Distortion Cost Function Research Articles

Domain Transformation of Distortion Costs for Efficient JPEG Steganography with Symmetric Embedding

Nowadays, most image steganographic schemes embed secret messages by minimizing a well-designed distortion cost function for the corresponding domain, i.e., the spatial domain for spatial image steganography or the JPEG (Joint Photographic Experts Group) domain for JPEG image steganography. In this paper, we break the boundary between these two types of schemes by establishing a theoretical link between the distortion costs in the spatial domain and those in the JPEG domain and thus propose a scheme for domain transformations of distortion costs for efficient JPEG steganography with symmetric embedding, which can directly convert the spatial distortion cost into its JPEG counterpart. Specifically, by formulating the distortion cost function for JPEG images in the decompressed spatial domain, a closed-form expression for a distortion cost cross-domain transformation is derived theoretically, which precisely characterizes the conversion from the distortion costs obtained by existing spatial steganographic schemes to those applied in JPEG steganography. Experimental results demonstrate that the proposed method outperforms other advanced JPEG steganographic schemes, e.g., JUNIWARD (JPEG steganography with Universal Wavelet Relative Distortion), JMiPOD (JPEG steganography by Minimizing the Power of the Optimal Detector), and DCDT (Distortion Cost Domain Transformation), in resisting the detection of various advanced steganalyzers.

Model Access Control Based on Hidden Adversarial Examples for Automatic Speech Recognition

DNNs have achieved remarkable success across various domains, and their commercial value has led to their classification as intellectual property (IP) for their creators. While model watermarking is commonly employed for DNN IP protection, it is limited to post-hoc forensics. In contrast, model access control offers a more effective, proactive approach to prevent IP infringement through authentication. However, existing model access control methods primarily focus on image classification models and are not suitable for automatic speech recognition (ASR) models, which are also widely used in commercial applications. To address the above limitation, inspired by audio adversarial examples, we propose the first model access control scheme for IP protection of ASR models, which utilizes audio adversarial examples with target labels as user identity information, serving as identity-proof samples. However, a unique challenge arises in the form of interception attacks, in which an attacker detects and hijacks an authorized sample to bypass the authentication process. To remedy it, we introduce the hidden adversarial examples (HAEs) for authentication, which embed the authorized information by slightly modifying the logits and behaving like clean audios, thereby making them difficult to be detected by analyzing the predicted results. To further evade detection by steganalysis, which can be employed for adversarial example detection, we design a distortion cost function inspired by adaptive steganography to guide the generation of HAEs. We conduct extensive experiments on the open-source ASR system DeepSpeech, demonstrating that our proposed scheme effectively protects ASR models proactively and is resistant to unique interception attacks.

An efficient distortion cost function design for image steganography in spatial domain using quaternion representation

It is well known that an effective steganographic distortion cost function is essential for spatial image steganography in the framework of minimal distortion embedding. To define the embedding cost by exploiting the texture complexity characterized by the high-pass filtered image residuals has been a popular approach. In this paper, by incorporating the quaternion representation of cover images, the image complexity is evaluated in terms of quaternion magnitude and phase, and a quaternion magnitude-based distortion (QMD) and a quaternion phase-based distortion (QPD) can be defined accordingly. With Hadamard product of QMD and QPD, a novel distortion cost function, namely QMP (Quaternion Magnitude-Phase), is obtained, aiming to efficiently allocate embedding modifications in complex regions of cover images with rich texture contents to improve the statistical undetectability (i.e., security). In addition, the generalized QMP (GQMP) in exponential form is further developed to boost the security performance. Experimental results demonstrate that both the proposed QMP and its generalized variant GQMP can outperform the state-of-the-art schemes, e.g., S-UNIWARD, HiLL, and MiPOD, in terms of empirical security performance against steganalysis.

