Table of Content

30 September 2016, Volume 34 Issue 5

Previous Issue    Next Issue

Special Issue: Information Security of Multimedia Contents

Prospect of Digital Steganography Research

ZANG Xin-peng, QIAN Zhen-xing, LI Sheng

2016, 34(5):  475-489.  doi:10.3969/j.issn.0255-8297.2016.05.001

Asbtract ( 1724 )   PDF (12113KB) ( 939 )  

References | Related Articles | Metrics

Data in cyber-space can be classified into three categories: data for describing objective world, data for recording human behaviors and data for describing virtual world. While current steganographic techniques mainly use the first class of data for hiding secret information, this paper surveys new steganographic techniques using the latter two categories of data as the camouflage, including semi-creative steganography, creative steganography and behavior steganography. Basic concepts and representative methods of the new types of steganography are described. With the development of information and network technology, the new paradigm of digital steganography deserves intensive studies in the future.

Self-Recovery Watermarking for e-Invoice Image Based on Block Classification and RS Coding

WEI Wei, HE Hong-jie, CHEN Fan

2016, 34(5):  490-502.  doi:10.3969/j.issn.0255-8297.2016.05.002

Asbtract ( 1000 )   PDF (19965KB) ( 321 )  

References | Related Articles | Metrics

Taking into account watermark size and quality of recovered image that was tampered, this work proposes a self-recovery watermarking algorithm for e-invoice images based on block classification coding. Image blocks in a test e-invoice image are classified into four types: blocks containing important contents, seal, and invoice format information, and blank ones. The block types and the binary information of all non-blank blocks are coded with RS(1,3) and RS(2,3) respectively to generate a recovery watermark. Based on a key, the recovery watermark is randomly embedded into the blocks in an order of seal, format, and blank until all watermark data are used up. To minimize the size of watermark, the recovery watermark based on the type code and binary information code is used for both authentication and image recovery without additional authentication information. Compared with the recent self-recovery approaches for natural images, the proposed method gives better recovery quality with the same watermark size, and produces a smaller watermark with the same recovery quality, and has better performance both in tamper detection and image recovery against general tampering as well as collage attacks.

Visible Watermarking for 3D Models Based on Boundary Adaptation and Mesh Subdivision

AN Xin-chen, NI Rong-rong, ZHAO Yao

2016, 34(5):  503-514.  doi:10.3969/j.issn.0255-8297.2016.05.003

Asbtract ( 913 )   PDF (20794KB) ( 1090 )  

References | Related Articles | Metrics

Digital watermarking can protect copyrights of digital products, while visible watermarking allows immediate perception so that copyright is openly declared. For threedimensional mesh models, extensive studies on invisible watermarking have been done, but few consider using visible watermarking to achieve direct identification of copyright. This paper proposes a visible watermarking algorithm using a 3D mesh model, which is based on boundary adaptation and grid refining of the 3D mesh model. The algorithm uses a TTF character library in the Windows operating system. It can directly input watermarking information including Chinese characters, English letters, and numbers for embedding. A smooth 3D mesh model is projected to a 2D plane, and then the mesh is broken down to embed the visible watermark. Experiments verify that Chinese characters, English letters and numbers can be shown with smooth edges, and they are robust. Thus copyright of the 3D model is protected.

Texture-Based Image Steganography Using Adaptive Pixel Pair Matching

LIU Shuo-wei, HU Yong-jian, LIU Bei-bei, HUA Xiao-qian

2016, 34(5):  515-526.  doi:10.3969/j.issn.0255-8297.2016.05.004

Asbtract ( 961 )   PDF (9700KB) ( 584 )  

References | Related Articles | Metrics

Compared with flat regions, texture-rich regions in an image are more randomlike, and therefore are more secure for hidden messages. Based on the adaptive pixel pair matching (APPM) scheme, we propose an improved steganographic algorithm that embeds secret data preferentially in texture regions. A criterion for image texture assessment is defined. According to the length of secret data, We use the the criterion to adaptively select appropriate regions for embedding. A pixel value adjustment strategy based on optimizing a minimum distortion function through local search is proposed to deal with possible abnormal situations in embedding. Experimental results show that the proposed algorithm outperforms the original APPM in terms of embedding efficiency and Kullback-Leibler divergence. It is also more robust against several representative steganalytic attacks than a number of least significant bits (LSB) algorithms and APPM.

