Nonlinear Dyn (2014) 75:807–816

DOI 10.1007/s11071-013-1105-3

O R I G I N A L PA P E R

A technique for digital steganography using chaotic maps

Amir Anees · Adil Masood Siddiqui ·
Jameel Ahmed · Iqtadar Hussain

Received: 4 July 2013 / Accepted: 30 September 2013 / Published online: 31 October 2013
© Springer Science+Business Media Dordrecht 2013

Abstract Chaos has been applied extensively in se- signal to noise ratio, and mean square error have also
cure communication over the last decade, but most been carried out and shown that it can survive against
of the chaotic security protocols defined, are crypto- various differential attacks such as the known message
graphically weak or slow to compute. Also, study of attack, known cover attack, known stego attack, and
chaotic phenomena as application in security area is stego only attack.
not discussed in detail. In this paper, we have intensely
studied chaos, their influence in secure communica- Keywords Chaotic sequence · Secure
tions and proposed a steganography technique in spa- communication · Statistical analysis
tial domain for digital images based upon chaotic
maps. By applying chaos effectively in secure com-
munication, the strength of the overall anticipated al- 1 Introduction
gorithm has been increased to a significant level. In
addition, few security statistical analyses such as cor- Steganography is the science that falls under the cat-
relation, entropy, energy, contrast, homogeneity, peak egory of secure communications. Its original theme
is to visually hide the secret data in a carrier dur-
ing communication [1–3]. It can be interwoven with
A. Anees (B) · A.M. Siddiqui two other types of security systems, namely cryptog-
Department of Electrical Engineering, Military College raphy and watermarking. The purpose of cryptogra-
of Signals, National University of Science and Technology,
phy and steganography is the same, i.e., to conceal
Islamabad, Pakistan
e-mail: [email protected] the information message, but the methodologies em-
ployed in these techniques are different. The method-
A.M. Siddiqui
e-mail: [email protected] ology of watermarking and steganography is the same,
but the purposes of these techniques are different. Wa-
J. Ahmed termarking deals with copyright protection of digital
Department of Electrical Engineering, HITEC University,
data while steganography concerns about the hiding of
Taxila, Pakistan
e-mail: [email protected] digital data. Although for last couple of millennium,
the science of cover writing (steganography) has been
I. Hussain used in diversified forms [4], but can be comprehen-
Department of Sciences and Humanities, National
sively divided into two categories. The first one is lin-
University of Computer & Emerging Sciences, Islamabad,
Pakistan guistic steganography, which uses the natural language
e-mail: [email protected] as a carrier in hiding the original message and widely

Downloaded from http://www.elearnica.ir

808 A. Anees et al.

Fig. 1 Generalized
proposed chaotic
steganographic algorithm
illustrating the basic idea of
embedding information
message into two parts of a
carrier

used in ancient time. In the early fifth century BC, His- Data hiding in Least Significant Bits (LSBs) is a
taiacus tried to conceal the secret information in the simple and straightforward technique that has also the
form of a tattoo made on slave’s shaved head and dis- advantage of embedding more information than the
patched after his hair grew back with the message [5]. transform techniques keeping the texture of digital
The second class, technical steganography, which is image almost unaffected. But the problem with spa-
more of a modern category employs the digital data as tial techniques is the low robustness against statistical
carrier, such as images, video, audio, and text. analysis and differential attacks. In this paper, we have
With the evolution of chaotic theory in recent employed chaos to propose a chaotic steganographic
decades, chaos has been applied extensively in se- algorithm in spatial domain, which not only proved se-
cure communications. Chaotic dynamics are the im- cure against different attacks, but also exhibited good
promptu behavior exhibited by some nonlinear dy- statistical analysis. Message, in binary data is broken
namical systems and can be used as a source of dif- into LSBs and MSBs and are placed in a carrier in its
two different areas, respectively, and further at random
fusion in security techniques [6]. It has been shown
positions in these areas defined by chaotic sequences.
that chaotic security algorithms have commended sev-
Figure 1 illustrates this idea.
eral advantages such as high security, speed, rea-
sonable computational overheads, and computational
power over the traditional algorithms. Also, steganal-
2 Mathematical modeling and chaotic maps
ysis which includes chosen stego attack, stego-only
attack, known stego attack, known cover attack, and Let suppose C denotes the carrier signal in which mes-
statistical measures of chaotic steganographic algo- sage signal, denoted by M has to be placed. A stego-
rithms illustrate exceptional results over the custom- function ϕ takes the input C and M and generates
ary algorithms. However, at the same time, this ex- stego-signal, S.
ponential growth in chaotic security area leads to the
rapid publication of those papers, which are not se- S = ϕ(C, M) (1)
cure enough and shows several flaws, especially the
early proposed analog chaotic security approaches. The carrier C can be in any form, from human beings
Recently published, a number of papers represent sta- themselves [2] to multimedia such as image, audio,
tistical analysis of chaotic security systems showing and video files. At the receiving end, original informa-
the flaws in the strength of secure algorithms and can tion message signal, M retrieved by applying inverse
stego-function, ϕ −1 on stego-signal S.
be easily broken in short computer times. Also, the
performance analysis and security issues did not take M = ϕ −1 (S) (2)
due attention in proposing these techniques, which pay
off as to be weak against differential attacks, while the The strength of any steganographic algorithm lies in
safer ones cannot be effectively implemented on the the stego-function, ϕ. Different people from differ-
given hardware. ent scientific backgrounds has presented various stego
A technique for digital steganography using chaotic maps 809

functions and showed the strength of their algorithms.

