Search

Camcorder piracy refers to the process of using a camcorder to record a screen that displays copyrighted content. In contrast to the previous works that aimed at detecting the occurrence of camcorder piracy, this paper conducts an in-depth study of the luminance flicker that is naturally present in camcorded videos due to the interplay between a liquid-crystal-display (LCD) screen and a camcorder. We first model the flicker signal and show that its parameters are tied to such internal characteristics of the pirate devices as the back-light frequency of the LCD screen and the read-out time of the camcorder. We then present new estimation techniques to recover these hidden parameters directly from camcorded videos and demonstrate that such forensic cues could provide intelligence on the pirate devices. We also discuss how to recover the shape of the low-power flicker signal itself and show that it could be used to infer which back-light technology employed in the pirate LCD screen.

Spatio-temporal desynchronization remains a major challenge for watermarking system as it could impair the detection of the hidden payload. Over the years, several (non-blind) registration techniques have been proposed to realign the analyzed content prior to watermark detection and thereby achieve robustness against severe attacks such as display-and-camcord. Such techniques rely on assumptions that may be invalidated by new attacks that had not been accounted for. In the case of video, for instance, best-in-class resynchronization techniques may have difficulties to cope with screencasting. In this paper, we report findings after an experimental study that involved five popular screencasting tools. The impact of screencasting can indeed be characterized in terms of frame drops and repeats. Moreover, the on-the-fly encoding process occurring during screen recapture inherently introduces dependencies between the likelihood of drop/repeat and the frame encoding type.

In order to understand and analyse the main components of watermarking security presented in the next chapters, we introduce in this chapter the different elements needed to embed a watermark or a message inside a host content. We first present a functional view of a watermarking scheme (the embedding function, the decoding/detection function) and then its geometrical interpretation. Then we present the most popular class of watermarking schemes: spread-spectrum watermarking and watermarking techniques based on the idea of dirty paper codes.

This chapter concludes this book on Watermarking Security and I hope that the reader will have a better view of the ins and outs of this domain. If watermarking security may look like a cat and mice game that is never ending, we can however state several important conclusions derived for the large variety of researches that have been conducted but there are also fascinating related problems that still need to be solved. The goal of this last chapter is to list the different outputs and open questions related to watermarking security but also general security of information forensics.

We propose a new method to detect resampled imagery. The method is based on examining the normalized energy density present within windows of varying size in the second derivative of the image in the frequency domain, and exploiting this characteristic to derive a 19-D feature vector that is used to train a SVM classifier. Experimental results are reported on 7500 raw images from the BOSS database. Comparison with prior work reveals that the proposed algorithm performs similarly for resampling rates greater than 1, and is superior to prior work for resampling rates less than 1. Experiments are performed for both bilinear and bicubic interpolations, and qualitatively similar results are observed for each. Results are also provided for the detection of resampled imagery with noise corruption and JPEG compression. As expected, some degradation in performance is observed as the noise increases or the JPEG quality factor declines.

Motion coherency has recently been identified as a desirable property for watermarks embedded within video streams in order to withstand temporal frame averaging along the motion axis. Nevertheless, no tool has been proposed to easily evaluate the motion coherency of a given watermarking system. Today, this assessment relies on a computationally expensive procedure, namely, (1) embed a watermark, (2) perform temporal frame averaging, and (3) check for the presence of the watermark. In this article, a novel oracle is designed to detect whether a video stream contains any motion-incoherent component or not. Since such incoherence can be introduced by nonmotion-coherent watermarking algorithms, this tool has proven to be most valuable to distinguish watermarked from nonwatermarked content. The oracle relies on some features extracted from error frames after motion compensation. Experimental results demonstrate the efficiency of the proposed method with uncompressed and compressed video streams.

Disparity coherent watermarking has been introduced as a means to address the unique characteristics of stereoscopic visual content. This strategy has been reported to notably improve robustness performances in prior works, especially with respect to virtual view synthesis. While disparity coherence has always been conjectured to also provide better improved fidelity, it had never been properly evaluated. This perceptual study investigates the perception of three alternate watermarking strategies for stereo visual content on a population of thirty three observers. The analysis of the results reveals that there are three categories of observers and one of these categories, which amounts to nearly one third of our observers, is extremely sensitive to watermark coherence even for low-power watermarks. Moreover, the sensitivity to watermark coherence is found to be content dependent.