Coverless Information Hiding Based on Bag-of-Words Model of Image

ZHOU Zhi-li, CAO Yi, SUN Xing-ming

2016, 34(5):  527-536.  doi:10.3969/j.issn.0255-8297.2016.05.005

Asbtract ( 2128 )   PDF (10744KB) ( 756 )  

References | Related Articles | Metrics

This paper introduces a coverless information hiding method based on the bagof-words model (BOW). To hide text information into an image, visual words are extracted to represent the text information. Visual words from an image set are extracted using a BOW model, and a mapping relation between keywords in the text information and visual words is established. Each image is then divided into several sub-images. For each sub-image, a histogram of visual words is computed, and visual words having the largest values in the histogram selected to represent the sub-image. According to the mapping relation, a set of sub-images with visual words related to the text information is found. The images containing these sub-images are used as stego-images for secret communication. The experimental results and analysis show that the proposed method has good performance in anti-steganalysis capability, robustness against common attacks, and security.

Robust Image Hashing Based on Genetic Algorithm and BP Network

JIANG Cui-ling, LIN Jia-jun

2016, 34(5):  537-546.  doi:10.3969/j.issn.0255-8297.2016.05.006

Asbtract ( 1093 )   PDF (10497KB) ( 246 )  

References | Related Articles | Metrics

This paper presents a robust image-hashing scheme based on genetic algorithm (GA) and back propagation (BP) neural network for content authentication. Lifting wavelet transform and discrete Fourier transform are used to extract the image's amplitude spectrum and phase spectrum information to create an image feature matrix. A GA-BP network model is established, and used to generate an image Hash sequence for content authentication. Experimental results show that the proposed Hashing method is robust against content-preserving operations such as random attack, rotation, JPEG compression and additive Gaussian noise. The proposed approach is significantly superior to other algorithms in terms of robustness and discrimination.

Differential Fault Analysis on Feistel Block Cipher in Internet of Things

ZHANG Bo-liang, ZHONG Wei-dong, YANG Xiao-yuan

2016, 34(5):  547-554.  doi:10.3969/j.issn.0255-8297.2016.05.007

Asbtract ( 1106 )   PDF (1742KB) ( 516 )  

References | Related Articles | Metrics

Light-weight block cipher provides protection of RDIF label and smart card in the communication of Internet of things, therefore its security is important. This paper analyzes fault propagation of the traditional light-weight block cipher with a Feistel structure and proposes a deep differential fault analysis strategy. The method is based on the half byte fault attack principle for a Feistel light-weight block cipher. The corresponding key information is obtained according to the difference between fault and correct cipher-texts. Three different attack models are established for specific cryptographic algorithms, i.e., single round half byte fault injection, multiple-rounds half byte fault injection, multiplerounds of two half byte fault injection, to obtain the most efficient way against attacks. Experimental results of the Feistel algorithms, LBlock, are presented.

Steganalysis of JPEG Image Based on High-Dimensional Feature Regularization

SUN Wu-yi, FENG Guo-rui

2016, 34(5):  555-563.  doi:10.3969/j.issn.0255-8297.2016.05.008

Asbtract ( 799 )   PDF (6122KB) ( 450 )  

References | Related Articles | Metrics

Steganalysis plays a vital role in information security. To improve recognition accuracy of JPEG stego-image, we establish a model of scatter matrix eigenspectrum, and purpose a new method to transfer training features of cover and stego images. The eigenspectrum is first modeled by its distribution and divided into three non-overlapping partitions: areas in which the eigenvalue drops rapidly, areas with stable eigenvalues, and areas of zero eigenvalue. The eigenvector is whitened and regularized in the three partitions with adaptive regularization. The result of these steps is a transfer matrix that transfers the input features. The features are now processed and extracted by picking the first t eigenvectors. The processed feature data are then trained with a Fisher linear discriminant (FLD) ensemble classifier. The result shows that recognition accuracy of JPEG stegoimages with the FLD ensemble is improved after the JPEG stego-image features are sorted, regularized and extracted.

