Abstract

Images play an important role in defining human perception, and the power to manipulate such images gives immense power to malicious users. The new advancement in Artificial Intelligence, has altogether worked on the quality and productivity in creating counterfeit face pictures; for instance, the face manipulated by GANs is sensible to such an extent that it is hard to recognize the validness, either by the computer or by people. To improve the accuracy of recognizing facial pictures created by AI from genuine facial ones, an enhanced model has been proposed in this paper which is dependent on profound learnings like Deep Learning, Convolutional Neural Network (CNN), and Error Level Analysis (ELA). Our findings push the boundaries of understanding DeepFake detection and our solution to detect these images is based on the concepts of image error level and Deep learning. Our model uses the Convolutional Neural Network (CNN) architecture that utilizes error level analysis (ELA) to pre-process the images. We have utilized a dataset comprising on 24,000 images with equal split of real and deepfake images to traing and test our model. We were able to achieve an accuracy of 99%. The proposed model has a shorter training time and higher efficiency than most other methods for DeepFake detection.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call