基于双向马尔可夫模型的JPEG图象隐写分析

提出了一种基于双向马尔可夫模型的JPEG图象的通用隐写分析方法,利用量化后分块DCT系数的中低频系数间的相关性,提取DCT块内和块间的特征,采用阈值贝叶斯分类方法进行识别,并且与SVM分类器的识别效果进行了比较。针对4种公认的JPEG嵌入方法——F5、Outguess、MB1和MB2进行隐写分析,在CorelDraw图象库上做实验,取得了很好的性能。     

基于多样性特征的JPEG图像隐写分析

随着隐写分析技术的发展,新的特征提取算法不断出现,但目前还没有一种较好的通用特征能对JPEG图像进行有效的隐写分析。针对上述问题,提出一种从多域空间提取特征的通用隐写分析算法。采用残差共生矩阵与直方图统计函数计算DCT域、空域、小波域各域系数(像素)之间的依赖性关系,并结合校准方式从中提取特征。对多样性特征维数高的问题,采用前向选择与穷举结合的方法对其降维,以提高分类精度与节约分类时间。对4种典型的JPEG隐写算法在小嵌入率下进行实验,结果表明,与已有的检测方法相比,多域空间提取的多样性特征检测准确率能提高2%以上,适应性更广。     

基于马尔可夫模型和特征融合的图像隐写分析

提出一种针对JPEG图像隐写的通用隐写分析方法.根据量化后分块DCT系数绝对值构造水平、垂直和zigzag方向的差分数组,利用三向差分数组马尔可夫模型挖掘量化后分块DCT块内邻近系数相关性,提取转移概率矩阵的特征.对三向特征加权融合后进行隐写分析,以提高分类性能.对安全性较高的JPEG隐写OutGuess和F5,在不同嵌入率下进行隐写分析.实验结果显示,引入特征融合后隐写分析的检出率明显提高.     

Feature mining and pattern classification for steganalysis of LSB matching steganography in grayscale images

In this paper, we present a scheme based on feature mining and pattern classification to detect LSB matching steganography in grayscale images, which is a very challenging problem in steganalysis. Five types of features are proposed. In comparison with other well-known feature sets, the set of proposed features performs the best. We compare different learning classifiers and deal with the issue of feature selection that is rarely mentioned in steganalysis. In our experiments, the combination of a dynamic evolving neural fuzzy inference system (DENFIS) with a feature selection of support vector machine recursive feature elimination (SVMRFE) achieves the best detection performance. Results also show that image complexity is an important reference to evaluation of steganalysis performance.     

基于MCM和HVS的彩色图像盲隐写分析算法

针对典型算法在描述DCT域相关性方面存在的不足,所引起的隐写检测综合性能不理想的问题,将MCM与HVS进行有机结合,提出一种JPEG彩色图像通用隐写分析算法。设计8-邻域MCM型,全面描述DCT系数之间的相关性,基于HVS分别在YCbCr模型空间Y分量抽取Markov状态转移矩、在其相应RGB三通道抽取Markov状态转移矩的主对角线邻域相似熵作为特征统计量,并进行合理“绑定”,采用PCA技术对其进行优化选择,构建高效分类特征向量。实验结果表明该算法对于Jsteg、F5、Outguess、MB1、MB2攻击具有较高的可靠性和检测正确率。     

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。     

基于对称α稳态模型的JPEG通用隐写分析

针对JPEG隐写术检测问题,提出一种基于DCT系数分布模型的通用隐写分析方法。根据DCT系数统计分布规律,建立双参数对称α稳态模型,并将模型参数估计问题转化为离散的函数参数优化问题,利用改进的遗传算法进行求解。提取分布模型的参数作为隐写分析的特征,并运用图像校准技术进行校准以提高特征的敏感性。设计与特征相匹配的、具有线性阶计算复杂度的单类分类器进行隐写判别。实验结果表明,该方法可以有效地检测JPEG隐写术,当嵌入率为25%时,平均检测率达到76.1%,相比传统基于模型的隐写分析方法提高5.5%,具有更高的检测性能。     

A novel blind detector for additive noise steganography in JPEG decompressed images

