Joachim J. Eggers

Scalar Costa scheme for information embedding

Research on information embedding, particularly information hiding techniques, has received considerable attention within the last years due to its potential application in multimedia security. Digital watermarking, which is an information hiding technique where the embedded information is robust against malicious or accidental attacks, might offer new possibilities to enforce the copyrights of multimedia data. In this paper, the specific case of information embedding into independent identically distributed (IID) data and attacks by additive white Gaussian noise (AWGN) is considered. The original data is not available to the decoder. For Gaussian data, in 1983, Costa proposed a scheme that theoretically achieves the capacity of this communication scenario. However, Costas scheme is not practical. Thus, several research groups have proposed suboptimal practical communication schemes based on Costas idea. The goal of this paper is to give a complete performance analysis of the scalar Costa scheme (SCS), which is a suboptimal technique using scalar embedding and reception functions. Information theoretic bounds and simulation results with state-of-the-art coding techniques are compared. Further, reception after amplitude scaling attacks and the invertibility of SCS embedding are investigated.

Blind watermarking applied to image authentication

To prevent image manipulations and fraudulent use of modified images, the embedding of semi-fragile digital watermarks into image data has been proposed. The watermark should survive modifications introduced by random noise or compression, but should not be detectable from non-authentic regions of the image The original image cannot be used by the watermark detector to verify the authenticity of the image. In this paper, we investigate the application of a recently developed quantization based watermarking scheme to image authentication. The watermarking technology, called Scalar Costa Scheme (SCS), allows reliable blind watermark detection from a small number of pixels, and thus enables the detection of local modifications to the image content.

Estimation of amplitude modifications before SCS watermark detection

New blind digital watermarking schemes that are optimized for additive white Gaussian noise (AWGN) attacks have been developed by several research groups within the last two years. Currently, the most efficient schemes, e.g., the scalar Costa scheme (SCS), involve scalar quantization of the host signal during watermarking embedding and watermark reception. Reliable watermark reception for these schemes is vulnerable to amplitude modification of the attacked host signal. In this paper, a method for the estimation of possible amplitude modifications before SCS watermark detection is proposed. The estimation is based on a securely embedded SCS pilot watermark. We focus on linear amplitude modifications, but investigate also the extension to nonlinear amplitude modifications. Further, the superiority of our proposal over an estimation method based on a spread-spectrum pilot watermark is demonstrated.

Asymmetric Watermarking Schemes

Unauthorized copying and distribution of digital data is a severe problem in protecting intellectual property rights. The embedding of digital watermarks into multimedia content has been proposed to tackle this problem, and many different schemes have been presented in the last years. However, almost all of them are symmetric, meaning the key used for watermark embedding must be available at the watermark detector. This leads to a security problem if the detectors are implemented in consumer devices that are spread all over the world. Therefore, the development of asymmetric schemes becomes important. In such a scheme the detector only needs to know a public key, which does not give enough information to make watermark removal possible. In this paper, we review recent proposals for asymmetric watermarking and analyze their performance.

Robustness of a blind image watermarking scheme

It was shown previously that, in some situations, blind watermarking can perform as well as watermarking schemes with the host signal available to the decoder. In this paper, blind watermarking of colored Gaussian host signals in the presence of filtering and additive Gaussian noise attacks is discussed. Three suboptimal but practical schemes are compared with a scheme where the host signal is available at the decoder. The performance is analyzed theoretically and experimentally for image watermarking.

Communications approach to image steganography

Steganography is the art of communicating a message by embedding it into multimedia data. It is desired to maximize the amount of hidden information (embedding rate) while preserving security against detection by unauthorized parties. An appropriate information-theoretic model for steganography has been proposed by Cachin. A steganographic system is perfectly secure when the statistics of the cover data and the stego data are identical, which means that the relative entropy between the cover data and the stego data is zero. For image data, another constraint is that the stego data must look like a typical image. A tractable objective measure for this property is the (weighted) mean squared error between the cover image and the stego image (embedding distortion). Two different schemes are investigated. The first one is derived from a blind watermarking scheme. The second scheme is designed specifically for steganography such that perfect security is achieved, which means that the relative entropy between cover data and stego data tends to zero. In this case, a noiseless communication channel is assumed. Both schemes store the stego image in the popular JPEG format. The performance of the schemes is compared with respect to security, embedding distortion and embedding rate.

