, which is being exposed. The D-CNN architecture highlights the remarkable accuracy of 92% in recognizing deepfake images, and its potential to mitigate the threat posed by deepfake technologies. While densenet yields an accuracy of 80%, and the mesonet yields 67%. This ...