Blind Detection of Image Splicing Based on Shadow Model

YANG Bin, CHEN Xian-yi, HU Wei-feng

2016, 34(5):  564-574.  doi:10.3969/j.issn.0255-8297.2016.05.009

Asbtract ( 908 )   PDF (18435KB) ( 202 )  

References | Related Articles | Metrics

Image splicing is a common type of forgery that takes some objects from other images. Splicing is hard to be detected for lack of references. When the objects have shadows, fake shadows are often created to conceal the tampering. A blind forensic method to detect spliced image based on a shadow model is proposed. The method is based on the fact that a natural shadow contains certain features absent in a fake shadow. A shadow model suitable for image forensics is established. Shadows are extracted and located by using an algorithm of image region dilation. A robust shadow scale factor is then obtained by using an energy minimization method, and interfering points removed. Thus, fake shadows can be determined from abnormal values of the shadow scale factor. Experiment results show effectiveness of the proposed method.

Plus-Minus One Modified Product Steganographic Coding Based on Extended Embeddable Columns

ZHANG Ling-yu, ZHAO Xian-feng

2016, 34(5):  575-584.  doi:10.3969/j.issn.0255-8297.2016.05.010

Asbtract ( 800 )   PDF (5940KB) ( 474 )  

References | Related Articles | Metrics

In the existing product coding technique, secret data are embedded into rows and columns of a particular Hamming code matrix in a sequential fashion using the (n, k) linear block codes generated with a parity-check matrix H. The product code is a type of parallel concatenated block code that uses short codes to construct a long code. In decoding, a relatively simple iterative technique is used. This paper proposes a novel algorithm of ±1 modified product steganographic code (±1 MPSC) by expending some embeddable columns. Compared with the product perfect code (PPC), ±1 MPSC uses 2^c-1-1 additional columns where c is the column number of the Hamming code matrix. It also uses a XOR data list constructed by LSB and second LSB of adjacent carrier elements to further reduce the number of flips to improve embedding efficiency. The theoretical analysis and experiments show that the embedding rate of the proposed method reaches 1.33 times that of PPC, and the total performance of D-E curve is improved nearly 2%.

Video Watermarking Robust Against Frame Operations Based on Hamming Code Correction

REN Shao-ying, YUAN Shi-jie, GAO Peng, LU Jian-feng, ZHANG Shan-qing, LI Li

2016, 34(5):  585-597.  doi:10.3969/j.issn.0255-8297.2016.05.011

Asbtract ( 935 )   PDF (10567KB) ( 273 )  

References | Related Articles | Metrics

A frame alignment algorithm is proposed using the (7,4) Hamming code correction to resist these attacking operations. In this algorithm, the watermark is extracted by identifying watermarked frames. The simple linear iterative clustering (SLIC) superpixels method is used to extract the frames' natural characteristics X, which is a 7-bit binary sequence representing the clustering centers obtained using SLIC via preprocessing and quantization. The approach takes advantage of (7,4) Hamming code correction that can hide three bits of the target code S by modifying only one bit. The watermark is then embedded into the video containing alignment marks with DCT-SVD decomposition. Experimental results show that the proposed approach can resist attacks such as frame dropping, frame insertion and recoding. It is also robust against common signal processing operations.

Universal Steganalysis against Adaptive Steganographic Algorithms

LIU Ge, HUANG Fang-jun, LI Zhong-hua

2016, 34(5):  598-604.  doi:10.3969/j.issn.0255-8297.2016.05.012

Asbtract ( 934 )   PDF (2500KB) ( 858 )  

References | Related Articles | Metrics

Digital image is a popular carrier in steganography. In reality, the specific steganographic algorithm used to hide secrete data is generally unknown. Therefore, universal steganalytic techniques capable of detecting hidden information with unknown steganographic algorithms are important. This paper proposes a universal steganalysis technique against adaptive steganographic algorithms for images. With feature extraction and training, the influence of various adaptive steganographic algorithms on statistical characteristics of the carrier image are captured so that unknown steganographic algorithms can be accurately detected. Experimental results show that high detection accuracy can be obtained even for previously unknown steganographic algorithms.

