Abstract:Recent advances in Generative Adversarial Networks (GANs) have led to the creation of realistic-looking digital images that pose a major challenge to their detection by humans or computers. GANs are used in a wide range of tasks, from modifying small attributes of an image (StarGAN [14]), transferring attributes between image pairs (CycleGAN [91]), as well as generating entirely new images (ProGAN [36], StyleGAN [37], SPADE/GauGAN [64]). In this paper, we propose a novel approach to detect, attribute and localize GAN generated images that combines image features with deep learning methods. For every image, co-occurrence matrices are computed on neighborhood pixels of RGB channels in different directions (horizontal, vertical and diagonal). A deep learning network is then trained on these features to detect, attribute and localize these GAN generated/manipulated images. A large scale evaluation of our approach on 5 GAN datasets comprising over 2.76 million images (ProGAN, StarGAN, CycleGAN, StyleGAN and SPADE/GauGAN) shows promising results in detecting GAN generated images.

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is to detect, attribute, and localize images generated by generative adversarial networks (GANs). Specifically, with the development of GAN technology, the generated digital images are becoming more and more realistic, which poses a significant challenge to human or computer detection. These images are applied in many tasks, from modifying small attributes of images to generating completely new images. However, this technology also brings the problem of image forgery, making it difficult to distinguish between real and fake images. To solve this problem, the paper proposes a new method that combines image features and deep - learning methods. For each image, the co - occurrence matrices of RGB - channel neighboring pixels in different directions (horizontal, vertical, and diagonal) are calculated. Then, a deep - learning network is trained to detect, attribute, and localize these GAN - generated / tampered images. Through large - scale evaluations on five GAN data sets containing more than 2.76 million images, the results show that this method has a good effect in detecting GAN - generated images. The main contributions of the paper include: - Proposing a new method for detecting, attributing, and localizing GAN - generated images, which combines deep - learning and co - occurrence matrices. - Calculating the co - occurrence matrices of images in different directions and using deep - learning for training. For detection and attribution, the matrices are calculated on the entire image; for localization, the co - occurrence matrices are calculated on image patches to generate heat maps. - Conducting tests on more than 2.7 million images. To the best of the authors' knowledge, this is the largest - scale evaluation of GAN - image detection. - Using t - SNE visualization to explain the effectiveness of the method. - Proving that this method can maintain performance under different JPEG compression factors and image - patch sizes, and is suitable for a variety of practical application scenarios.

Detection, Attribution and Localization of GAN Generated Images

No One Can Escape: A General Approach to Detect Tampered and Generated Image

Towards Discovery and Attribution of Open-world GAN Generated Images

Detection of GAN Generated Image Using Color Gradient Representation

Detecting GAN-generated Face Images Via Hybrid Texture and Sensor Noise Based Features

AI vs. AI: Can AI Detect AI-Generated Images?

Detecting GAN-generated Imagery using Color Cues

Generative adversarial networks (GANs): Introduction, Taxonomy, Variants, Limitations, and Applications

Detecting GAN-generated Images by Orthogonal Training of Multiple CNNs

Towards Robust GAN-generated Image Detection: a Multi-view Completion Representation

Detecting Generated Images by Real Images Only

Generative adversarial network: An overview of theory and applications

Image plagiarism detection using GAN - (Generative Adversarial Network)

Detecting GAN generated errors

Learning on Gradients: Generalized Artifacts Representation for GAN-Generated Images Detection

Generic image application using GANs (Generative Adversarial Networks): A Review

GAN in digital image processing

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

T-GD: Transferable GAN-generated Images Detection Framework

Research on Image Tampering Detection and Localization Based on Iterative GAN

Detection of GAN-Generated Images by Estimating Artifact Similarity