Correlation-based watermark detection inherently assumes that a pseudo-random watermark pattern is present in the cover signal at a given location. However, disparity-coherent watermarks for stereo video content are broken into pieces which are offset one with respect to the others. Acoustic transmission yields echoes due to multi-path propagation and may thus introduce watermark replicates in the recorded audio track. In such cases, the watermark energy is not concentrated in a single element of the cross-correlation array but rather distributed over a number of them. As a result, it is necessary to devise aggregation strategies to recover these scattered watermarks. In this paper, we present a method that relies on outlier detection to identify which elements of the correlation array are associated to the embedded watermark and should be aggregated. Experimental results clearly demonstrate the ability of this technique to recover the watermark without introducing undesired false positives, even in noisy conditions.

Cryptography-based content protection is an efficient means to protect multimedia content during transport. Nevertheless, content is eventually decrypted at rendering time, leaving it vulnerable to piracy e.g. using a camcorder to record movies displayed on an LCD screen. Such type of piracy naturally imprints a visible flicker signal in the pirate video due to the interplay between the rendering and acquisition devices. The parameters of such flicker are inherently tied to the characteristics of the pirate devices such as the back-light of the LCD screen and the read-out time of the camcorder. In this article, we introduce a forensic methodology to estimate such parameters by analyzing the flicker signal present in pirate recordings. Experimental results clearly showcase that the accuracy of these estimation techniques offers efficient means to tell-tale which devices have been used for piracy thanks to the variety of factory settings used by consumer electronics manufacturers.

International audience; Radial-based 3D watermarking alters the distances between the center of mass of the 3D mesh and its vertices. These watermarking systems are inherently sensitive to cropping. To address this limitation, this paper introduces a complementary blind resynchronization module to transmit critical synchronization information to the watermark decoder. Spherical patterns formed by several secret landmark vertices are embedded alongside the payload and blindly retrieved by the decoder, thereby conveying the synchronization information needed. Experimental results showcase significant improvement against cropping, while preserving performances against valumetric attacks thanks to a control parameter that automatically switches between alternate resynchronization modes.

Digital watermarking has been introduced in the 1990s as a complementary technology for copyright protection. In an effort to anticipate hostile behavior of adversaries, the research community is constantly introducing new attacks to benchmark watermarking systems. In this paper, we present a generic attack strategy based on block replacement. As multimedia content is often highly repetitive, the attack exploits signal's self-similarities to replace each signal block with another, perceptually similar one. Guided by the principles of the proposed attack framework, we implemented three attack algorithms for different types of multimedia content: video shots, audio tracks and still images. Finally, considering the effectiveness of the proposed algorithms, we identify the properties that a watermark should have to counter this attacking strategy

Geometrical distortions are the Achilles heel for many watermarking schemes. Most countermeasures proposed in the literature only address the problem of global affine transforms (e.g., rotation, scaling, and translation). In this paper, we propose an original blind watermarking algorithm robust to local geometrical distortions such as the deformations induced by Stirmark. Our method consists in adding a predefined additional information to the useful message bits at the insertion step. These additional bits are labeled as resynchronization bits or reference bits and they are modulated in the same way as the information bits. During the extraction step, the reference bits are used as anchor points to estimate and compensate for small local and global geometrical distortions. The deformations are approximated using a modified basic optical flow algorithm.

Camcording a screen is a common and easy way for a pirate to capture a protected video and circumvent any cryptography-based protection system like DRM. Therefore, antipiracy methods like watermarking should be robust to this type of attack. One effect very often encountered is the flicker effect: image luminance varies periodically along the video, and scrolling stripes may also be observed. Such an effect is due to aliasing, because the higher frequency backlight of the screen is sampled at a lower rate by the camcorder. The stripe effect appears because each line of the image is captured with a small delay with CMOS sensors. We show that the amplitude of the flicker depends on the luminance of the displayed image, as well as on the luminosity of the screen and on the exposure setting of the camcorder. We propose a method to blindly estimate the flicker parameters by studying the spectrum of the camcorded video; then we show how the flicker can be selectively removed without impairing other frequential components of the video. Experiments on camcorded videos show that removing the flicker enable significant improvement in the watermarking detection performances.