An anti-steganalysis adaptive steganography for HEVC video based on PU partition modes

Calibration-based steganalysis methods are a fatal threat to current HEVC steganography algorithms. The current steganography algorithms based on prediction unit (PU) still lack resistance to this type of steganalysis. The reason is that the improvements in HEVC codec provide more potential steganographic space as well as introduce more detectable distortion. Therefore, a HEVC video steganography method that has resistance to PU shift-based steganalysis is proposed in this paper. First, the PU shift phenomenon that exists in PU-based steganography is introduced. Then the existing calibration-based steganalysis method is enhanced. According to the encoding rules of HEVC, the distortion cost function according to the difference of rate distortion cost (RD cost) during recompression is designed, which aims to obtain better resistance to calibration-based steganalysis. Experimental results demonstrate that the steganography algorithm designed in this paper provides excellent resistance ability for calibration-based steganalysis.

Adaptive Audio Steganography Based on Improved Syndrome-Trellis Codes

Syndrome-Trellis Code (STC) is a near-optimal convolutional method for adaptive steganography. Hitherto, the existing adaptive steganography commonly depends on the carefully designed distortion cost function, which controls the embedding position of the message in the cover signal. From another point of view, we implement adaptive steganography by improving the STC coding process (named Adaptive-STC). The parity-check matrix is the key to the encoding and extraction process of STC. In this work, we prove that the average embedding change probability of corresponding elements can be changed by adjusting its submatrix. Following that, we specially design an adaptive parity-check matrix to replace the designed distortion cost to restrict the embedding position. The generation of the adaptive parity-check matrix can be formulated as a multi-constrained integer programming problem in which the width of the submatrix is allocated at a fixed height. To solve this particular problem, we propose a targeted intelligent optimization algorithm (named GOAS) that can adaptively generate the parity-check matrix according to different audio cover. The experimental results show that the proposed method outperforms the state-of-the-art adaptive steganography with reduced embedding changes and improved audio quality while ensuring the ability against steganalysis.

Selecting Optimal Submatrix for Syndrome-Trellis Codes (STCs)-Based Steganography With Segmentation

Syndrome-Trellis Codes (STC), as the de facto state-of-the-art steganography framework, has drawn huge attention recently. However, the original STC merely developed with the general form of cover and message, and almost all the subsequent improvements for STC focused on the cover selection and design of the distortion cost function. How to select the optimal submatrix by the features for the given cover and message is an important issue and requires to be solved urgently. In this work, we propose to divide both the cover and message into segments and embed each message segment into its optimally matched cover segment. We name the proposed method as Segment-STC steganography. Specifically, by investigating the features of the submatrix used in STC, we first select the optimal submatrix, which could effectively reduce the number of distorted elements during the embedding. Then, the given cover and message will be split into a series of segments. Each message segment is adaptively matched with an optimal submatrix. We conduct the experiments on both the benchmark BOSS dataset and our collected on-line songs dataset. The experimental results show that, compared with the original STC, our method could effectively reduce the distortion and improve the undetectability of the stego.

Reversible data hiding in JPEG images based on zero coefficients and distortion cost function

Recently, reversible data hiding (RDH) in joint photographic experts group (JPEG) images has received a great deal of attention since the JPEG image is one of the most popularly used image formats nowadays. Generally in JPEG image, the quantized discrete cosine transform (DCT) coefficients with the value of zero are far more than those with nonzero coefficients. However, most existing methods are avoided to change these zero coefficients for fear of JPEG file size increase, and the designed distortion cost functions are not perfect. In this paper, we propose a new JPEG RDH method in which the numerous zero coefficients can be modified. Based on an improved distortion cost function, a strategy which can measure the distortion for each coefficient and select coefficient positions which are most suitable for embedding is proposed. With this strategy, zero coefficients are applied to data embedding and the embedding performance is improved. Experimental results have proved that the proposed method can effectively reduce the embedding distortion and increase the embedding capacity while maintaining a relative good file size.

