Content uploaded by Siwei Lyu
Author content
All content in this area was uploaded by Siwei Lyu on Dec 04, 2015
Content may be subject to copyright.
A preview of the PDF is not available
Natural Image Statistics in Digital Image Forensics
Abstract and Figures
The fundamental problem in digital image forensics is to differentiate tampered images from those of untampered ones. A general solution framework can be obtained using the statistical properties of natural photographic images. In the recent years, applications of natural image statistics in digital image forensics have witnessed rapid developments and led to promising results. In this chapter, we provide an overview of recent developments of natural image statistics, and focus on three applications of natural image statistics in digital image forensics as (1) differentiating photographic images from computer-generated photorealistic images, (2) generic steganalysis; (3) rebroadcast image detection. © Springer Science+Business Media New York 2013. All rights are reserved.
Figures - uploaded by Siwei Lyu
Author content
All figure content in this area was uploaded by Siwei Lyu
Content may be subject to copyright.
... Image Forensics Traditional image forensics uses the natural statistics of images to detect tampered media (Fridrich, 2009;Lyu, 2013). A common approach is steganalysis (Lukáš et al., 2006;Fridrich, 2009;Bestagini et al., 2013), where high-frequency residuals are used to detect manipulations. ...
... While early GAN-generated images were easily distinguishable from real images, newer generations fool even human observers (Simonite, 2019). To facilitate the development of automated methods for recognizing fake images, we take inspiration from traditional image forensics (Lyu, 2013) and examine GAN-generated images in the frequency domain. ...
Deep neural networks can generate images that are astonishingly realistic, so much so that it is often hard for humans to distinguish them from actual photos. These achievements have been largely made possible by Generative Adversarial Networks (GANs). While these deep fake images have been thoroughly investigated in the image domain-a classical approach from the area of image forensics-an analysis in the frequency domain has been missing so far. In this paper, we address this shortcoming and our results reveal that in frequency space, GAN-generated images exhibit severe artifacts that can be easily identified. We perform a comprehensive analysis, showing that these artifacts are consistent across different neural network architectures, data sets, and resolutions. In a further investigation, we demonstrate that these artifacts are caused by upsampling operations found in all current GAN architectures, indicating a structural and fundamental problem in the way images are generated via GANs. Based on this analysis, we demonstrate how the frequency representation can be used to identify deep fake images in an automated way, surpassing state-of-the-art methods.
... To eliminate incorrectly recovered images, regularities in multi-scale representation of images are utilized. This is realized by computing first-and higher-order statistics from sub-band coefficients and using them as features to build a classification model [26]. ...
... To identify JPEG file fragments, we applied the ensemble classifier introduced in Sec.III-C to cand-jpg data segments which led to identification of 1.97 GiB of JPEG encoded data. Then each candidate JPEG encoded data bearing segment is carved, and the resulting partial image is validated to exhibit natural image statistics [26]. This process resulted with 53,387 partial images recovered from 1.71 GiB of data comprising orphaned JPEG file fragments. ...
Orphaned file fragment carving is concerned with recovering contents of encoded data in the absence of any coding metadata. Constructing an orphaned file carver requires addressing three challenges: a specialized decoder to interpret partial file data; the ability to discriminate a specific type of encoded data from all other types of data; and comprehensive prior knowledge on possible encoding settings. In this work, we build on the ability to render a partial image contained within a segment of JPEG coded data to introduce a new carving tool that addresses all these challenges. Towards this goal, we first propose a new method that discriminates JPEG file data from among 993 file data types with 97.7% accuracy. We also introduce a method for robustly delimiting entropy coded data segments of JPEG files. This in turn allows us to identify partial JPEG file headers with zero false rejection and 0.1% of false alarm rate. Secondly, we examine a very diverse image set comprising more than 7 million images. This ensures comprehensive coverage of coding parameters used by 3,269 camera models and a wide variety of image editing tools. Further, we assess the potential impact of the developed tool on practice in terms of the amount of new evidence that it can recover. Recovery results on a set of used SD cards purchased online show that our carver is able to recover 24% more image data as compared to existing file carving tools. Evaluations performed on a standard dataset also show that Jpg
Scraper
improves the state-of-the-art significantly in carving JPEG file data.
