Escolar Documentos
Profissional Documentos
Cultura Documentos
DOI 10.1007/s11277-016-3436-0
Tariq Shah1
Abstract In this paper, the system of non-linear ordinary differential equations which
defines a continuous-time dynamical system that shows the fractal characteristics of
attractor is used to construct chaotic S-box. In this new digital watermarking technique the
priority is of importance of robustness along with chaos to create confusion. The inclusion
of chaos along with watermarking in frequency domain ensures robustness. As we have
proposed a frequency domain watermarking as which we embed watermark into the low or
middle frequencies, these changes will be spread all over the image. The strength of
fractional S-box is evaluated with the help of bit independence criterion, nonlinearity
analysis, strict avalanche criterion, linear approximation probability and differential
approximation probability. Additionally, some security analyses in the form of correlation,
contrast, energy, entropy, homogeneity, mean square error and peak signal to noise ratio
are performed for validity of proposed watermarking scheme. The confidence measure
after these analyses indicates the survival aligned with malicious attacks like noise,
cropping and compression.
1 Introduction
The availability of information through global computer networks for the communi-
cation of digital material and documents have changed the standing of digital media.
But free access of digital media communication improperly also provides practically
consummate opportunities to use illegally exclusive rights data. In this proposed
technique, substitution box is constructed and employed for new digital watermarking
123
2034 S. S. Jamal et al.
technique using frequency domain. Over the last millennium, the idea of using a
digital watermark to identify and trace copyright violations with particular ownership
has inspired significant interests among experts of different fields which include
economic, politics, education, entertainment, business and many more. In recent
decades, for secure communication the chaos theory which actually observes the
performance of dynamical systems has been applied in meteorology, engineering,
physics, biology and other fields. Diffusion and confusion are the features of cryp-
tographic structures and chaotic sequences. The unarranged behavior exhibited by
non-linear dynamical systems may be considered as a basis of diffusion [1]. Con-
siderably dissimilar future performance of chaotic systems is observed due to very
small perturbation of the trajectory indicates the sensitivity to initial conditions.
Chaotic communication which uses chaos theory having a feature of unpredictability
planned to offer security in the communication of information attained through
broadcasting tools. The major initiative for combination of chaos and cryptography
was taken by Shanon [2]. Manipulating a new method for chaos based secure com-
munication is an attractive area for researchers. To design new cryptosystems,
properties of chaotic dynamical systems that includes sensitivity to initial conditions
really encourage the duality of chaos and cryptography. The substitution boxes is one
of the applications of chaotic systems in block cryptosystems and an application of
chaotic cryptology [3]. The chaotic systems approach to utilize uncertainty makes it
striking choice in generating mix-up and additional in execution nonlinear alterations
among various approaches of uncertainty. The statistical properties of the plaintext
can be used by applying nonlinear transformation that end result in a remarkable
development of unpredictability, as shown by the scrambled data.
Certain examples of chaotic maps based block ciphers are revealed in literature.
Resistance to cryptanalysis is mainly due to chaotic maps for generation of strong
block ciphers. We may use diverse types of chaotic maps to develop strong block
cipher [4]. Three-dimensional baker’s maps and heuristic approach for substitution box
is being used by Chen [5]. The idea to use chaotic differential equation for generating
the substitution boxes is proposed by Özkaynak et al. [6]. The chaotic algorithms have
shown numerous advantages such as, speed, reasonable computational overheads, high
security and computational power over the conventional algorithms. In this proposed
algorithm, strong cryptographic properties of fractional chaotic Rössler system help us
to generate a S-box having ability of creating confusion and capability of adding
randomness [7]. Then we employed this chaotic S-box to propose a new watermarking
technique in which watermark is embedded by using discrete Fourier cosine transform.
Simulation results are calculated for original and watermark image, statistical analysis
have been done and depict good results. The robustness of proposed algorithm is
analyzed by applying several image processing operations and then tries to extract
watermark which also exhibits good results.
Mathematical background for the chaotic Rössler system and the proposed algorithm
combining S-box and watermarking are discussed in Sect. 2. In Sect. 3, the performance
analysis for the outcome of designed S-box is performed. Section 4 is for the statistical
analysis of original and watermarked image. In Sect. 5 robustness in opposition to image
processing attacks are done. Last section is for conclusion.