Improved PVO-k Algorithm for Reversible Data Hiding

YAN Bing, BAI Sen, YANG Yi, JI Xiao-yong

2016, 34(5):  605-615.  doi:10.3969/j.issn.0255-8297.2016.05.013

Asbtract ( 1081 )   PDF (4351KB) ( 413 )  

References | Related Articles | Metrics

In recent years, a number of reversible hiding information (RDH) algorithms based on pixel-value-ordering (PVO) and some revised versions have appeared. This paper proposed an improved RDH method to overcome drawbacks in the PVO-k algorithm. The proposed improvement is more suitable for the information embedded into flat blocks due to the improved embedding rules. High fidelity of the stego-image is guaranteed by using the improved pixel-block-selection. Experimental results show that the proposed method has better performance in terms of embedding capacity and image fidelity as compared with PVO-k, especially for relatively smooth images.

Fingerprint Liveness Detection Based on Weber Binarized Perception Features

LÜ Rui, XIA Zhi-hua, CHEN Xian-yi, SUN Xing-ming

2016, 34(5):  616-624.  doi:10.3969/j.issn.0255-8297.2016.05.014

Asbtract ( 839 )   PDF (2526KB) ( 444 )  

References | Related Articles | Metrics

Based on Weber binarized perception features (WBPF), we propose a fingerprint liveness detection method. It consists of two components: local binary differential excitation (LBDE) for extracting perception features by Weber's law, local binary gradient orientation (LBGO) for extracting gradient based orientation features. The features are used to train SVM classifiers on two publicly available databases used in the Fingerprint Liveness Detection Competition 2011 and 2013. Experimental results show that the proposed method outperforms the state-of-the-art liveness detection techniques.

Steganography by Constructing Texture Images

PAN Lin, QIAN Zhen-xing, ZHANG Xin-peng

2016, 34(5):  625-632.  doi:10.3969/j.issn.0255-8297.2016.05.015

Asbtract ( 789 )   PDF (10606KB) ( 445 )  

References | Related Articles | Metrics

This paper proposes a new steganography method using texture construction. The data-hider first constructs a database of elements containing many different patterns and features used to represent secret bits. The data-hider selects patterns and features from the database according to the secret message. The selected elements are painted onto specified positions on a white paper, and some background elements added to the painted paper. With pre-defined transforming functions, the data-hider transforms the painted image to construct a stego-image. On the receiver side, secret messages can be extracted from the stego-image using the inverse operations. Experimental results show that the constructed stego-image has good visual effects, and the secret messages can be properly hidden into the image. Meanwhile, any amount of message can be hidden by choosing different parameters.

Reversible Audio Watermarking Based on Non-causal Prediction and Histogram Shifting

XIANG Shi-jun, CHEN Zhao-long

2016, 34(5):  633-650.  doi:10.3969/j.issn.0255-8297.2016.05.016

Asbtract ( 918 )   PDF (10505KB) ( 948 )  

References | Related Articles | Metrics

Prediction algorithms play an important role in reversible audio watermarking. This paper presents a novel prediction algorithm for reversible watermarking. A prediction parameter is obtained from relations among each group of three consecutive samples, and an optimum prediction coefficient is then calculated by averaging all these parameters in a clip. The prediction value of the current sample and the corresponding prediction error are computed. Since the prediction error is non-integer, we use an improved histogram shifting method to embed the additional data. The proposed scheme can better explore audio samples' correlations by using non-causal prediction and computing prediction parameter in a statistical way. We demonstrate validity of the scheme with a set of six standard audio files. Compared with three state-of-art schemes, the proposed scheme has lower embedding distortion with the same payload.