A large portion of pirate movies illegally shared over the Internet is either a camcorded copy of a projection in a digital cinema or is directly ripped from optical discs such as DVDs and Blu-ray discs. In this paper, we introduce a classifier that automatically discriminates between these two types of piracy in an effort to provide tools that help streamlining the whole forensic analysis process. This oracle relies on tell-tale visual artifacts that reveal the occurrence of camcording. We survey three alternate discriminative features relating to temporal flicker, color gamut, and edge orientation and detail how to combine them to obtain accurate classification. Experiments conducted on a large corpus of real pirated movies clearly demonstrate the feasibility of such classification.

To be relevant in copyright protection scenarios, watermarking systems need to provide appropriate levels of security. This paper investigates the security of a popular 3D watermarking method that alters the histogram of distances between the vertices of a surface mesh and its center of mass using a quadratic programming formulation.We study two conventional security mechanisms, namely (i) obfuscating the support of the content used for watermarking and (ii) relying on random projections to obfuscate the watermarking subspace. The different attacks surveyed throughout the paper clearly highlight the limitations of this family of 3D watermarking systems with respect to security.

Watermarking digital video material is usually considered as watermarking a sequence of still images. However, such a frame-by-frame approach is very risky since straightforward embedding strategies can result in poor performance in terms of security i.e., against hostile attacks. As examples, two very common video-watermarking systems will be presented as well as the associated intra-video collusion attacks which defeat them. Then, both watermark modulation and embedding strength modulation will be surveyed to design alternative embedding strategies which exhibit superior performance against such attacks. Nevertheless, it will also be shown that an expert attacker can still construct an effective watermark removal attack. Finally, there will be a discussion to assert whether or not security against intra-video collusion can be achieved with such blind frame-by-frame embedding strategies.

We describe a new watermarking system based on the principles of informed coding and informed embedding. This system is capable of embedding 1380 bits of information in images with dimensions 240/spl times/368 pixels. Experiments on 2000 images indicate the watermarks are robust to significant valumetric distortions, including additive noise, low-pass filtering, changes in contrast, and lossy compression. Our system encodes watermark messages with a modified trellis code in which a given message may be represented by a variety of different signals, with the embedded signal selected according to the cover image. The signal is embedded by an iterative method that seeks to ensure the message will not be confused with other messages, even after addition of noise. Fidelity is improved by the incorporation of perceptual shaping into the embedding process. We show that each of these three components improves performance substantially.

Digital watermarking was introduced at the end of the 20th century to provide means of enforcing copyright protection once the use and distribution of digital multimedia data have exploded. This technology has first been intensively investigated for still images and recent efforts have been put to exhibit unifying characteristics. On the other hand, the situation is rather different in the context of video watermarking, where unrelated articles are scattered throughout the literature. The purpose of this paper is consequently to give an in-depth overview of video watermarking and to point out that it is not only a simple extension of still images watermarking. New applications have to be considered, specific challenges have to be taken up and video-driven approaches have to be investigated.

Collaborative creation workflows are calling for softer protection mechanisms than conventional encryption. Geometry-preserving encryption relates to protection techniques for 3D objects that alter the visual representation of the object after rendering but preserves some geometrical primitives, e.g. the bounding box. The advantage of this approach is the ability of legacy renderers to readily display composite scenes that contain both protected and unprotected objects. In this paper, we introduce a collection of geometry-preserving encryption algorithms, whose visual impact can be adjusted with a tuning parameter, and then report on the influence of the deformation on the actual rendering time of the object. Experimental results clearly highlight that special care needs to be taken in order to preserve interactivity, that is a de facto requirement for collaborative creation.

Content owners are reluctant to deliver premium content to open devices due to the risk of piracy and the corresponding potential for loss of revenue. To lift this deadlock, there is increasing interest in incorporating watermarking-based tracing mechanisms onto mobile devices such as smart phones and tablets. The scarcity of resources calls for innovative watermarking paradigms. In this demo, we will exemplify how two-step bitstream video watermarking could be exploited to address this challenge and why it needs to be adapted to cope with the characteristics of HTTP adaptive streaming.