Some of the earlier proposed methods are hidden tat-
toos, invisible inks, covered writings, null ciphers,
character arrangement, code words, forms of covert
channels, digital signatures, microdots, and spread-
spectrum communications.
In the proposed algorithm, ϕ has used three chaotic
sequences, namely Logistic, Tangent Delay Ellipse
Reflecting Cavity Map System (TD-ERCS) and Non-
linear Chaotic Algorithm (NCA) to embed the infor-
mation signal, M at different positions of carrier sig-
nal, C.
To show the strength of proposed algorithm, it is
necessary to demonstrate the importance of chaotic se-
quences and their influence in secure communication, Fig. 2 Bifurcation diagram of logistic map, i.e., graph plot-
take the example of the logistic chaotic map. ted for logistic sequence against each value of parameter r for
The logistic map is a model of population growth r = 1 to 4 with spacing of 0.0005 and with initial condition of
x0 = 0.5 and r = 3.7
first proposed in [7]. It is derived from the continuous
form of differential equation defined as

dx Figure 2 shows the graph of logistic sequence, xn

= rx(1 − x) (3) verses r. It assembles all the information of xn against
dt
r for r = 1 to 4 in one figure. The values of parameter
The discrete version in the form of difference equation r are presented on horizontal axis from left to right
is described as and values of logistic sequence against each value of
r with spacing of 0.0005 are plotted on vertical axes.
xn+1 = rxn (1 − xn ) (4) The sequences are plotted after 40 iterations to see the
long term behavior against each value of r. Based on
where the initial parameters are
Fig. 2, the interval r can be divided into three segments
listed below [11]:
r ∈ (0, 4)
x0 ∈ (0, 1) • When r ∈ (0, 3), the iterative sequence for xn
reaches to fixed value after some number of iter-
The parameter r is the rate of population growth, ations showing the stable state. Figure 3(a) shows
or in physical term, defines the rate of heating in a the plot of logistic sequence for r = 1 to 3. It can
convection equation or may be velocity of fluid in a be visualized that for every value of r, the sequence
mechanical rotating circle of convection. The charac- remains constant on single value. This effect is men-
teristic of logistic equation is heavily dependent upon tioned in Figs. 3(b)–(d), where sequence xn is plot-
parameter r. Robert May [8–10] analyzed at length the ted against number of iterations for different values
conduct of logistic equation based upon r. After plot- of r = 1.1, 1.5, 2.0, 2.5, respectively.
ting the performance of logistic iterative parameter xn • When r ∈ [3, 3.6), the iterative sequence swings be-
as a function of r, it was noticed that when r is low, tween several points and system shows periodic-
the map settles on steady state after some iterations. ity. Figure 4(a) shows the plot of logistic sequence
When r is high, the stable state breaks into bifurcation, for r = 2.9 to 3.6. At r = 3, the logistic sequence
into a two-state periodic form; this bifurcation is fur- swings between two values showing periodicity of
ther divided into a four- state periodic form and then period two as illustrated in Figs. 4(b) and (c) for
into eight. On added value of r, the map sequence en- r = 3 and r = 3.3, respectively. By increasing the
ters into an unpredictable behavior region, the chaotic value to r = 3.5, the periodicity increased to four as
region. can be seen in Fig. 4(d) and at r = 3.55, the period-
810 A. Anees et al.

Fig. 3 Bifurcation diagram of logistic map for r = 0 to 3 show- gistic sequence against number of iterations for different values
ing the stable behavior (a) showing the plot for logistic se- of r = 1.1, 1.5, 2.0, 2.5, respectively
quence against parameter r and (b)–(d) shows the plot for lo-

icity further increases by the factor of two and the small difference, i.e., x0 = 0.10000 and x0 = 0.10001.
sequence swings between the eight values as men- The two graphs matched detail for detail until the 22nd
tioned in Fig. 4(e). iteration, then they started to separate apart and by the
• When r ∈ [3.6, 4), the periodicity disappears from end of 50th iteration, they are completely out of phase
iterative sequence and exhibits chaotic behavior. with the initial difference as small as one part in a ten
Figure 5(a) shows the graph plotted for logistic se- thousand.
quence against values of r from r = 3.5 to 4, i.e.,
Not only have map sequences differ for different
zooming the Fig. 2 for the chaotic sequence. Fig-
values of x0 , but also due to the different values of r.
ures 5(b)–(d) shows the plots for different values of
r showing the chaotic sequences against the number Figure 6(b) shows two graphs laid on each other for
of iterations. logistic map with constant value of x0 and two values
of r with small difference, i.e., r = 3.7000 and r =
One of the essential properties of any chaotic se-
3.7001. Again numeric data of logistic map behaves
quence is sensitive dependence on initial conditions.
as same as earlier, two graphs have same values until
It is must for every chaotic map to exhibit this prop-
the 22nd iteration and then stays differ vigorously up
erty as for the logistic map as well. After sketching
the graph for logistic map with slightly different initial to 50th iteration.
values, we can easily visualize the effect of sensitive This property makes chaotic maps effective in se-
dependence on initial values. Figure 6(a) illustrates cure communication. It is very difficult to predict the
two graphs laid on each other for logistic map with initial conditions based upon chaotic sequences fetch
constant value of r = 3.7 and two values of x0 with after cipher text attack; this is the reason why our an-
A technique for digital steganography using chaotic maps 811

