A Blind Steganalysis on JPEG Gray Level Image Based on Statistical Features and its Performance Analysis

International Journal of Engineering Research and Development
e-ISSN: 2278-067X, p-ISSN: 2278-800X, www.ijerd.com
Volume 11, Issue 09 (September 2015), PP.27-31
27
A Blind Steganalysis on JPEG Gray Level Image Based on
Statistical Features and its Performance Analysis
1
Mrs. Swagota Bera, 2
Dr. Monisha Sharma
1
Associate Professor, Dept. of Electronics & Tele. SSIET, Durg, India
2
Professor, Dept. of Electronics & Tele. SSCET, Bhilai, India
Abstract:- This paper presents a blind steganalysis technique to effectively attack the JPEG steganographic
schemes i.e. Jsteg, F5, Outguess and DWT Based. The proposed method exploits the correlations between
block-DCTcoefficients from intra-block and inter-block relation and the statistical moments of characteristic
functions of the test image is selected as features. The features are extracted from the BDCT JPEG 2-array.
Support Vector Machine with cross-validation is implemented for the classification.The proposed scheme gives
improved outcome in attacking.
Keywords:- Steganography, Steganalysis, Cover image, Stego image, cover Image, Attack, Least Significant
Bit (LSB), DCT,DWT
I. INTRODUCTION
Steganography is the science for secret data concealing. If the data hiding is done after applying DCT
and quantization to the image pixel, comes under the transform domain steganography. Since JPEG ( Joint
Photographic Expert Group) format is the most dominant image format for image storage and exchange at this
time, the JPEG steganography is attracting attention of the researcher .Several steganographic in transform
domain for JPEG images has been developed. In this paper we focus on attacking three well known and most
advanced steganographic methods, i.e., Outguess [1], F5 [2], and the Jsteg [11] DWT[12] . Jsteg[12] is JPEG
hiding technique in which the zero and one coefficient is not used for hiding. OutGuess [1] is a universal
steganographic scheme that embeds hidden information into the redundant bits of data sources. It preserves the
global histogram of BDCT. It adjust untouched coefficient to preserve the histogram. F5[2] works on JPEG by
modifying the block-DCT coefficients to embed messages. This technique is based on straddling and matrix
coding. Straddling scatter the message as uniformly distribution and matrix coding improves embedding
efficiency. DWT Based steganography [12] hides the secret data bits in the wavelet coefficients such that the
global histogram is preserve after hiding. In reverse process detection of hidden data is known as steganalysis.
Various approaches are discussed by the different researchers in the area of steganalysis. Broadly, there are two
approaches to the problem of steganalysis, and one is to come up with a steganalysis method specific to a
particular steganographic algorithm known as embedding algorithm based steganalysis techniques. The other
technique is more general class of steganalysis techniques pioneered independently can be designed to work
with any steganographic embedding algorithm, even an unknown algorithm. Such techniques have been called
universal steganalysis techniques or blind steganalysis techniques.
Features of typical natural images which can get violated when an image undergoes some embedding
process. Hence, designing a feature classification based universal steganalysis technique consists of tackling two
independent problems. The first is to find and calculate features which are able to capture statistical changes
introduced in the image after the embedding process. The second is coming up with a strong classification
algorithm which is able to maximize the distinction captured by the features and achieve high classification
accuracy. Prediction accuracy can be interpreted as the ability of the measure to detect the presence of a hidden
message with minimum error on average. Similarly, prediction monotonicity signifies that the features should
ideally be monotonic in their relationship to the embedded message size. This image features should be
independent on the type and variety of images supplied to it .Embedding techniques affect different aspects of
images.
Farid [3] proposed a universal steganalyzer based on image’s high order statistics .Quadrature mirror filters are
used to decompose the image into wavelet subbands and then the high order statistics are calculated for each
high frequency subband. The second set of statistics is calculated for the errors in an optimal linear predictor of
the coefficient magnitude.
In [6], Shi et al presented a universal steganalysis system. The statistical moments of characteristic
functions of the image, its prediction-error image, and their discrete wavelet transform (DWT) subbands are
selected as features. All of the low-low wavelet subbands are also used in their system. This steganalyzer can
provide a better performance than [3] in general.