HTTP adaptive streaming has become the dominant solution for broadband streaming delivery. Essentially, the server delivers small chunks of content in response to queries from a player, that can alternate between different qualities of content depending on its rendering capabilities and the network condition. In this paper, we highlight that the frequent switches of this popular delivery mechanism are deemed to interfere with the payload modulation of bit-stream video watermarking systems. We then propose a generic strategy to cope with such transmission and showcase its efficiency in practice with a watermarking system that operates directly in the H.264/AVC CABAC bit-stream. Eventually, we evaluate the robustness of the proposed solution against standard attacks.

Stereo video content calls for new watermarking strategies, e.g. to achieve robustness against virtual view synthesis. Prior works focused either on inserting the watermark in an invariant domain or on guaranteeing that the watermarks introduced in the left and right views are coherent with the disparity of the scene. However, the first approach raises fidelity issues while the second requires side information at detection i.e. the detector is not blind. In this paper, we propose a new blind detection procedure for disparity-coherent watermarks. In a nutshell, the detector relies on cross-correlation to aggregate the scattered pieces of the embedded reference watermark pattern rather than warping the reference pattern according to the parameters of the current view prior to detection. Reported experimental results indicate that this revisited detector successfully manages to retrieve embedded watermarks even after lossy compression.

Content protection relies on several security mechanisms: (i) encryption to prevent access to the content during transport, (ii) trusted computation environment to prevent access during decoding, and we can also add (iii) forensic watermarking to deter content re-acquisition at rendering. With the advent of next generation video and the ever increasing popularity of embedded devices for content consumption, there is a need for new content protection solutions that rely less on hardware. In this context, we propose an architecture that combines the ARM TrustZone technology, an hypervised environment built on Genode and a bit stream watermarking algorithm that inserts serialization marks on the fly in an embedded device. As a result, an attacker cannot get access to video assets in clear form and not watermarked. Reported performances measurements indicate that the induced computational overhead is reasonable.

International audience; A watermarking strategy for triangle surface meshes consists in modifying the vertex positions along the radial directions, in order to adjust the distribution of radial distances and thereby encode the desired payload. To guarantee that watermark embedding does not alter the center of mass, prior work formulated this task as a quadratic programming problem. In this article, we contribute to the generalization of this formulation with: (i) integral reference primitives, (ii) arbitrary relocation directions to alter the vertex positions, and (iii) alternate distortion metrics to minimize the perceptual impact of the embedding process. These variants are benchmarked against a range of attacks and we report both improved robustness performances, in particular for simplification attacks, and improved control over the embedding distortion.

International audience; Modulating the distances between the vertices and the center of mass of a triangular mesh is a popular approach to watermark 3D objects. Prior work has formulated this approach as a quadratic programming problem which minimizes the geometric distortion while embedding the watermark payload in the histogram of distances. To enhance this framework, we introduce two watermarking components, namely the spread transform and perceptual shaping based on roughness information. Benchmarking results showcase the benefits of these add-ons with respect to the fidelity-robustness trade-off.

Nowadays, forensic watermarks are introduced at projection time in movie theater. This provides the Entertainment industry with a tracing mechanism to identify indelicate cinema owners, who would have let camcorders enter their premises. However, the watermarking technologies deployed in film-based theaters and digital cinemas differ. For cost efficiency during forensic analysis, it is desirable to have a reliable oracle that can accurately discriminate between film and digital projection recordings. Prior work focused on tell-tale artifacts present in the luminance channel to build a classifier. In this paper, we investigate whether color information could also be exploited. We introduce a number of features that relate either to color saturation or color bias, and assess their discriminative power in a systematic fashion. Experimental results on a data set containing nearly 250 real-life full length camcorded movies suggest that the color bias seems to provide the highest classification potential.