This paper proposes a novel universal steganalyzer for additive noise steganography in JPEG decompressed images. On the basis of the influence of the message embedding on the statistical distributions of alternating current (ac) discrete cosine transform (DCT) coefficients, we first develop a new steganalytic feature which is defined as the ratio between different ranges of the normalized ac coefficients histogram. Then a powerful blind detector is constructed with the proposed one-dimensional (1-D) feature. Extensive experimental results demonstrate that the proposed steganalyzer outperforms the existing state-of-the-art schemes significantly and even can detect the additive noise steganography effectively at a very low embedding rate. In addition, our method using a 1-D feature is not only practical and real-time, but also can provide a better control of the false positive rate and the false negative rate by adjusting the detection threshold. Moreover, the proposed feature can also be used to identify JPEG compression besides steganalysis, which indicates that our method has a great promise in practical applications.     

基于DCT域共生矩阵的JPEG图像隐写分析

提出一种新的针对JPEG图像的通用隐写分析方法，该方法直接提取DCT系数，利用共生矩阵去挖掘出块中低频系数的相关性，最后形成120维特征并用SVM进行分类。针对4种公认的安全性较高的JPEG类嵌入方法F5、Outguess、MB（未去除分块特性MB1，去除分块特性MB2），对CorelDraw1096张图库进行了实验，结果表明，该方法的识别率和运算速度明显优于现有算法。     

Detection of Double-Compression in JPEG Images for Applications in Steganography

This paper presents a method for the detection of double JPEG compression and a maximum-likelihood estimator of the primary quality factor. These methods are essential for construction of accurate targeted and blind steganalysis methods for JPEG images. The proposed methods use support vector machine classifiers with feature vectors formed by histograms of low-frequency discrete cosine transformation coefficients. The performance of the algorithms is compared to selected prior art.     

彩色图像通用隐写分析的多类统计特征

提出了用于彩色图像通用隐写分析的3类统计特征。针对隐写所导致的图像DCT系数分布、空域像素值平坦性以及颜色空间一致性的改变,从系数分布模型背离程度、空域相关性和不同颜色分量间的相关性3个方面设计能够体现原始载体和含密图像差异的特征,提取了10维特征向量。这些特征对隐写行为较为敏感、涵盖面广,且维数低、计算方便。在这些特征的基础上,可利用支持向量机分析彩色JPEG图像是否含有秘密信息。实验结果表明,基于这些特征的通用隐写分析方法可以准确、高效地检测以Jstge、F5或MB隐写方法嵌入的秘密信息,辨别隐写方法种类,具有很好的通用性。     

Calibration based universal JPEG steganalysis

For steganalysis of JPEG images,features derived in the embedding domain appear to achieve a prefer-able performance. However,with the existing JPEG steganography,the minor changes due to the hid-den secret data are not easy to be explored directly from the quantized block DCT (BDCT) coefficients in that the energy of the carrier image is much larger than that of the hidden signal. In this paper,we present an improved calibration-based universal JPEG steganalysis,where the microscopic and macro-scopic calibrations are combined to calibrate the local and global distribution of the quantized BDCT coefficients of the test image. All features in our method are generated from the difference signal be-tween the quantized BDCT coefficients of the test image and its corresponding microscopic calibrated image,or calculated as the difference between the signal extracted from test image and its correspond-ing macroscopic calibrated image. The extracted features will be more effective for our classification. Moreover,through using the Markov empirical transition matrices,both magnitude and sign dependen-cies along row scanning and column scanning patterns existed in intra-block and inter-block quantized BDCT coefficients are employed in our method. Experimental results demonstrate that our proposed scheme outperforms the best effective JPEG steganalyzers having been presented.     

基于流形学习的JPEG图像定量隐写分析算法

提出了一种基于DCT系数统计特性的JPEG图像定量隐写分析算法。该算法在对JPEG图像DCT系数的统计模型进行研究的基础上,提取了能够反映嵌入容量变化规律的特征参数[α]。以特征参数[α]为基础,提出了基于流形学习的特征提取算法,通过LIB-SVM分类器进行训练,估计隐写对DCT系数的更改比率。实验结果表明,与传统的定量分析算法相比,提出的算法具有更高的估计准确率和稳定性。     

