Investigation Of Comparison on Modified CNN Techniques to Classify Fake Face in Deepfake Videos

Deepfake is an advanced deep learning technology that will allow us to swap faces. With recent advancements in Neural network algorithms such as Generative Adversarial Neu- ral networks, hyper-realistic images and videos can be generated. It will be difficult to differentiate between fake and real images orvideos for naked eyes. These fake images and videos pose a real threat in various societal platforms. To reduce these issues various deep learning technologies can be used to classify deepfakes. Here we have compared three neural net models such as ResNext., Xception, and Ensemble of both. Our study has shown that the Ensemble method works better compared to others.