A Blind Steganalysis on JPEG Gray Level Image Based on Statistical Features and its Performance Analysis
28
In [4], Fridrich has proposed a set of distinguishing features from the BDCT domain and spatial
domain aiming at detecting information embedded in JPEG images.The statistics of the original image are
estimated by decompressing the JPEG image followed by cropping the four rows and four columns on the
boundary, and then recompressing the cropped image to JPEG format using the original quantization table.
Designed specifically for detecting JPEG steganography. This scheme performs better than [3,5] in attacking
JPEG steganography .
In [7], a new scheme is proposed, in which the inter-pixel and intra-pixel dependencies are used and a Markov
chain model is adopted . The empirical transition matrix of a given test image is formed. The average transition
probability matrix is calculated for the horizontal, vertical, main diagonal and minor diagonal difference JPEG
2-array[4].
The proposed technique is an improved steganalysis scheme to effectively attack the advanced JPEG
steganographic methods. In our scheme, the correlations between block-DCT coefficients in both intra-block
and inter-block sense and the statistical moments of characteristic functions of the test image is selected as
features. The embedding processes often decrease the dependencies of the intra and inter pixel values exiting in
original cover data to some extent. These changes are captured by comparing these statistical parameters. The
first and second order statistical parameters and statistical moment parameter is used as features which is
calculated from JPEG 2-array. Finally we evaluate the proposed features with support vector machines (SVM)
as classifier by conducting experiments over a diverse data set of 4000 JPEG images. The superior results have
demonstrated the effectiveness of our proposed scheme.
The rest of this paper is organized as follows. Section II discusses the proposed scheme for feature
generation. Classification performance results are presented in Section III and conclusions are drawn in Section
IV.
II. PROPOSED SCHEME FOR FEATURE GENERATION
Steganographic embedding causes disturbance on the smoothness, regularity, contuinity, consistency
and periodicity and therefore correlation among the cover image. There exist inter and intra block correlation
among the image pixel which maintain the above features of the image. Any statistical parameter which includes
this relationship may become a good tool for the detection purpose.
First Order Features
The statistical features are calculated from the DCT coefficient .The simplest first order statistic of
DCT coefficients is the histogram. Suppose , dk(i, j) is the DCT coefficient array with quantized value . Q(i, j),
i, j = 1,…,8, k = 1, …, B represents the quantized value of the JPEG file .The symbol dk(i, j) denotes the (i, j)-
th quantized DCT coefficient in the k-th block (there are total of B blocks). The global histogram of all 64k
DCT coefficients will be denoted as Hr, where r = L, …, R, L = mink,i,j dk(i, j) and R = maxk,i,j dk(i, j). Many
of the steganographic programs preserves the global histogram but fails to preserve the histogram of the
individual DCT modes.Thus, we add individual histograms for low frequency DCT modes to our set of
functionals. For a fixed DCT mode (i, j), let hr
ij
, r =L, …, R, denote the individual histogram of values dk(i, j),
k = 1, …, B. We only use histograms of low frequency DCT coefficients because histograms of coefficients
from medium and higher frequencies are usually statistically unimportant due to the small number of non-zero
coefficients. For a fixed coefficient value d, the dual histogram is an 8×8 matrix gij
d
where δ(u,v)=1 if u=v and
0 otherwise. In words , gij
d
is the number of how many times the value d occurs as the (i, j)-th DCT coefficient
over all B blocks in the JPEG image. The dual histogram captures how a given coefficient value d is distributed
among different DCT modes[4].
Second Order Features
The natural images can exhibit higher-order correlations over distances larger than 8 pixels, individual
DCT modes from neighboring blocks are not independent. Thus, the features that capture inter-block
dependencies can be violated by the various steganographic algorithms. Let Ir and Ic denote the vectors of block
indices while scanning the image “by rows” and “by columns”, respectively. The first functional capturing inter-
block de-pendency is the “variation” V defined as
Most steganographic techniques in some sense add entropy to the array of quantized DCT coefficients
and thus are more likely to increase the variation V than decrease. Embedding changes are also likely to increase

