Deepfake Detection Using SVM
Authors:
- 4 August 2021
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1245-1249
Abstract
In recent years the Deep generative networks have made it easy to create real face swaps in images and videos with less traces of manipulation, significantly improving the quality of these deepfakes. This improvement in fake media have gained more concern as for their use in fake terrorism, blackmail etc. This has drawn attention of both industry and government to distinguish and restrict their utilization. In spite of the way that these AI generated fake media produce realistic images upon detailed assessment, the proposed method can find some features that are unnatural which are not visible to the naked eyes. This paper uses a method known as frequency domain analysis after which a classifier will be used to differentiate the real and fake image. This paper evaluates our method on dataset of deepfake images collected from different website. Our work can show promising performance for detecting these deepfake images.Keywords
This publication has 5 references indexed in Scilit:
- Smart Farming Techniques for New Farmers Using Machine Learning2021
- Malicious Application Detection in Android using Machine Learning2020
- Deepfake Video Detection Using Recurrent Neural Networks2018
- Exposing Digital Image Forgeries by Illumination Color ClassificationIEEE Transactions on Information Forensics and Security, 2013
- Image Forgery Localization via Fine-Grained Analysis of CFA ArtifactsIEEE Transactions on Information Forensics and Security, 2012