... We use three frequency transforms that have been applied successfully in both traditional image forensics (Lyu, 2008) and deepfake detection: discrete Fourier transform (DFT), discrete cosine transform (DCT), and the reduced spectrum (Durall et al., 2020;Dzanic et al., 2020;Schwarz et al., 2021), which is as a 1D representation of the DFT. While DFT and DCT visualize frequency artifacts, the reduced spectrum can be used to identify spectrum discrepancies. ...
Diffusion models (DMs) have recently emerged as a promising method in image synthesis. They have surpassed generative adversarial networks (GANs) in both diversity and quality, and have achieved impressive results in text-to-image and image-to-image modeling. However, to date, only little attention has been paid to the detection of DM-generated images, which is critical to prevent adverse impacts on our society. Although prior work has shown that GAN-generated images can be reliably detected using automated methods, it is unclear whether the same methods are effective against DMs. In this work, we address this challenge and take a first look at detecting DM-generated images. We approach the problem from two different angles: First, we evaluate the performance of state-of-the-art detectors on a variety of DMs. Second, we analyze DM-generated images in the frequency domain and study different factors that influence the spectral properties of these images. Most importantly, we demonstrate that GANs and DMs produce images with different characteristics, which requires adaptation of existing classifiers to ensure reliable detection. We believe this work provides the foundation and starting point for further research to detect DM deepfakes effectively.
... Deep fake detection: A comprehensive statistical studying of natural images shows that regularities always exist in natural images due to the strong correlations among pixels, see [29]. However, such regularity does not exist in synthesized images. ...
Generative Adversarial Networks (GANs) have paved the path towards entirely new media generation capabilities at the forefront of image, video, and audio synthesis. However, they can also be misused and abused to fabricate elaborate lies, capable of stirring up the public debate. The threat posed by GANs has sparked the need to discern between genuine content and fabricated one. Previous studies have tackled this task by using classical machine learning techniques, such as k-nearest neighbours and eigenfaces, which unfortunately did not prove very effective. Subsequent methods have focused on leveraging on frequency decompositions, i.e., discrete cosine transform, wavelets, and wavelet packets, to preprocess the input features for classifiers. However, existing approaches only rely on isotropic transformations. We argue that, since GANs primarily utilize isotropic convolutions to generate their output, they leave clear traces, their fingerprint, in the coefficient distribution on sub-bands extracted by anisotropic transformations. We employ the fully separable wavelet transform and multiwavelets to obtain the anisotropic features to feed to standard CNN classifiers. Lastly, we find the fully separable transform capable of improving the state-of-the-art.
The advent in graphic rendering software and technological progress in hardware can generate or modify photorealistic computer graphic (CG) images that are difficult to identify by human observers. Computer-generated images are used in magazines, film and advertisement industry, medical and insurance agencies, social media, and law agencies as an information carrier. The forged computer-generated image created by the malicious user may distort social stability and impacts on public opinion. Hence, the precise identification of computer graphic and photographic image (PG) is a significant and challenging task. In the last two decades, several researchers have proposed different algorithms with impressive accuracy rate, including a recent addition of deep learning methods. This comprehensive survey presents techniques dealing with CG and PG image classification using machine learning and deep learning. In the beginning, broad classification of all the methods in to five categories is discussed in addition to generalized framework of CG detection. Subsequently, all the significant works are surveyed and are grouped into five types: image statistics methods, acquisition device properties-based techniques, color, texture, and geometry-based methods, hybrid methods, and deep learning methods. The advantages and limitations of CG detection methods are also presented. Finally, major challenges and future trends in the CG and PG image identification field are discussed.