123
A Watermarking Technique with Chaotic Fractional S-Box… 2035
Rössler was motivated by the geometry of flows in dimension three and, specifically, by
the reinjection principle, which consists of relaxation-type systems to often present a Z-
shaped slow manifold in their phase space. In dimension three, the reinjection can persuade
chaotic behavior if the motion is spiraling out on one branch of the slow manifold. In this
way, Rössler invented a series of systems which is applied for construction of chaotic
system [8]. Mathematically we have system of differential equations.
dx
¼ y z
dt
dy
¼ x þ ay ð1Þ
dt
dz
¼ b cz þ zx
dt
This dynamical system is basically for creating confusion in the plaintext to accomplish
secure transmission. Rössler system generates chaotic attractor with a single lobe as
compared to the Lorenz attractor which has two lobes. The chaotic behavior of Rössler
system is depicted in Fig. 1. In Eq. (1), three variables ðx; y; zÞ that evolve in the contin-
uous time t and having three parameters (a, b, c). Figures 2, 3 and 4 represent the graphs of
x, y and z variables of Rossler system, respectively. Initial conditions are given as:
xð0Þ ¼ xo ; yð0Þ ¼ yo ; zð0Þ ¼ zo
where xo ; yo ; zo are constants.
The space plots obtained from the equations represented in (1) are given in Figs. 1, 2, 3
and 4. Here the parameters are a = 0.1, b = 0.1 and c = 14. The intervals used in the state
123
2036 S. S. Jamal et al.
Fig. 2 The Rössler system for x along t-axis for a = 0.1, b = 0.1, c = 14
Fig. 3 The Rössler system for y along t-axis for a = 0.1, b = 0.1, c = 14
of the system are 0 x 1600, 0 y 1600 and 0 z 1600. The Rossler system shows
chaotic performance for the particular parameters and intervals.
In Fig. 5 the algorithm of the chaos based S-box design and then watermarking by using
frequency domain is presented. A substitution box or S-box is one of the basic components
123
A Watermarking Technique with Chaotic Fractional S-Box… 2037
Fig. 4 The Rössler system for z along t-axis for a = 0.1, b = 0.1, c = 14
of symmetric key cryptography. In all-purpose, an S-box takes p input bits and convert
them into q output bits. We say it as p q S-box and is shown in Table 1. A p q
Substitution box is basically a function f from p binary inputs to q binary outputs, i.e.,
GF ð2p Þ ! GF ð2q Þ where GF represents Galois field. These S-boxes are vigilantly selected
to defend against linear and differential cryptanalysis. The outcome of Rössler’s system
creates the trajectories and dimension of system indicates the total number of orbits. The
sensitivity of initial conditions is of great importance to produce chaos. Matlab is operated
to get solution of the chaotic system of equations. In order to construct the fractional S-box
which alternate 8 bits of data, every trajectory is sampled at 8 bit resolution and in this way
we achieve 256 different values whose range is from 0 to 255.
Table 1 represents the values of S-box attained by proposed fractional system. The
behavior of chaotic trajectories of the Rössler system in xy plane is depicted in Fig. 1. By
making codes in Mat Lab the answers of S-box are achieved from the preferred trajectories
having specific initial conditions. As we know, the dynamical systems sensitivity to initial
conditions we selected conditions with immense effort to get chaos. The trajectories are
attained from 1000 data samples By substituting watermark image values with the help of
chaotic fractional S-box we obtain the altered watermark image. Now to embed altered
watermark image in host image we have options in the form of spatial and frequency
domain. Because of not as much of robustness ability of spatial domain against dissimilar
attacks it may not be capable to protect image processing attacks although it has the
capacity of hiding data around 6.25 %. The key motive of less significant robustness in
spatial domain is that watermark is not spread around the whole image so watermark can
be altered or destroyed. On the other hand, in spectrum-(or frequency) domain watermarks
are inserted in coefficients of the transformed image. We have the options of discrete
cosine transform (DCT), discrete Fourier transform (DFT) or discrete wavelet transform
(DWT). The embedding scheme of watermarking into the low or middle frequencies, the
whole image examine the change due to all over allocation of watermark as greater part of
the energy lies in the low-frequency components. In our paper, we will apply discrete
123
2038 S. S. Jamal et al.
Fourier transform on original image and then embed watermark in the frequency coeffi-
cients of image.
