Identification of deep network generated images using disparities in color components

Abstract

With the powerful deep network architectures, such as generative adversarial networks and variational autoencoders, large amounts of photorealistic images can be generated. The generated images, already fooling human eyes successfully, are not initially targeted for deceiving image authentication systems. However, research communities as well as public media show great concerns on whether these images would lead to serious security issues. In this paper, we address the problem of detecting deep network generated (DNG) images by analyzing the disparities in color components between real scene images and DNG images. Existing deep networks generate images in RGB color space and have no explicit constrains on color correlations; therefore, DNG images have more obvious differences from real images in other color spaces, such as HSV and YCbCr, especially in the chrominance components. Besides, the DNG images are different from the real ones when considering red, green, and blue components together. Based on these observations, we propose a feature set to capture color image statistics for detecting the DNG images. Moreover, three different detection scenarios in practice are considered and the corresponding detection strategies are designed. Extensive experiments have been conducted on face image datasets to evaluate the effectiveness of the proposed method. The experimental results show that the proposed method is able to distinguish the DNG images from real ones with high accuracies. Index TermsImage generative model, generative adversarial networks, fake image identification, image statistics.

Exposing GAN-synthesized Faces Using Landmark Locations

“…Compared to popularity of exploring strategies for synthesizing face images with GANs, methodologies to differentiate the real and synthesized images are far from satisfactory. Li et al [10] observed the color mismatch in H, S, V and Cb, Cr, Y channels between real and GAN-generated images. Similarly, McCloskey and Albright identified the frequency of saturated pixels and color image statistics of the GAN-generated images are different from the ones captured by cameras [13].…”

“…Unlike previous image/video manipulation methods, realistic images are generated completely from random noise through a deep neural network. Current detection methods are based on low level features such as color disparities [10,13], or using the whole image as input to a neural network to extract holistic features [19].…”

Detecting GAN generated Fake Images using Co-occurrence Matrices

“…Several methods have been proposed in the area of image forensics over the past years [25,26,27,28,29]. Recent approaches have focused on applying deep learning based methods to detect tampered images [30,31,32,33,34,9,35] The detection of GAN images is a new area in image forensics and there are very few papers in this area [36,37,38,39,40,41,42,43,44,45]. Related fields also include detection of computer generated (CG) images [46,47,48,49].…”

“…The top results they obtained using a combination of residual features [50,9] and deep learning [51]. Similar to [36], the authors in [38] compute the residuals of high pass filtered images and then extract co-occurrence matrices on these residuals, which are then concatenated to form a feature vector that can distinguish real from fake GAN images. In contrast to these approaches, our approach does not need any image resid-uals to be computed.…”

Incremental learning for the detection and classification of GAN-generated images

“…Some of them exploit specific facial artifacts, like asymmetries in the colour of the eyes, or artifacts arising from an imprecise estimation of the underlying geometry, especially on areas around the nose, the border of the face, and the eyebrows [7]. Color information is instead exploited in [8], [9]. In particular, [8] proposes to use features shared by different GAN architectures, based on the way they transform a multichannel feature map into a 3-channel color image.…”