29
the discontinuities along the 8×8 block boundaries. In fact, this property has proved very useful in steganalysis
in the past . Thus, we include two blockiness measures Bα, α = 1, 2, to our set of functionals. The blockiness is
calculated from the decompressed JPEG image and thus represents an “integral measure” of inter-block
dependency over all DCT modes over the whole image:
In the expression above, M and N are image dimensions and xij are grayscale values of the decompressed JPEG
image[4].
Statistical Moment Feature
The histogram of an image is essentially the probability mass function (pmf) of the image. Multiplying
each component of the pmf by a correspondingly shifted unit impulse results in the probability density function
(pdf). The pf pdf is exchangeable. Thus, the pdf can be thought as the normalized version of a histogram. The
characteristic function (CF) is the Fourier transform of the pdf .The statistical moment get varies for different
JPEG 2-array coeffiecient . This property is desirable for steganalysis. The statistical moments of the CFs of an
image is defined as follows.
where H(fi) is the characteristic function component at frequency fi, N is the total number of points in the
horizontal axis of the histogram. Note that we have purposely excluded the zero frequency component of the
CF, i.e., H(f0), from calculating the moments because it represents only the summation of all components in the
discrete histogram. For an image, it is the total number of pixels. For a JPEG 2-array, it is the total number of
the coefficients[6].
III. EXPERIMENTS
Image set
An image set consisting of 4000 JPEG images with quality factors ranging of 90 is used in our
experimental work. Each image was cropped (central portion) to the dimension of either 640 X 480. Some
sample images are given in Fig.(1).

30
Fig.(1) Some Sample Images used in this Experimental Work
Stego images generation
On the basis of above approaches the steganalysis algorithm is designed using the MATLAB
software and implemented to the stego image database , where database includes few different images of
different size and formats encoded with JPEG Steganography technique Outguess, F5, Jsteg and DWT Based of
different capacities 0.05 , 0.1, 0.2 bpnc are used[1,2,11,12].
Experimental results for first and second order statistics
The image from the database has been used for both training and testing of the SVM classifier. The
cross-validation technique is used in which 90 % of the data is used for training and rest 10 % is used for testing
purpose. All the images in the dataset becomes the training and testing data simultaneously. Fridrich’s first and
second order[4] , shi’s statistical moment[6] and proposed steganalyzer is implemented for the detection of
Jsteg[11], F5[2], outguess[1] and DWT Based[12]. The classification result is shown in the Table 1. for the
proposed scheme and the obtained result is compared with the existing one in Table 2.
Table 1. Performance of the SVM classifier for the proposed one where bpnc stand for bit per non zero
coefficient, Sen. stands for sensitivity, Pre. stands for precision, Accu. stands for accuracy, Speci. stands
for specifity and Detec. Relia. Stands for detection reliability.
Hiding
Method
Bpnc. Sen. Pre. Accu. Speci. Detec.
Relia.
DWT
Based
0.05 0.656 0.448 65.5518 0.67 0.024
0.1 0.67 0.453 67 0.67 0.004
0.2 0.672 0.456 67.2241 0.67 0.004
Outguess 0.05 0.657 0.442 65.5629 0.66 0.014
0.1 0.656 0.444 65.667 0.66 0.02
0.2 0.663 0.455 66.333 0.67 0.02
F5 0.05 0.666 0.443 66.5541 0.67 0
0.1 0.669 0.447 66.8896 0.67 0
0.2 0.679 0.783 67.893 0.68 0.02
Jsteg 0.05 0.661 0.456 66.1017 0.67 0.024
0.1 0.674 0.459 67.4497 0.68 0.004
0.2 0.695 0.483 69.5205 0.70 0