123
A Watermarking Technique with Chaotic Fractional S-Box… 2039
X
1 M 1 X
N 1 h ux vyi
Fðu; vÞ ¼ f ðx; yÞexp j2p þ ð2Þ
MN X¼0 y¼0 M N
XX
M1 N1 h ux vyi
f ðx; yÞ ¼ Fðu; vÞexp j2p þ ð3Þ
u¼0 v¼0
M N
Algorithm
A.1: By using initial conditions and chaotic parameters, substitution box is obtained
from solution of the Rössler system by mapping each output of x values of numerical
solutions in the range from 0 to 255.
A.2: From the values of watermark image pixels, we generate the indices of the
substitution box. Values from these indices location are replaced with corresponding
original values of image pixels.
A.3: The chaotic y-sequence of Rössler system identify the embedding positions of
S-box substituted watermark into frequency coefficients (obtained by discrete Fourier
transform) of Host image.
A.4: Finally, the watermark image is attained by applying inverse Fourier transform.
The evaluation of fractional S-box make available guarantee of its efficiency and strength
[9, 10]. We analyze proposed chaotic system generated with the help of Rössler system in
Sect. 3. Among different available tests we selected non linearity, strict avalanche crite-
rion, bit independence criterion (BIC), linear approximation probability (LP), and
123
2040 S. S. Jamal et al.
differential approximation probability (DP). Different results indicates that the S-box
approximately fulfill all the parameters near to optimal values. Next subsections corre-
spond to the analyses and brief the results related to the strength of proposed S-box
assessment.
The avalanche effect is viewed if more or less half output bits have changed results due to
single input deviation and owing to this output bits transform the entire substitution-
permutation system view the series of variations [11]. The resistance in opposition to
cryptanalysis and the strength of the cipher are mainly because of above defined changes.
These changes are observed when strict avalanche criterion is applied to the nonlinear
S-box transformation of proposed algorithm. The results of the strict avalanche criterion
are shown in Table 2.
The bit independence criterion (BIC) is analyzed if some of its input bits remain
unchanged. This adjustment of input bits and the avalanche vectors independent perfor-
mance of pair wise variables are the assets of this criterion. The non-linearity of BIC is
tested, and the outcomes are given in Table 3. Table 4 gives the BIC of SAC of proposed
chaotic S-box.
3.3 Nonlinearity
In this analysis, all affine functions are utilized to compare the distance from given
Boolean function. Non-linearity represents the bits which are modified in the truth table to
attain least distance from set of affine function. The nonlinearity calculation is bounded by,
max
NLg ¼ 2n1 1 2n S ð gÞ ðw Þ ð4Þ
w 2 GF ð2n Þ
where
X
SðgÞ ðwÞ ¼ ð1ÞgðxÞ¿:w ð5Þ
wr F2m
123
A Watermarking Technique with Chaotic Fractional S-Box… 2041
The highest value of the disproportion of an event between input and output bits is
enumerated by linear approximation probability. The masks Cl and Cm which shows the
parity of input and output bits respectively are used to get the linear probability. It can be
given as,
123
2042 S. S. Jamal et al.
#fp=p p ¼ Sð pÞ Cn g 1
LP ¼ max ð6Þ
Cm ;Cn 6¼0 2e 2
It is essential to judge the both original image and watermarked image by using available
statistical analyses. In this paper, we look at images with the help of homogeneity, cor-
relation, energy, contrast, entropy, mean square error and peak signal to noise ratio. To find
out the security of images the analyses are made on 256x256 Host images of Lena, baboon
and pepper and 50x50 watermark image. We discuss all analyses in detail and talk about
the obtained results.
