Digital Video Steganalysis Exploiting Statistical Visibility in the Temporal Domain

In this paper, we present effective steganalysis techniques for digital video sequences based on interframe collusion that exploits the temporal statistical visibility of a hidden message. Steganalysis is the process of detecting, with high probability, the presence of covert data in multimedia. Present image steganalysis algorithms when applied directly to video sequences on a frame-by-frame basis are suboptimal; we present methods that overcome this limitation by using redundant information present in the temporal domain to detect covert messages embedded via spread-spectrum steganography. Our performance gains are achieved by exploiting the collusion attack that has recently been studied in the field of digital video watermarking and pattern recognition tools. Through analysis and simulations, we evaluate the effectiveness of the video steganalysis based on linear collusion approaches. The proposed steganalysis methods are successful in detecting hidden watermarks bearing low energy with high accuracy. The simulation results also show the improved performance of the proposed temporal-based methods over purely spatial methods     

生成对抗网络驱动的图像隐写与水印模型

目的 图像信息隐藏包括图像隐写术和图像水印技术两个分支。隐写术是一种将秘密信息隐藏在载体中的技术,目的是为了实现隐秘通信,其主要评价指标是抵御隐写分析的能力。水印技术与隐写术原理类似,但其是通过把水印信息嵌入到载体中以达到保护知识产权的作用,追求的是防止水印被破坏而尽可能地提高水印信息的鲁棒性。研究者们试图利用生成对抗网络（generative adversarial networks,GANs）进行自动化的隐写算法以及鲁棒水印算法的设计,但所设计的算法在信息提取准确率、嵌入容量和隐写安全性或水印鲁棒性、水印图像质量等方面存在不足。方法 本文提出了基于生成对抗网络的新型端到端隐写模型（image information hiding-GAN,IIH-GAN）和鲁棒盲水印模型（image robust blind watermark-GAN,IRBW-GAN）,分别用于图像隐写术和图像鲁棒盲水印。网络模型中使用了更有效的编码器和解码器结构SE-ResNet（squeeze and excitation ResNet）,该模块根据通道之间的相互依赖性来自适应地重新校准通道方式的特征响应。结果 实验结果表明隐写模型IIH-GAN相对其他方法在性能方面具有较大改善,当已知训练好的隐写分析模型的内部参数时,将对抗样本加入到IIH-GAN的训练过程,最终可以使隐写分析模型的检测准确率从97.43%降低至49.29%。该隐写模型还可以在256×256像素的图像上做到高达1 bit/像素（bits-per-pixel）的相对嵌入容量;IRBW-GAN水印模型在提升水印嵌入容量的同时显著提升了水印图像质量以及水印提取正确率,在JEPG压缩的攻击下较对比方法提取准确率提高了约20%。结论 本文所提IIH-GAN和IRBW-GAN模型在图像隐写和图像水印领域分别实现了领先于对比模型的性能。     

基于条件熵的多链路到达序列编码隐写分析

网络隐写是一种以计算机网络通信数据为载体的隐蔽通信技术。近年来,研究人员对网络隐写的研究重点由原来的注重隐写可实施性向尽量提高隐写隐蔽性转变,基于多链路到达序列编码的隐写即在此背景下提出,目前尚无有效的检测算法。基于此,分析了基于多链路到达序列编码的隐写方法,然后通过对正常多链路数据流的建模分析,提出了基于条件熵的隐写分析方法。实验结果表明,所提方法在观测窗口达到1 200个数据包时对该隐写具有较好的检测效果。     

引入超分辨率下采样误差的图像边信息估计隐写

目的由于空域图像下采样过程中提供的量化误差边信息能够有效提升隐写安全性,为了得到下采样之前的高分辨率图像,提出一种基于超分辨率网络的空域图像边信息估计隐写方法。方法受原始下采样边信息隐写方法的启发,使用超分辨率网络生成被称为预载体的高分辨率图像。同时利用现有的空域图像对称失真算法得到每个像素点的修改失真,然后以浮点型精度对预载体下采样,得到和载体同分辨率的图像形式,利用对应像素点间的差值指导像素点的修改方向,实现基于初始失真的非对称失真调整。首先以峰值信噪比和极性估计准确率为指标对比了多种超分辨率网络以及基于传统插值方法的上采样性能,并通过调整初始失真分别进行隐写和隐写分析实验,选择使安全性提升最大的残差通道注意力机制网络及其对应调整系数作为本文的下采样边信息估计隐写方法。结果使用隐写领域中常用的3个数据库、两种传统初始失真函数以及两类隐写分析方法进行实验。在跨数据集的隐写安全性上,相比传统隐写方法,在对抗基于手工特征和基于深度学习的隐写分析时,本文方法的安全性均有显著提升,如在测试集载体图像上,嵌入率为0.5 bit/像素时,安全性分别提升6.67%和6.9%;在训练集载体图像上,本...     