This paper proposes a CNN-based (Convolutional Neural Network based) network to detect altered face picture, which can cover the most common face swap methods. The network uses an autoencoder which is pre-trained on the original images to reconstruct the input images. The reconstructed one and the input image are then processed by the SRM filter which can extract the noise distribution of images. We then feed the minus result of two processed results into a CNN architecture to predict whether the input image is original or tampered. The model was trained and evaluated in FaceForensics dataset and state-of-art face swap method. Experimental results demonstrate the effectiveness of our network.
The statistics of photographic images, when represented using multiscale (wavelet) bases, exhibit two striking types of non-Gaussian behavior. First, the marginal densities of the coefficients have extended heavy tails. Second, the joint densities exhibit variance dependencies not captured by second-order models. We examine properties of the class of Gaussian scale mixtures, and show that these densities can accurately characterize both the marginal and joint distributions of natural image wavelet coefficients. This class of model suggests a Markov structure, in which wavelet coefficients are linked by hidden scaling variables corresponding to local image structure. We derive an estimator for these hidden variables, and show that a nonlinear "normalization" procedure can be used to Gaussianize the coefficients.
The tutorial starts with an overview of the concepts of VC dimension and structural risk minimization. We then describe linear Support Vector Machines (SVMs) for separable and non-separable data, working through a non-trivial example in detail. We describe a mechanical analogy, and discuss when SVM solutions are unique and when they are global. We describe how support vector training can be practically implemented, and discuss in detail the kernel mapping technique which is used to construct SVM solutions which are nonlinear in the data. We show how Support Vector machines can have very large (even infinite) VC dimension by computing the VC dimension for homogeneous polynomial and Gaussian radial basis function kernels. While very high VC dimension would normally bode ill for generalization performance, and while at present there exists no theory which shows that good generalization performance is guaranteed for SVMs, there are several arguments which support the observed high accuracy of SVMs, which we review. Results of some experiments which were inspired by these arguments are also presented. We give numerous examples and proofs of most of the key theorems. There is new material, and I hope that the reader will find that even old material is cast in a fresh light.
Techniques for information hiding have become increasingly more sophisticated and widespread. With high-resolution digital images as carriers, detecting hidden messages has become considerably more difficult. This paper describes an approach to detecting hidden messages in images that uses a wavelet-like decomposition to build higher-order statistical models of natural images. Support vector machines are then used to discriminate between untouched and adulterated images.
This paper describes pyramid solutions to graphics problems that have proven difficult in other image representations. The 'physics simulation' approach grows more out of the physics and mathematical modeling traditions. Greater realism can be achieved by using the physics simulation approach, but the complexity and computation time are vastly increased over the multiresolution pyramid approaches described here. Pyramid solutions to some graphics problems that have proven difficult in other image representations are described, viz. : (1) image analysis problems: (a) interpolation to fill missing pieces of an image, (b) smooth merging of several images to form mosaics; (2) creation of realistic looking images: (a) shadows and shading, (b) fast generation of natural looking textures and scenes using fractals
This paper is a revised version of an article by the same title and author which appeared in the April 1991 issue of Communications of the ACM. For the past few years, a joint ISO/CCITT committee known as JPEG (Joint Photographic Experts Group) has been working to establish the first international compression standard for continuous-tone still images, both grayscale and color. JPEG’s proposed standard aims to be generic, to support a wide variety of applications for continuous-tone images. To meet the differing needs of many applications, the JPEG standard includes two basic compression methods, each with various modes of operation. A DCT-based method is specified for “lossy’ ’ compression, and a predictive method for “lossless’ ’ compression. JPEG features a simple lossy technique known as the Baseline method, a subset of the other DCT-based modes of operation. The Baseline method has been by far the most widely implemented JPEG method to date, and is sufficient in its own right for a large number of applications. This article provides an overview of the JPEG standard, and focuses in detail on the Baseline method. 1