Hybrid Adaptive Video Steganography Scheme Under Game Model

The steganographic model based on game theory is improved. A hybrid steganography model that satisfies the Kerckhoffs principle by combining the strategy adaptation of the game theory model and the content adaptation in the cover model is proposed. This model is used to design a high-security steganography algorithm for an H.264 video. The distortion function in UNIWARD algorithm is improved according to the H.264 video coding characteristics. The improved distortion function is used to generate the bias function under the game theory model, and the embedding probability of each position in the cover is calculated. The new distortion function under the hybrid model is generated by combining the embedding probability of the cover position and the distortion cost function under the improved UNIWARD algorithm and using it to embed information in the H.264 residual DCT coefficient block. The experiment proves that the algorithm has higher security than the UNIWARD algorithm. With the same analysis algorithm, the detection error rate is significantly improved compared to the UNIWARD algorithm. The algorithm has a large embedding capacity and good invisibility. When fully embedded, the capacity reaches 1.9% or more, the visual effect does not change significantly, and the PSNR value decreases within 2 dB. After the secret message is embedded in the information, the video bit rate is increased by 9%.

Adaptive Spatial Steganography Based on the Correlation of Wavelet Coefficients for Digital Images in Spatial Domain

How to define the pixel’s embedding distortion is a key problem of image adaptive steganography. This paper proposes an adaptive spatial steganography in spatial domain based on the correlation of wavelet coefficients for digital images according to the principle of minimizing a suitable defined distortion. Firstly, one-dimensional high-pass and low-pass filters are used to construct directional filters. Secondly, do direction filtering to images through the direction of horizontal, vertical and diagonal, and design the distortion cost function according to the correlation of wavelet coefficients. Finally, the secret message is embedded by STC (Syndrome Trellis Code) according to the pixel’s embedding distortion. The experiment results illustrate that the proposed method not only make the embedding region focus on the texture regions whose content is complex, but also can resist the common steganalysis detection effectively.

Rendering Distortion Estimation Model for 3D High Efficiency Depth Coding

A depth map represents three-dimensional (3D) scene geometry information and is used for depth image based rendering (DIBR) to synthesize arbitrary virtual views. Since the depth map is only used to synthesize virtual views and is not displayed directly, the depth map needs to be compressed in a certain way that can minimize distortions in the rendered views. In this paper, a modified distortion estimation model is proposed based on view rendering distortion instead of depth map distortion itself and can be applied to the high efficiency video coding (HEVC) rate distortion cost function process for rendering view quality optimization. Experimental results on various 3D video sequences show that the proposed algorithm provides about 31% BD-rate savings in comparison with HEVC simulcast and 1.3 dB BD-PSNR coding gain for the rendered view.

Intra mode decision and frame layer Lagrange multiplier adjustment scheme for H.264 based on structural similarity

H.264 takes rate distortion optimisation (RDO) technique to perform intra and inter mode decision and achieves higher coding efficiency, but the objective distortion metric such as mean square error (MSE) is employed in traditional RDO framework, which cannot acquire optimal subjective quality. In this paper, structural similarity (SSIM)-based subjective distortion is applied to RDO-based intra mode decision in H.264 I frame video coding, and a linear SSIM distortion model is firstly proposed and SSIM-based rate distortion cost function for intra mode decision is defined. Furthermore, a content adaptive frame layer Lagrange multiplier adjustment scheme is proposed to balance the tradeoff between rate and SSIM distortion better. Experimental results show that, the proposed method encodes image structural information more effectively and thus acquires better perceptual quality and subjective RDO performance compared with objective distortion-based RDO method. Under the same perceptual quality, our scheme achieves about 8·03% I frame bit rate reduction on average for various sequences over MSE-based RDO employed in JM reference software.