Current status and key issues in image steganography: A survey

Steganography and steganalysis are the prominent research fields in information hiding paradigm. Steganography is the science of invisible communication while steganalysis is the detection of steganography. Steganography means “covered writing” that hides the existence of the message itself. Digital steganography provides potential for private and secure communication that has become the necessity of most of the applications in today’s world. Various multimedia carriers such as audio, text, video, image can act as cover media to carry secret information. In this paper, we have focused only on image steganography. This article provides a review of fundamental concepts, evaluation measures and security aspects of steganography system, various spatial and transform domain embedding schemes. In addition, image quality metrics that can be used for evaluation of stego images and cover selection measures that provide additional security to embedding scheme are also highlighted. Current research trends and directions to improve on existing methods are suggested.     

一种基于时空冗余统计不可见性的视频隐写分析方法

视频中的扩频嵌入和MSU隐写是两种典型的隐写方法,可抵抗压缩等多种攻击。为有效检测利用此方法嵌入的秘密信息,根据视频的时空冗余特性,提出了一种实时的视频隐写分析方法。用大小为L+1的滑行窗口获取视频帧估计值,并提取相应的DCT和马尔可夫特征,使用神经网络、支持向量机等多种分类方法对隐写视频进行检测。结果表明,根据DCT和马尔可夫特征来分析,检测率较高。将支持向量机和时、空冗余特性等应用于视频隐写分析,有很大前景。     

Steganalysis of stochastic modulation steganography

Steganography[1] conceals the occurrence and real purpose of communication by em- bedding message into the cover medium, such as an image, audio or video. Because im- age is widely used in Internet and has much redundancy, it has been one of the most im- …     

一种基于线性规划特征选择和集成分类器的图像隐写分析方法

隐写分析是防范由隐写术进行信息隐藏所带来危害的有效方法。图像隐写分析方法主要用于检测图像是否被隐写术嵌入隐秘信息。通用型图像隐写分析能够针对广泛类型的隐写术进行检测,该类方法一般采用从图像提取的统计特征和分类器模型进行。当前的高性能隐写分析一般采用高维特征和集成分类器进行。高维特征能够较好地表达图像统计特性中被隐写术扰动的成分,但另一方面,高维特征具有较多的冗余和无效成分,因此进行特征选择能较好的提升效率。本文提出一种使用线性规划的特征选择模型,该模型可与集成分类器协同使用,同时考虑集成分类器中子分类器的检测精度和多个子分类器使用特征的多样性。实验证明,本文提出的方法对多个隐写术的检测性能有较好的提升。     

发送方可否认图像隐写

目的 图像隐写是指将秘密信息隐藏到载体图像中,生成含密图像并在公共信道中传输。隐写分析旨在识别图像中是否隐藏秘密信息。不论何种隐写方案,都会在一定程度上被隐写分析识破,从而导致胁迫攻击,即攻击者找到发送方或接收方,胁迫其提交经过验证的秘密信息。为了保护秘密信息的隐蔽通信,对抗胁迫攻击的可否认方案亟待研究。在密码学领域,为了对抗胁迫攻击,已经提出了可否认加密的概念及相关方案并受到关注与研究。而在隐写领域,有研究提出可否认隐写的概念并设计了接收方可否认隐写的框架,但没有发送方可否认隐写的相关研究。对此,本文讨论发送方可否认隐写。方法 设计方案的通用框架,并构造两个方案:基于可逆网络的发送方可否认图像隐藏方案和基于可否认加密的发送方可否认图像隐写方案。在发送方可否认隐写的框架下,发送方可使用虚假的秘密信息生成与攻击者手中相同的含密图像,以欺骗攻击者,逃脱胁迫攻击,保护真实的秘密信息。结果 实验结果表明,两个方案都是可行且有效的,与原始隐写方案相比,可否认方案达到了发送方可否认功能的同时,均不会显著降低含密图像的视觉质量(峰值信噪比(peak signal-to-noise ratio,PSN...     