International audience; We propose a robust thickness estimation approach for 3D objects based on the Shape Diameter Function (SDF). Our method first applies a modified strategy to estimate the local diameter with increased accuracy. We then compute a scale-dependent robust thickness estimate from a point cloud, constructed using this local diameter estimation and a variant of a robust distance function. The robustness of our method is benchmarked against several operations such as remeshing, geometric noise and artifacts common in triangle soups. The experimental results show a more stable local thickness estimation than the original SDF, and consistent segmentation results on defect-laden inputs.

Video watermarking systems operating directly in the compressed bit stream are receiving increasing interest due to their ease of integration into existing content delivery workflows. Such algorithms, however, usually involve a strong dependency to the underlying video codec. In other words, the throughput of the system, e.g. the embedding rate, is heavily tied to the input bit stream. In this paper, we consider a watermarking system for H.264/AVC CABAC and study its throughput depending on the content mastering process. More precisely, we explore the influence of the compression ratio, the resolution, and some encoding parameters.

Despite the academic focus on blind detection, many forensic watermarking systems actually operate in a non-blind fashion to accommodate for real-life desynchronization attacks. In other words, the original content is needed to perform watermark detection e.g. to re-align the pirate sample with the original content prior to watermark decoding. A somehow intermediate strategy, that could be coined `semi-blind', only requires access to a condensed representation of the original content i.e. content fingerprints. In this paper, we combined a variant of the dynamic time warping algorithm together with multi-dimensional temporal fingerprints to realign pirate video samples. Reported experimental results clearly demonstrate improved accuracy compared to state-of-the-art registration methods for a number of synthetic and real-life attack scenarios.

A number of technologies claim to be robust against content re-acquisition with a camera recorder e.g. water- marking and content ngerprinting. However, the benchmarking campaigns required to evaluate the impact of the camcorder path are tedious and such evaluation is routinely overlooked in practice. Due to the interaction between numerous devices, camcording displayed content modi es the video essence in various ways, including geometric distortions, temporal transforms, non-uniform and varying luminance transformations, saturation, color alteration, etc. It is necessary to clearly understand the di erent phenomena at stake in order to design ef- cient countermeasures or to build accurate simulators which mimic these e ects. As a rst step in this direction, we focus in this study solely on luminance transforms. In particular, we investigate three di erent alterations, namely: (i) the spatial non uniformity, (ii) the steady state luminance response, and (iii) the transient luminance response.

AC-3 audio compression find applications in home cinema, DVD, BD and digital television. This paper addresses the challenge of transparently watermarking AC-3 bit streams instead of the conventional decoding — baseband watermarking — re-encoding approach. We discuss the alternate possibilities to embed watermarks by altering the MDCT coefficients delivered in the bit stream, which consists of mantissas and exponents. We then introduce an AC-3 bit stream watermarking system that is accomplished by solely modifying mantissas to obtain a low computational complexity. We subsequently present the associated watermark detector, which proves to be interoperable with a legacy watermarking system that operates in a different embedding domain. Finally, we compare the proposed bit stream system with the regular decoding — watermarking — re-encoding chain to verify the quality and robustness.

Digital camcording in the premises of cinema theaters is the main source of pirate copies of newly released movies. To trace such recordings, watermarking systems are exploited in order for each projection to be unique and thus identifiable. The forensic analysis to recover these marks is different for digital and legacy cinemas. To avoid running both detectors, a reliable oracle discriminating between cams originating from analog or digital projections is required. This article details a classification framework relying on three complementary features : the spatial uniformity of the screen illumination, the vertical (in)stability of the projected image, and the luminance artifacts due to the interplay between the display and acquisition devices. The system has been tuned with cams captured in a controlled environment and benchmarked against a medium-sized dataset (61 samples) composed of real-life pirate cams. Reported experimental results demonstrate that such a framework yields over 80% classification accuracy.

This paper presents an approach to evaluate the acoustic path transmission of watermarked audio tracks through large scale simulations. The multitude of signal alterations performed implicitly via acoustic path transmission are aggregated through the measurement of impulse responses. These impulse responses are integrated in a test suite in order to be able to perform large scale automated tests as a replacement of the time intensive and expensive individual measurements. The reliability of the approach is demonstrated by the comparison of measurements and simulations.