Quantization effects on digital watermarks

Abstract Digital watermarking is the process of introducing small modifications into a copy of a digital document that can be detected later. The embedded information can be used to determine the documents owner or simply to distinguish several copies. However, coincidental or malicious “attacks” can degrade the robustness of watermark detection. Here, uniform scalar quantization of watermarked documents is investigated theoretically, extending results from theory of dithered quantization, and experimentally. The watermark is embedded by an independent additive pseudo-noise sequence. The statistical distribution of the quantization errors depending on the statistics of the host signal and the watermark is used to determine the robustness of watermark detection via correlation. Experiments with JPEG compression of an image with a DCT-domain additive watermark demonstrate the usefulness of the presented theory.

Analysis of digital watermarks subjected to optimum linear filtering and additive noise

Using a theoretical approach based on random processes, signal processing, and information theory, we study the performance of digital watermarks subjected to an attack consisting of linear shift-invariant filtering and additive colored Gaussian noise. Watermarking is viewed as communication over a hostile channel, where the attack takes place. The attacker attempts to minimize the channel capacity under a constraint on the attack distortion (distortion of the attacked signal), and the owner attempts to maximize the capacity under a constraint on the embedding distortion (distortion of the watermarked signal). The distortion measure is frequency-weighted mean-squared error (MSE). In a conventional additive-noise channel, communication is most difficult when the noise is white and Gaussian, so we first investigate an effective white-noise attack based on this principle. We then consider the problem of resisting this attack and show that capacity is maximized when a power-spectrum condition (PSC) is fulfilled. The PSC states that the power spectrum of the watermark should be directly proportional to that of the original signal. However, unlike a conventional channel, the hostile attack channel adapts to the watermark, not vice versa. Hence, the effective white-noise attack is suboptimal. We derive the optimum attack, which minimizes the channel capacity for a given attack distortion. The attack can be roughly characterized by a rule-of-thumb: At low attack distortions, it adds noise, and at high attack distortions, it discards frequency components. Against the optimum attack, the PSC does not maximize capacity at all attack distortions. Also, there is no unique watermark power spectrum that maximizes capacity over the entire range of attack distortions. To find the watermark power spectrum that maximizes capacity against the optimum attack, we apply iterative numerical methods, which alternately adjust the watermark power spectrum and re-optimize the parameters of the optimum attack. Experiments using ordinary MSE distortion lead to a rule-of-thumb: White watermarks perform nearly optimally at low attack distortions, while PSC-compliant watermarks perform nearly optimally at high attack distortions. The effect of interference from the original signal in suboptimal blind watermarking schemes is also considered; experiments examine its influence on the optimized watermark power spectra and the potential increase in capacity when it can be partially suppressed. Additional experiments demonstrate the importance of memory, and compare the optimum attack with suboptimal attack models. Finally, the rule-of-thumb for the defense is extended to the case of frequency-weighted MSE as a distortion measure.

Channel model for watermarks subject to desynchronization attacks

One of the most important practical problems of blind Digital Watermarking is the resistance against desynchronization attacks, one of which is the Stirmark random bending attack in the case of image watermarking. Recently, new blind digital watermarking schemes have been proposed which do not suffer from host-signal interference. One of these quantization based watermarking scheme is the Scalar Costa Scheme (SCS). We present an attack channel for SCS which tries to model typical artefacts of local desynchronization. Within the given channel model, the maximum achievable watermark rate for imperfectly synchronized watermark detection is computed. We show that imperfect synchronization leads to inter-sample-interference by other signal samples, independent from the considered watermark technology. We observe that the characteristics of the host signal play a major role in the performance of imperfectly synchronized watermark detection. Applying these results, we propose a resynchronization method based on a securely embedded pilot signal. The watermark receiver exploits the embedded pilot watermark signal to estimate the transformation of the sampling grid. This estimate is used to invert the desynchronization attack before applying standard SCS watermark detection. Experimental results for the achieved bit error rate of SCS watermark detection confirm the usefulness of the proposed resynchronization algorithm.

Watermark Detection after Quanization Attacks

The embedding of additive noise sequences is often used to hide information in digital audio, image or video documents. However, the embedded information might be impaired by involuntary or malicious “attacks.” This paper shows that quantization attacks cannot be described appropriately by an additive white Gaussian noise (AWGN) channel. The robustness of additive watermarks against quantization depends strongly on the distribution of the host signal. Common compression schemes decompose a signal into sub-signals (e.g., frequency coefficients) and then adapt the quantization to the characteristics of the sub-signals. This has to be considered during watermark detection. A maximum likelihood (ML) detector that can be adapted to watermark sub-signals with different robustness is developed. The performance of this detector is investigated for the case of image watermark detection after JPEG compression.