负载自适应的批量隐写研究

为了进一步提高批量隐写的安全性, 针对以往自适应批量隐写方案的不足设计了一种简单可靠的新方案。首先借助基于随机森林的集成分类器确定当前隐写分析技术下图像的隐写容量, 在此基础上确定各个图像嵌入的信息量, 方案最大程度地利用了载体资源, 并通过对秘密信息进行分割分组进一步增强了安全性。实验结果表明, 隐写容量计算中的安全判定方法准确性高, 在保证低漏检率的同时避免了较高的虚警率, 且运行时间更短。     

数字图像鲁棒隐写综述

随着智能设备和社交网络的飞速发展,通过网络传输的数字图像成为了实施隐蔽通信的新型重要载体,适应网络信道的图像隐写技术有望成为开放网络环境下可靠、隐蔽传递信息的一种重要方式。然而,数字图像通过Facebook、Twitter、微信、微博等社交网络传输的过程中,往往会遭受压缩、缩放、滤波等处理,对传统信息隐藏技术在兼顾鲁棒性与抗检测性方面提出了新的挑战。为此,研究者经过多年的努力探索,提出了可抵抗多种图像处理攻击和统计检测的新型鲁棒隐写技术。本文结合网络有损信道中隐蔽通信应用需求,对现有的数字图像鲁棒隐写技术进行综述。首先简要介绍本领域的研究背景,并从图像水印和隐写两方面对图像信息隐藏技术的基本概念、相关技术和发展趋势进行了简要总结。在此基础上,将图像鲁棒隐写技术的研究架构分为载体图像选择、鲁棒载体构造、嵌入代价度量、嵌入通道选择、信源/信道编码以及应用安全策略等方面,并分别对相关方法的基本原理进行了归纳和阐述。随后,对具有代表性的相关方法进行了对比测试,并结合应用场景需求给出了推荐的鲁棒隐写方法。最后,指出了数字图像鲁棒隐写技术有待进一步研究解决的问题。     

3D网格隐写与隐写分析回顾与展望

在计算机图形学中,3D形状可有多种表示形式,包括网格、体素、多视角图像、点云、参数曲面和隐式曲面等。3D网格是常见的表示形式之一,其构成3D物体的顶点、边缘和面的集合,通常用于表示数字3D物体的曲面和容积特性。在过去的20年中,基于3D网格载体的虚拟现实、实时仿真和交叉3维设计已经在工业,医疗和娱乐等场景得到广泛应用,以3D网格为载体的水印技术、隐写和隐写分析技术也受到研究者的关注。相比于图像与音视频等载体的隐写,3D网格具备嵌入方式灵活与载体形式多变等其自身的优势。本文回顾了3D网格隐写和隐写分析的发展,并对现有研究工作进行了系统的总结和分类。根据嵌入方式和嵌入位置将隐写算法分成4类：两态调制隐写、最低位隐写、置换隐写和变换域隐写;根据特征提取角度将隐写分析算法分为2类：通用型隐写分析和专用型隐写分析。随后,介绍了每个类别的技术,综合安全性、鲁棒性、容量以及运算效率分析了各类算法的优劣性,总结当前的发展水平,并提供了不同嵌入率下两种数据集上隐写分析算法之间的性能比较。最后讨论了3D隐写和隐写分析现有技术的局限性,并探讨了潜在的研究方向,旨在为后续学者进一步推动3D隐写和隐写分析技术提供指导。     

Kerckhoffs-Based Embedding Security Classes for WOA Data Hiding

It has recently been discovered that using pseudorandom sequences as carriers in spread-spectrum techniques for data-hiding is not at all a sufficient condition for ensuring data-hiding security. Using proper and realistic apriori hypothesis on the messages distribution, it is possible to accurately estimate the secret carriers by casting this estimation problem into a blind source separation problem. After reviewing relevant works on spread-spectrum security for watermarking, we further develop this topic to introduce the concept of security classes which broaden previous notions in watermarking security and fill the gap with steganography security as defined by Cachin. We define four security classes, namely, by order of creasing security: insecurity, key security, subspace security, and stegosecurity. To illustrate these views, we present two new modulations for truly secure watermarking in the watermark-only-attack (WOA) framework. The first one is called natural watermarking and can be made either stegosecure or subspace secure. The second is called circular watermarking and is key secure. We show that circular watermarking has robustness comparable to that of the insecure classical spread spectrum. We shall also propose information leakage measures to highlight the security level of our new spread-spectrum modulations.     

