Deep Fake Detection Using CNN

Sudip Ghosh; Saikat Dey

Deep Fake Detection Using CNN

Author : Sudip Ghosh, Saikat Dey Journa Name: International Journal of Science, Engineering and Technology Country : India Volume: 12 issue: 3 Year: 2024 Views : 450

Abstract:

This study introduces a deep learning approach for predicting Deep fakes using Convolutional Neural Networks (CNNs). The methodology entails training a CNN model on a dataset comprising both authentic and manipulated images sourced from Kaggle. Subsequently, transfer learning is applied by leveraging the pre-trained Xception model, which has been trained on the extensive Image Net dataset. Through this process, the model learns to differentiate between real and fake images by discerning unique patterns and features inherent to each category. Preliminary results indicate that the proposed CNN-based approach demonstrates satisfactory performance in identifying fake images. Efforts are ongoing to further enhance the accuracy of the model with the aim of achieving even better results.