Matthias Kirchner

Can we trust digital image forensics

Compared to the prominent role digital images play in nowadays multimedia society, research in the field of image authenticity is still in its infancy. Only recently, research on digital image forensics has gained attention by addressing tamper detection and image source identification. However, most publications in this emerging field still lack rigorous discussions of robustness against strategic counterfeiters, who anticipate the existence of forensic techniques. As a result, the question of trustworthiness of digital image forensics arises. This work will take a closer look at two state-of-the-art forensic methods and proposes two counter-techniques; one to perform resampling operations undetectably and another one to forge traces of image origin. Implications for future image forensic systems will be discussed.

Hiding Traces of Resampling in Digital Images

Resampling detection has become a standard tool for forensic analyses of digital images. This paper presents new variants of image transformation operations which are undetectable by resampling detectors based on periodic variations in the residual signal of local linear predictors in the spatial domain. The effectiveness of the proposed method is supported with evidence from experiments on a large image database for various parameter settings. We benchmark detectability as well as the resulting image quality against conventional linear and bicubic interpolation and interpolation with a sinc kernel. These early findings on ldquocounter-forensicrdquo techniques put into question the reliability of known forensic tools against smart counterfeiters in general, and might serve as benchmarks and motivation for the development of much improved forensic techniques.

Fast and reliable resampling detection by spectral analysis of fixed linear predictor residue

This paper revisits the state-of-the-art resampling detector, which is based on periodic artifacts in the residue of a local linear predictor. Inspired by recent findings from the literature, we take a closer look at the complex detection procedure and model the detected artifacts in the spatial and frequency domain by means of the variance of the prediction residue. We give an exact formulation on how transformation parameters influence the appearance of periodic artifacts and analytically derive the expected position of characteristic resampling peaks. We present an equivalent accelerated and simplified detector, which is orders of magnitudes faster than the conventional scheme and experimentally shown to be comparably reliable.

Counter-Forensics: Attacking Image Forensics

This chapter discusses counter-forensics, the art and science of impeding or misleading forensic analyses of digital images. Research on counter-forensics is motivated by the need to assess and improve the reliability of forensic methods in situations where intelligent adversaries make efforts to induce a certain outcome of forensic analyses. Counter-forensics is first defined in a formal decision-theoretic framework. This framework is then interpreted and extended to encompass the requirements to forensic analyses in practice, including a discussion of the notion of authenticity in the presence of legitimate processing, and the role of image models with regard to the epistemic underpinning of the forensic decision problem. A terminology is developed that distinguishes security from robustness properties, integrated from post-processing attacks, and targeted from universal attacks. This terminology is directly applied in a self-contained technical survey of counter-forensics against image forensics, notably techniques that suppress traces of image processing and techniques that synthesize traces of authenticity, including examples and brief evaluations. A discussion of relations to other domains of multimedia security and an overview of open research questions concludes the chapter.

On resampling detection in re-compressed images

Resampling detection has become a standard tool in digital image forensics. This paper investigates the important case of resampling detection in re-compressed JPEG images. We show how blocking artifacts of the previous compression step can help to increase the otherwise drastically reduced detection performance in JPEG compressed images. We give a formulation on how affine transformations of JPEG compressed images affect state-of-the-art resampling detectors and derive a new efficient detection variant, which better suits this relevant detection scenario. The principal appropriateness of using JPEG pre-compression artifacts for the detection of resampling in re-compressed images is backed with experimental evidence on a large image set and for a variety of different JPEG qualities.

Tamper hiding: defeating image forensics

This paper introduces novel hiding techniques to counter the detection of image manipulations through forensic analyses. The presented techniques allow to resize and rotate (parts of) bitmap images without leaving a periodic pattern in the local linear predictor coefficients, which has been exploited by prior art to detect traces of manipulation. A quantitative evaluation on a batch of test images proves the proposed methods efficacy, while controlling for key parameters and for the retained image quality compared to conventional linear interpolation.

Efficient estimation of CFA pattern configuration in digital camera images

This paper proposes an efficient method to determine the concrete configuration of the color filter array (CFA) from demosaiced images. This is useful to decrease the degrees of freedom when checking for the existence or consistency of CFA artifacts in typical digital camera images. We see applications in a wide range of multimedia security scenarios whenever inter-pixel correlation plays an important role. Our method is based on a CFA synthesis procedure that finds the most likely raw sensor output for a given full-color image. We present approximate solutions that require only one linear filtering operation per image. The effectiveness of our method is demonstrated by experimental results from a large database of images.

Linear row and column predictors for the analysis of resized images

This paper adds a new perspective to the analysis and detection of periodic interpolation artifacts in resized digital images. Instead of relying on a single, global predictor, we discuss how the specific structure of resized images can be explicitly modeled by a series of linear predictors. Characteristic periodic correlations between neighboring pixels are then measured in the estimated predictor coefficients itself. Experimental results on a large database of images suggest a superior detection performance compared to state-of-the-art methods.

Multimedia Forensics Is Not Computer Forensics

The recent popularity of research on topics of multimedia forensics justifies reflections on the definition of the field. This paper devises an ontology that structures forensic disciplines by their primary domain of evidence. In this sense, both multimedia forensics and computer forensics belong to the class of digital forensics, but they differ notably in the underlying observer model that defines the forensic investigators view on (parts of) reality, which itself is not fully cognizable. Important consequences on the reliability of probative facts emerge with regard to available counter-forensic techniques: while perfect concealment of traces is possible for computer forensics, this level of certainty cannot be expected for manipulations of sensor data. We cite concrete examples and refer to established techniques to support our arguments.

Spectral methods to determine the exact scaling factor of resampled digital images

This paper combines analytical models of periodic interpolation artifacts with recent empirical findings on the spectral energy distribution of rescaled images to infer exact transformation parameters in a passive-blind forensic setting. We present a measure to solve a long-known ambiguity between upscaling and downscaling in the forensic analysis of resampled signals and thus substantially limit the range of candidate scaling factors. The effectiveness of our method is backed with empirical evidence on a large set of images and scaling factors.