4.1 Homogeneity
The ability of combinations of pixel brightness results in tabular form is given by the gray
level co-occurrence matrix (GLCM). The imminence of the distribution in the GLCM to
diagonal of GLCM is focused by using homogeneity. The gray levels frequency is sur-
mised from GLCM table. The homogeneity is given by:
X pði; jÞ
Hom ¼ ð8Þ
1 þ ji jj
4.2 Contrast
Contrast analysis gives help to on looker intensely make out the objects in texture of an
image. Chaotic algorithm better values are obtained with the help of contrast values. It is
defined as
X
C¼ ji jj2 pði; jÞ ð9Þ
4.3 Energy
Energy is obtained by taking square values of ith row and jth column of gray pixels and
sum all those results
123
Table 6 Statistical Analysis Original Image And Watermarked Image
Statistical Proposed technique Amir’s technique
analysis
Pepper Pepper Lena Lena Baboon Baboon Pepper Pepper Lena Lena Baboon Baboon
Host Watermarked Host Watermarked Host Watermarked Host Watermarked Host Watermarked Host Watermarked
Homogeneity 0.8902 0.8902 0.8651 0.8811 0.7294 0.7294 0.8917 0.8916 0.8727 0.8687 0.8423 0.8427
Contrast 0.3311 0.3311 0.4141 0.3371 1.0004 1.0004 0.3181 0.3188 0.3778 0.3857 0.3540 0.3531
Energy 0.1330 0.1330 0.0942 0.1130 0.0817 0.0817 0.1233 0.1220 0.1311 0.1288 0.1387 0.1395
A Watermarking Technique with Chaotic Fractional S-Box…
Entropy 7.5612 7.5613 7.7021 7.7023 7.3903 7.3903 7.6003 7.4115 7.4677 7.2512 7.1872 7.1012
Correlation 0.9207 0.9207 0.9444 0.9443 0.6607 0.6607 0.9295 0.9307 0.8991 0.8933 0.8398 0.8380
2043
123
2044 S. S. Jamal et al.
X
E¼ pði; jÞ2 ð10Þ
4.4 Entropy
Entropy gives the randomness of the digital image as it is the magnitude of the uncertainty
of a random variable. It is worth watching to approximate that whether the digital image is
the same as original image. Mathematically,
X
C¼ pðxi Þ log2 pðxi Þ ð11Þ
The proposed technique employed chaos with the frequency domain embedding of
watermark. This approach to the best of our knowledge is new and has certain advantages
over the previous works. Moreover, the proposed work has been compared with one of the
state of art techniques that uses almost the same methodology [12]. The comparison is
given in Table 6.
4.5 Correlation
The method to check similarity of host image and watermarked image is named as cor-
relation.Mathematically it is represented as:
X ði liÞðj ljÞpði; jÞ
Corr ¼ ð12Þ
rirj
The pixel value of ith row and jth column is ði; jÞ, where i; j represents Image position
pixels. Variance and standard deviation are denoted by r and l respectively. The corre-
lation of different images is given in Table 6.
Two digital images difference is obtained with the help of Mean squared Error (MSE).
Table 7 gives the value of MSE for different images. Mathematically it is defined as
1X
MSE ¼ ðxi x i Þ2 ð13Þ
n
The logarithm of the fraction between the signal strength and difference between the
images (MSE) gives the better comparative statistical analysis. It is named as peak signal
123
A Watermarking Technique with Chaotic Fractional S-Box… 2045
to noise ratio (PSNR). Table 7 gives the value of PSNR for images of Pepper, Lena and
Baboon respectively. We can define it as:
MAXI2
PSNR ¼ 10 log10 ð14Þ
MSE
Pictorial results of proposed embedding and extraction of watermarking technique.
Fig. 7 Watermark
123
2046 S. S. Jamal et al.
In embedding scheme, Fig. 6 represents the Host image and Fig. 7 depicts the watermark
image. The substitution of S-box in watermark is given in Fig. 8 and after following the
watermarking scheme in frequency domain using DFT, the Fig. 9 shows the watermarked
image, which has nearly the identical appearance as of Fig. 7 of original image.
The basic idea of inverse proposed watermark algorithm is authentication. The inverse
algorithm is then applied on Watermarked image shown in Fig. 9. The result of inverse
process is the host image given in Fig. 10. In addition to this, Fig. 11 depicts the substituted
watermark (substituted with S-box) and Fig. 12 is the original watermark image.