基于支持向量机的电路故障诊断模型

提出了一种基于遗传编程和支持向量机的故障诊断模型。通过遗传编程对时域指标进行特征选择和提取，得到更能反映信号本质的特征信号，该特征信号可作为识别特征输入多类支持向量机，实现对模拟电路不同类型软故障的识别。实验结果表明，同传统时域指标相比，经过遗传选择和提取的特征对模拟电路的软故障具有更好的识别能力，进而提高了多类支持向量机的分类准确性。     

针对特定测试样本的隐写分析方法

现今主流的图像隐写分析方法主要聚焦于设计检测特征,用以提高通用盲检测（UBD,Universal Blind Detection）模型的检测准确率,这类检测方法与待测图像无关,难以做到精准检测。本文在拥有大数据训练资源的前提下,研究了隐写对图像特征的影响,找出了隐写分析与图像特征之间的重要关系,基于此提出了一种为测试样本选择专用训练集的隐写分析方法。以经典的JPEG隐写算法nsF5和主流的JPEG隐写分析特征（CC-PEV、CC-Chen、CF^*、DCTR和GFR）为例组织实验,结果表明本文方法的检测准确率高于其他同类方法。     

改进的小波高阶统计量的JPEG图像隐写分析

提出了一种针对JPEG图像隐写的盲检测改进方法。直接把JPEG图像的DCT系数分解到小波域,计算出各个子带小波系数的一系列高阶统计特征。对得到小波高阶统计特征进行校准,以使统计特征能更好地反映信息嵌入以后的变化。校准后的统计特征作为图像最后的特征。应用基于核技巧的支持向量机SVM作为分类器,判断图像是否含有隐藏信息。针对两种典型的JPEG类嵌入方法Outguess和JP Hide & Seek进行了实验,结果表明该方法有效地实现了信息隐藏的盲检测分析。     

基于图像相邻像素相关性的LSB匹配隐写分析

LSB匹配隐写是图像隐写分析中的重点研究问题。根据图像相邻像素的相关性, 提出了一种新的隐写分析算法。通过图像复原算法计算出复原图像, 利用高阶Markov链模型分别对待检测图像和复原图像建模, 根据LSB匹配隐写对高阶Markov链模型经验矩阵的影响, 提取复原图像和待检测图像的统计特征组合成新的27维特征向量对支持向量机进行训练。实验表明提出的算法对LSB匹配隐写有较好的分析效果, 特别在嵌入率低的情况下, 算法具有较好的分析能力。     

一种JPEG图像的DCT域隐写分析方法

提出了一种基于JPEG图像DCT系数差分矩阵统计特征的隐写分析方法。该算法保留了以往算法选用的DCT系数水平和竖直方向的差分矩阵相关特征,通过增加其主副对角线方向上差分矩阵来提取和计算特征向量,进而利用SVM分类器进行分类。实验结果表明,该方法能够有效地对JPEG图像进行检测,并且具有较高的检测正确率。     

以直方图差值序列为特征的JPEG MB2隐写分析

基于JPEG图像的离散余弦变换(DCT)系数直方图差值序列,提出一种新的检测JPEG MB2隐写术的方法。该算法定义JPEG图像的DCT系数一阶直方图与其Cauchy拟合分布模型的系数直方图之间的差值序列作为相关性度量,在此基础上构造区分隐写图像和载体图像的分类特征序列。在特征分析中,通过Hilbert-Huang变换对特征序列进行经验模式分解,构建了基于Hilbert谱的特征向量。实验结果表明,基于该特征向量的支持向量机(SVM)对MB2隐写算法检测平均准确率达到84.48%,该方法也适用于JPEG MB1的检测。     

Digital image splicing detection based on Markov features in DCT and DWT domain

Image splicing is very common and fundamental in image tampering. To recover people's trust in digital images, the detection of image splicing is in great need. In this paper, a Markov based approach is proposed to detect this specific artifact. Firstly, the original Markov features generated from the transition probability matrices in DCT domain by Shi et al. is expanded to capture not only the intra-block but also the inter-block correlation between block DCT coefficients. Then, more features are constructed in DWT domain to characterize the three kinds of dependency among wavelet coefficients across positions, scales and orientations. After that, feature selection method SVM-RFE is used to fulfill the task of feature reduction, making the computational cost more manageable. Finally, support vector machine (SVM) is exploited to classify the authentic and spliced images using the final dimensionality-reduced feature vector. The experiment results demonstrate that the proposed approach can outperform some state-of-the-art methods.