A New Approach to in Ensemble Method for Deepfake Detection
Serhat Ataş, Mehmet Karaköse · 2023
With the great development of technology and deep learning gaining competence in many areas, recently forgery of images has started to pose great threats and become the main topic of most technology companies. But here, with the ongoing race between good and evil, new approaches have emerged in fraud detection. Since this technological development is bilateral, fraud detection is constantly trying to be developed and trying to catch fraud. In this regard, various approaches have emerged by many different companies and people to contribute to society. In the method we have proposed to contribute to these detection processes and to detect forgery, feature extraction is provided on images using the D-CNN model, and with these features, SVM, Estimation is done on features with Random Forest and Logistic Regression. Finally, using the Ensemble method, an estimation process is performed by taking all the estimates together with the sampling on the images. Thanks to this proposed method, the determinations made are supported in a layered way and precision is ensured at the accuracy rate.