The numerical estimate of two watermarks is the similarity of extracted and original
watermark, Cox et al. [13]. High correlation between two watermark images guarantees the
robustness of watermarking algorithm. In addition to this, if the value is on higher side, it
represents the more correlation between two watermark images. It is given as:
Fig. 10 Extracted host image
123
A Watermarking Technique with Chaotic Fractional S-Box… 2047
P
ti :si
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Sim ¼ pP P ffi ð15Þ
ti : s2i :
2
where ti and si represent the corresponding ith element of extracted and original watermark
respectively. The perfect correlation between extracted and original watermark is observed
by seeing the numerical value of confidence measure which is 98.09. With the help of
certain image processing operations on watermarked image and on extracting watermark
we may have the better idea of similarity. Figure 13a–c shows the image processing effects
of compression, noise and cropping respectively on baboon image.
Gaussian, Poisson and speckle are different types of noise attack. We consider and add salt
and pepper noise to check the noise attack on our watermarked image.
Among different compression attack we consider the Joint Photographic Experts Group
(JPEG) for our proposed algorithm.
Table 8 Confidence measure values of selected images against different image processing attacks
Attacks Pepper Lena Baboon
123
2048 S. S. Jamal et al.
6 Conclusion
The best available method for protected copyright of digital contents is Digital water-
marking with certain drawbacks. This new idea shows the importance of utilization of
chaotic system of differential equation for construction of S-box in watermarking. It is
almost impossible to recognize watermark once the watermarking is done with frequency
domain include the chaos from numerical solution of differential equation. The embedding
of watermark (secret signature) has been done in the frequency domain of the original
image whose copyrights is to be protected. The embedding positions are defined by the
random values generated through the chaotic map. The embedding through chaos is our
idea which differentiates our work from others. In addition to this, the outcomes of security
statistical analysis along with robustness test with the help of confidence measure really
support the new idea of watermarking. The similarity after image processing tests which is
between 42 and 77 % proposed scheme is an example of semi fragile watermarking
technique.
References
1. Chen, G., & Ueta, T. (1999). Yet another chaotic attractor. International Journal of Bifurcation and
Chaos, 9(07), 1465–1466.
2. Shannon, C. E. (1949). Communication theory of secrecy systems*. Bell System Technical Journal,
28(4), 656–715.
3. Sam, I. S., Devaraj, P., & Bhuvaneswaran, R. S. (2012). An intertwining chaotic map based image
encryption scheme. Nonlinear Dynamics, 69(4), 1995–2007.
4. Tang, G., & Liao, X. (2005). A method for designing dynamical S-boxes based on discretized chaotic
map. Chaos, Solitons & Fractals, 23(5), 1901–1909.
5. Chen, G. (2008). A novel heuristic method for obtaining S-boxes. Chaos, Solitons & Fractals, 36(4),
1028–1036.
6. Özkaynak, F., & Özer, A. B. (2010). A method for designing strong S-Boxes based on chaotic Lorenz
system. Physics Letters A, 374(36), 3733–3738.
7. Hussain, I., Shah, T., Mahmood, H., & Gondal, M. A. (2012). Construction of S 8 Liu J S-boxes and
their applications. Computers & Mathematics with Applications, 64(8), 2450–2458.
8. Rössler, O. E. (1976). An equation for continuous chaos. Physics Letters A, 57(5), 397–398.
9. Adams, C., & Tavares, S. (1989). Good S-boxes are easy to find. In: Advances in cryptology—
CRYPTO’89 proceedings (pp. 612–615). Springer New York.
10. Wang, Y., Xie, Q., Wu, Y., & Du, B. (2009). A software for S-box performance analysis and test. In:
Electronic commerce and business intelligence, 2009. ECBI 2009. International Conference on (pp.
125–128). IEEE.
11. Khan, M., Shah, T., Mahmood, H., Gondal, M. A., & Hussain, I. (2012). A novel technique for the
construction of strong S-boxes based on chaotic Lorenz systems. Nonlinear Dynamics, 70(3),
2303–2311.
12. Anees, A. (2015). An image encryption scheme based on Lorenz system for low profile applications. 3D
Research, 6(3), 1–10.
13. Cox, I. J., Kilian, J., Leighton, F. T., & Shamoon, T. (1997). Secure spread spectrum watermarking for
multimedia. Image Processing, IEEE Transactions on, 6(12), 1673–1687.
123
A Watermarking Technique with Chaotic Fractional S-Box… 2049
123