31
Table 2. Comparison of the Accuracy of the proposed detection technique with the reference.
IV. DISCUSSION AND CONCLUSIONS
(1) In [6] the statistical moment of the test image and their wavelet sub bands are used as features and in [4].
the first order and second order statistical parameters are calculated from the JPEG-2 array of the test
image. Since JPEG which use DCT and quantization is widely used for the data restoration and its
transmission, so in the proposed scheme the statistical moment is calculated form the JPEG-2 array of the
image. For the feature extraction the MATLAB software.
(2) The obtained features are used for the SVM classification with the help of weka data mining software. The
cross-validation is selected for the better result.
(3) For developing Jsteg [11] stego image set, the optimized quantization table is used on the Block DCT
coefficient for the implementation of conventional jsteg.
(4) For DWT based, F5, Jsteg and Ouguess hiding technique, the detection accuracy is higher than [6 ] but
slightly higher than [ 4].
(5) Among all the four the value is high for Jsteg[11] for the proposed one since it is an conventional one which
does not restore the global histogram , but other three outguess[1] , F5[2] and DWT Based [12]have the
global restoration property.
REFERENCES
[1]. http://www.outguess.org/
[2]. http://wwwrn.inf.tu-dresden.de/~westfeld/f5.html.
[3]. Lyu. S. and Farid. H. 2003.Detecting hidden messages using higher-order statistics and support
vector machines. Information Hiding, Springer. 2578: 340–354.
[4]. Fridrich . J. 2005 . Feature based steganalysis for JPEG images and its implications for future
design of steganographic schemes. Information Hiding, Springer. 67–81.
[5]. Farid . H. and Lyu . S. 2006. Steganalysis using higher - order image statistics. IEEE
Trasactions on Information Forensics and Security. 1(1): 111–119.
[6]. Chunhua Chen1, Yun Q. Shi1, Wen Chen1, Guorong Xuan.2006. statistical moments based universal
steganalysis using jpeg 2-d array and 2-d characteristic function.IEEE international conference on
image processing.105-108.
[7]. Shi. Y. Q., Chen. C. and Chen. W. 2007. A markov process based approach to effective attacking
JPEG steganography . Lecture Notes in Computer Science , Information Hiding, Springer. 4437:
249–264.
[8]. Fu. D. Shi. Y. Et.al. 2007. JPEG steganalysis using empirical transition matrix in block DCT
domain. IEEE Workshop on Multimedia Signal Processing: 310–313.
[9]. Chen . C. and Shi. Y. 2008. JPEG image steganalysis utilizing both intrablock and interblock
correlations. IEEE International Symposium on in Circuits and Systems: 3029- 3032.
[10]. Kumar M. 2011. Steganography and steganalysis of joint picture expert group (JPEG) images
.Ph.D. Thesis, University of Florida.
[11]. Bera.S and Sharma .M.2012.Frequency Domain Steganography System Using Modified Quantization
Table. International Journal of Advanced and Innovative Research,1(7):193-196.
[12]. Bera.S and Sharma .M.2013. Development and Analysis of Stego Image Using Discrete Wavelet
Transform. International Journal of Science & Research .2(1): 142-148.
Hiding
Method
bpnc Fridrich's Shie et al's Proposed
DWT
Based
0.05 64.88 60.20 65.5518
0.1 65.00 58.67 67
0.2 65.89 58.86 67.2241
Outguess 0.05 64.90 58.00 65.5629
0.1 65.33 58.61 65.667
0.2 65.67 60.00 66.333
F5 0.05 66.81 57.19 66.5541
0.1 66.89 57.43 66.8896
0.2 67.56 66.89 67.893
Jsteg 0.05 63.73 60.00 66.1017
0.1 63.76 60.07 67.4497
0.2 65.75 62.67 69.5205

A Blind Steganalysis on JPEG Gray Level Image Based on Statistical Features and its Performance Analysis

More Related Content

What's hot (16)

Similar to A Blind Steganalysis on JPEG Gray Level Image Based on Statistical Features and its Performance Analysis (20)

More from IJERD Editor (20)

Recently uploaded (20)

A Blind Steganalysis on JPEG Gray Level Image Based on Statistical Features and its Performance Analysis