3D video content has been receiving increasing interest over the last few months and created challenges regarding how to protect such high valued items. For instance, depth-image-based rendering techniques allowing for the creation of virtual views may impair underlying watermarks embedded within individual views. In contrast with previous works, this article focus on defining a view-independent watermark embedding domain rather than a watermarking system coherent with the disparity across several views, thus permitting blind detection. More specifically, based on the fact that pixel displacements across views most often reduce to horizontal shifts, the average luminance value along the rows of the video frames have been selected for watermark embedding. Experimental results clearly prove the robustness of the watermark against virtual views generation.

We propose a new method to detect re-sampled imagery. The method is based on examining the normalized energy density present within windows of varying size in the second derivative of the frequency domain, and exploiting this characteristic to derive a 19-dimensional feature vector that is used to train a SVM classifier. Experimental results are reported on 7,500 raw images from the BOSS database. Comparison with prior work reveals that the proposed algorithm performs similarly for re-sampling rates greater than 1, and is superior to prior work for re-sampling rates less than 1. Experiments are performed for both bilinear and bicubic interpolation, and qualitatively similar results are observed for each. Results are also provided for the detection of re-sampled imagery that subsequently undergoes JPEG compression. Results are quantitatively similar with some small degradation in performance as the quality factor is reduced.

The weighted overlap add (WOLA) technique is routinely used in audio processing to avoid introducing audible clicks. This process introduces inter-dependencies between WOLA coefficients and it is thus necessary to revisit watermarking algorithms to accommodate for such dependencies and prevent introducing self-inflicted interferences. In this paper, we focus on an existing phase quantization based audio watermarking system. Instead of the original computationally expensive ad-hoc process for dealing with the inter-dependencies between WOLA coefficients, we introduce an alternative strategy. Experimental results show a notable reduction of the complexity while maintaining robustness performances.

Psychoacoustic models are routinely used in audio watermarking algorithms to adjust the changes induced by the watermarking process to the sensitivity of the ear. The performances of such models in audio watermarking applications are tightly related to the determination of tonal and noise-like components. In this paper, we present an improved tonality estimation and its integration into a psychoacoustic model. Instead of conventional binary classification, we exploit bi-modal prediction for more precise tonality estimation. Experimental results show improved robustness of the considered audio watermarking algorithm integrating the new tonality estimation, while preserving the high quality of the audio track.

In this industrial exhibit, Technicolor will showcase various technologies developed to protect digital audio-visual material all along the media value chain, from production to distribution, including forensics services in case pirated content is detected. The exhibit will be organized around Technicolor's proprietary technologies in video fingerprinting and audio/video watermarking.

Prior works on re-quantization detection were mainly focused on still images and videos, implying that the involved quantization is uniform. In this paper, we examine non-uniform re-quantization, and then investigate the automatic detection of re-compressed speech signals. Based on Fisher Linear Discriminant (FLD), two detection algorithms are described in the time-domain and in the DFT-domain respectively. Comparative experiments indicate that both detection algorithms produce reliable results with AUC values higher than 0.9744 for a set of different experimental setups. In general, time-domain detection performs slightly better than DFT-domain detection. However, the latter is superior in the less dimensionality of input vectors.

Re-quantization commonly occurs when digital multimedia content is being tampered with. Detecting requantization is therefore an important element for assessing the authenticity of digital multimedia content. In this paper, we introduce three features based on the observation that re-quantization (i) induces periodic artifacts and (ii) introduces discontinuities in the signal histogram. After validating the discriminative potential of these features with synthetic signals, we propose a system to detect JPEG re-compression. Both linear (FLD) and non-linear (SVM) classifications are investigated. Experimental results clearly demonstrate the ability of the proposed features to detect JPEG re-compression, as well as their competitiveness compared to prior approaches to achieve the same goal.

In this paper we argue that improved interoperability between DRM systems is likely to benefit to all the actors in the content value chain. Moreover, we describe a domain-based approach to DRM interoperability. We discuss in some detail configuration aspects of such interoperable domains, and how they affect certain social aspects such as marriage and divorce. We conclude that interoperable domains offer a much needed flexibility, allowing users to more easily move between vertical silos and offering content owners a more flexible environment for selling content.