Phoneme Selective Speech Enhancement Using Parametric Estimators and the Mixture Maximum Model: A Unifying Approach

This study presents a ROVER speech enhancement algorithm that employs a series of prior enhanced utterances, each customized for a specific broad level phoneme class, to generate a single composite utterance which provides overall improved objective quality across all classes. The noisy utterance is first partitioned into speech and non-speech regions using a voice activity detector, followed by a mixture maximum (MIXMAX) model which is used to make probabilistic decisions in the speech regions to determine phoneme class weights. The prior enhanced utterances are weighted by these decisions and combined to form the final composite utterance. The enhancement system that generates the prior enhanced utterances comprises of a family of parametric gain functions whose parameters are flexible and can be varied to achieve high enhancement levels per phoneme class. These parametric gain functions are derived using 1) a weighted Euclidean distortion cost function, and 2) by modeling clean speech spectral magnitudes or discrete Fourier transform coefficients by Chi or two-sided Gamma priors, respectively. The special case estimators of these gain functions are the generalized spectral subtraction (GSS), minimum mean square error (MMSE), two-sided Gamma or joint maximum a posteriori (MAP) estimators. Performance evaluations performed over two noise types and signal-to-noise ratios (SNRs) ranging from <formula formulatype="inline" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex Notation="TeX">${-}$</tex> </formula> 5 dB to 10 dB suggest that the proposed ROVER algorithm not only outperforms the special case estimators but also the family of parametric estimators when all phoneme classes are jointly considered.

General Architecture for MPEG-2/H.263/H.264/AVC to H.264/AVC Intra Frame Transcoding

The latest international video-coding standard H.264/AVC significantly achieves better coding performance compared to prior video coding standards such as MPEG-2 and H.263, which have been widely used in today's digital video applications. To provide the interoperability between different coding standards, this paper proposes an efficient architecture for MPEG-2/H.263/H.264/AVC to H.264/AVC intra frame transcoding, using the original information such as discrete cosine transform (DCT) coefficients and coded mode type. Low-frequency components of DCT coefficients and a novel rate distortion cost function are used to select a set of candidate modes for rate distortion optimization (RDO) decision. For H.263 and H.264/AVC, a mode refinement scheme is utilized to eliminate unlikely modes before RDO mode decision, based on coded mode information. The experimental results, conducted on JM12.2 with fast C8MB mode decision, reveal that average 58%, 59% and 60% of computation (re-encoding) time can be saved for MPEG-2, H.263, H.264/AVC to H.264/AVC intra frame transcodings respectively, while preserving good coding performance when compared with complex cascaded pixel domain transcoding (CCPDT); or average 88% (a speed up factor of 8) when compared with CCPDT without considering fast C8MB. The proposed algorithm for H.264/AVC homogeneous transcoding is also compared to the simple cascaded pixel domain transcoding (with original mode reuse). The results of this comparison indicate that the proposed algorithm significantly outperforms the mode reuse algorithm in coding performance, with only slightly higher computation.

Transform-Domain Rate-Distortion Optimization Accelerator for H.264/AVC Video Encoding

In H.264/AVC video encoding, rate-distortion optimization for mode selection plays a significant role to achieve outstanding performance in compression efficiency and video quality. However, this mode selection process also makes the encoding process extremely complex, especially in the computation of the rate- distortion cost function, which includes the computations of the sum of squared difference (SSD) between the original and reconstructed image blocks and context-based entropy coding of the block. In this paper, a transform-domain rate-distortion optimization accelerator based on fast SSD (FSSD) and VLC-based rate estimation algorithm is proposed. This algorithm could significantly simplify the hardware architecture for the rate-distortion cost computation with only ignorable performance degradation. An efficient hardware structure for implementing the proposed transform-domain rate-distortion optimization accelerator is also proposed. Simulation results demonstrated that the proposed algorithm reduces about 47% of total encoding time with negligible degradation of coding performance. The proposed method can be easily applied to many mobile video application areas such as a digital camera and a DMB (Digital Multimedia Broadcasting) phone.