A Deep-Neural-Network-Based Approach To Detecting Forgery Images Generated From Various Generative Adversarial Networks

Abstract

In this paper, the deep learning-based method for forgery image detection is presented. First, we respectively do discrete Fourier transform for both real images and the forgery images generated from the generative adversarial networks. Then the obtained Fourier spectrums are fed to deep neural networks for model training. In order to enhance the detection capability of the model, we incorporate contrastive learning to make the model directly learns the difference between real and forgery images. Four kinds of generative adversarial networks (GANs), namely DCGAN, CycleGAN, AutoGAN, and Mixed GAN, are chosen to generate forgery images for testing our proposed method. The experimental results reveal that the average accuracy rate reaches 99.5% using our proposed method to detect the four kinds of GAN-generated images. Compared with the state-of-the-art forgery image detection method, our proposed method can more widely detect forgery images derived from different sources.