We compare the performance of three steganalysis system for detection of plusmn1 steganography. We examine the relative performance of each system on three commonly used image databases. Experimental results clearly demonstrate that both absolute and relative performance of all three algorithms vary considerably across databases. This sensitivity suggests that considerably more work is needed to develop databases that are more representative of diverse imagery. In addition, we investigate how performance varies based on a variety of training and testing assumptions, specifically (i) that training and testing are performed for a fixed and known embedding rate, (ii) training is performed at one embedding rate, but testing is over a range of embedding rates, (iii) training and testing are performed over a range of embedding rates. As expected, experimental results show that performance under (ii) and (iii) is inferior to (i). The experimental results also suggest that test results for different embedding rates should not be consolidated into a single score, but rather reported separately. Otherwise, good performance at high embedding rates may mask poor performance at low embedding rates.

International audience; Benchmarking watermarking systems now goes beyond only evaluating the ability of the embedded watermark to with- stand common signal primitives such as filtering, resampling, lossy compression, D/A-A/D conversions, etc. Evaluation procedures have to consider how much information leaks from a watermarking system since such knowledge could prove most helpful to design very powerful attacks. This paper further refines an attack on dirty paper watermarking schemes which relies on security weaknesses i.e. information leakage. In particular, additional constraints are introduced to be able to handle ‘complex' trellises. Moreover, the ef- ficiency of this attack has been evaluated for different trel- lis configurations. Quite counter-intuitively, increasing the number of states in the trellis seems to enhance both the robustness and the security of the system.

Recently Zhang et al described an algorithm for the detection of plusmn1 LSB steganography based on the statistics of the amplitudes of local extrema in the greylevel histogram. Experimental results demonstrated performance comparable or superior to other state-of-the-art algorithms. In this paper, we describe improvements to this algorithm to (i) reduce the noise associated with border effects in the histogram, and (ii) extend the analysis to amplitudes of local extrema in the 2D adjacency histogram. Experimental results on a composite database of 7125 images, averaged over a 20-fold cross validation, with classification based on Fisher linear discriminant analysis, demonstrate that the improved algorithm exhibits significantly better performance. The experimetal results are reported in the form of receiver operating characteristic (ROC) curves and summarized by computing the area under the ROC curve (AUC). The new algorithm, using 10 features derived from the ID and 2D histograms, has an AUC value of 0.77 compared to 0.57 for the original algorithm. It also significantly outperforms other state-of-the-art steganalysers.

Lattice codes and quantization-based techniques have received considerable attention as a method of digital watermarking with side information. However, these algorithms are also recognized to be highly sensitive to common signal processing primitives such as valuemetric scaling, e.g. changes in volume of a song, or re-quantization, e.g. JPEG compression. Hence, it is reasonable to investigate alternative codes which may offer the potential for superior performance. In previous work, we designed a new class of codes, called dirty paper trellis codes (DPTC), which are by construction immune to valuemetric scaling. Although DPTC have shown themselves to be resistant to certain classes attacks, the trellises were generated randomly, thus leading to suboptimal codes. In this paper, Ungerboeck's work on trellis coded modulation (TCM) is exploited to significantly improve the performance of the DPTC. Experimental results show that the resulting trellises significantly outperform the original design.

Considerable progress has been made in the detection of steganographic algorithms based on replacement of the least significant bit (LSB) plane. However, if LSB matching, also known as -1 embedding, is used, the detection rates are considerably reduced. In particular, since LSB embedding is modeled as an additive noise process, detection is especially poor for images that exhibit high-frequency noise - the high-frequency noise is often incorrectly thought to be indicative of a hidden message. To overcome this, we propose a targeted steganalysis algorithm that exploits the fact that after LSB matching, the local maxima of an images graylevel or color histogram decrease and the local minima increase. Consequently, the sum of the absolute differences between local extrema and their neighbors in the intensity histogram of stego images will be smaller than for cover images. Experimental results on two datasets, each of 2000 images, demonstrate that this method has superior results compared with other recently proposed algorithms when the images contain high-frequency noise, e.g. never-compressed imagery such as high-resolution scans of photographs and video. However, the method is inferior to the prior art when applied to decompressed imagery with little or no high-frequency noise.