基于拉普拉斯模型的序列隐写密钥估计

图像序列隐写是指利用载体图像特征数据（包括频域数据、空域数据）连续嵌入信息的隐藏方法。本文提出了一种针对图像扩展频谱序列隐写的密钥估计算法。该方法基于序列分析与突变检测的理论,利用序列概率比累积和检验方法（CUSUM-SPRT）对变化进行检测。考虑图像DCT系数满足拉普拉斯分布,给出了理想平稳拉普拉斯分布信号扩展频谱隐藏密钥估计的模型。采用随机微分方程法（SDE）生成拉氏分布的随机序列进行实验。对于非平稳信号的图像数据,在低信噪比（SNR）下,利用当地最有效序列检测法,给出了拉普拉斯分布的密钥估计模型。实验显示,该方法不但能检测出扩展频谱隐写,估计嵌入密钥,而且比Trivedi的方法更有效。     

Digital steganography: hiding data within data

Digital steganography is the art of inconspicuously hiding data within data. Steganography's goal in general is to hide data well enough that unintended recipients do not suspect the steganographic medium of containing hidden data. The software and links mentioned in this article are just a sample of the steganography tools currently available. As privacy concerns continue to develop along with the digital communication domain, steganography will undoubtedly play a growing role in society. For this reason, it is important that we are aware of digital steganography technology and its implications. Equally important are the ethical concerns of using steganography and steganalysis. Steganography enhances rather than replaces encryption. Messages are not secure simply by virtue of being hidden. Likewise, steganography is not about keeping your message from being known - it's about keeping its existence from being known     

基于视觉感知机理的大容量彩色图像盲隐写算法

针对现有彩色图像隐写算法大多将灰度图像的算法直接用于彩色通道,未能运用彩色视觉感知机理的问题,提出一种YUV空间上的、基于视觉感知机理的大容量彩色图像盲隐写算法。主要方法是在Y分量上通过标准差来分析载体图像的局部复杂性,利用视觉感知机理将信息分成两部分隐写于Y和V分量中,灵活性较差的U分量则作为隐写信息的通道指示器,无需大的改变。大量实验结果表明,所提算法与其他方法相比在嵌入大容量信息后依然能够保持很好的不可感知性,而且可以抵抗直方图对比和RS隐写分析方法,因而所提算法是合理、简洁、有效的。     

Digital video steganalysis by subtractive prediction error adjacency matrix

Video has become an important cover for steganography for its large volume. There are two main categories among existing methods for detecting steganography which embeds in the spatial domain of videos. One category focuses on the spatial redundancy and the other one mainly focuses on the temporal redundancy. This paper presents a novel method which considers both the spatial and the temporal redundancy for video steganalysis. Firstly, model of spread spectrum steganography is provided. PEF (Prediction Error Frame) is then chosen to suppress the temporal redundancy of the video content. Differential filtering between adjacent samples in PEFs is employed to further suppress the spatial redundancy. Finally, Dependencies between adjacent samples in a PEF are modeled by a first-order Markov chain, and subsets of the empirical matrices are then employed as features for a steganalyzer with classifier of SVM (Support Vector Machine). Experimental results demonstrate that for uncompressed videos, the novel features perform better than previous video steganalytic works, and similar to the well-known SPAM (Subtractive Pixel Adjacency Model) features which are originally designed for image steganalysis. For videos compressed with distortion, the novel features perform better than other features tested.     

一种通用数字水印模型

本文提出了基于信息加密技术、扩频通信技术、视觉识别特性及噪声理论的一种通用数字水印模型，并简要介绍了几种水印方法。     

Steganalysis Frameworks of Embedding in Multiple Least-Significant Bits

Replacement of least-significant bit plane is one of the popular steganography techniques in digital images because of its extreme simplicity. But it is more difficult to precisely estimate the rate of secret message embedded by replacement of multiple least-significant bit (MLSB) planes of a carrier object. In order to model the MLSB embedding, a lemma is introduced to prove the transition relationships among some trace subsets. Then, based on these transition relationships, two novel steganalysis frameworks are designed to detect two kinds of distinct MLSB embedding methods. A series of experiments show that the proposed steganalysis frameworks are highly sensitive to MLSB steganography, and can estimate the rate of secret message with higher accuracy. Furthermore, these frameworks can fully meet the need to distinguish stego images under low false positive rate, especially when the embedded message is short.      

Digital watermarking of text, image, and video documents

The ease of reproduction, distribution, and manipulation of digital documents creates problems for authorized parties that wish to prevent illegal use of such documents. To this end, digital watermarking has been proposed as a last line of defense. A digital watermark is an imperceptible, robust, secure message embedded directly into a document. The watermark is imperceptible both perceptually and statistically. Robustness means that the watermark cannot be removed or modified unless the document is altered to the point of no value. The watermark is secure if unauthorized parties cannot erase or modify it. Current watermarking schemes may be viewed as spread-spectrum communications systems, which transmit a message redundantly using a low-amplitude, pseudo-noise carrier signal. An example highlights the basic mechanisms and properties of spread spectrum and their relation to watermarking. Finally, specific issues in watermarking of text, images, and video are discussed, along with watermarking examples.