Fig. 4 Bifurcation diagram of logistic map for r = 3 to 3.5 quence with periodicity two for r = 3.0, 3.3, (d) periodicity four
showing the periodic behavior (a) showing the plot for logis- for of r = 3.5, and (e) periodicity eight for of r = 3.5
tic sequence against parameter r, (b)–(c) shows the periodic se-


ticipated algorithm is so strong against the traditional y0 = μ 1 − x02 (8)
algorithms.
x0 2
The rest of two algorithms have the same properties k0 = − μ (9)
described earlier for logistic map; the mathematics of y0
these maps are defined as follows. tan α + k0
TD-ERCS system was developed in 2004 by Yuan k0 = − (10)
1 − k0 tan α
Sheng Lee [12, 13] defined as ⎧
⎪
⎪ μ ∈ (0, 1]
⎧ ⎪
⎪
⎪ 2kn−1 yn−1 + xn−1 (μ2 − kn−1
2 ) ⎨x ∈ [−1, 1]
⎪
⎪ = − 0
⎨ n
x are the system seed parameters
μ2 + kn−1
2 ⎪
⎪ α ∈ (0, π)
⎪
⎪
⎪
⎪ 2kn−m − kn−1 + kn−1 kn−m
2 ⎩m = 2, 3, 4, 5, . . .
⎪
⎩kn = , n = 1, 2, 3, . . .

1 + 2kn−1 kn−m − kn−m
2

(5) The Nonlinear Chaotic Algorithm (NCA) map was

proposed especially for image encryption [11] and
where stated as

− xyn−1 μ2 n<m

kn−m = n−1 (6)  α 1 β
− xyn−m μ2 n≥m xn+1 = 1 − β −4 . cot . 1+
n−m 1+β β
yn = kn−1 (xn − xn−1 ) + yn−1 (7) × tan(αxn ).(1 − xn )β (11)
812 A. Anees et al.

Fig. 5 Bifurcation diagram of logistic map for r = 3.56 to 4 gistic sequence against number of iterations for different values
showing the chaotic behavior (a) showing the plot for logistic of r = 3.75, 3.80, 3.85, 3.90, respectively
sequence against parameter r and (b)–(d) shows the plot for lo-

Fig. 6 Iterative sequence of logistic map showing the property ferent initial values of x = 0.10000 and x = 0.10001. (b) Plot
of sensitive dependence on initial conditions. (a) Plot of logistic of logistic map with constant initial value of x = 0.1 and two
map with constant initial value of r = 3.7 and two slightly dif- slightly different initial values of r = 3.7000 and r = 3.7001
A technique for digital steganography using chaotic maps 813

where the seed parameters can be 4 Statistical security analysis

⎧
⎪ Some of the statistical security analyses such as cor-
⎨xn ∈ (0s, 1)
relation, entropy, energy, contrast, homogeneity, peak
α ∈ (0, 1.4]
⎪
⎩ signal to noise ratio, and mean square error have
β ∈ [5, 43] been done on the proposed algorithm for original and
steganographic to show the strength of it. The re-
or
sults for first five analyses for red, green, and blue
⎧ frames are listed in Tables 1–3, respectively. The val-
⎪
⎨xn ∈ (0, 1) ues for MSE and PSNR as a comparison of original
α ∈ (1.4, 1.5]
⎪
⎩
and steganographic images are listed in Table 4.
β ∈ [9, 38]
4.1 Correlation
or
⎧ The most fundamental method used in determining the
⎪
⎨xn ∈ (0, 1) similarity between two images is the correlation anal-
α ∈ (1.5, 1.57] ysis. The correlation of an image is given as
⎪
⎩
β ∈ [3, 15] (i − μi)(j − μj )p(i, j )
Corr = (12)
σj σj
i,j

3 Proposed algorithm and simulated results where i, j corresponds to image pixels positions;
p(i, j ) is pixel value at ith row and j th column; μ
is the variance; σ is the standard deviation.
The proposed algorithm shown in the form of flow-
chart in Fig. 7 employs in spatial domain and embeds 4.2 Entropy
the information in LSBs of carrier image. The carrier
image is first broken into two parts, upper and lower, Entropy is a magnitude of the uncertainty of a random
respectively. The information signal is converted into variable to come in a random process and can be used
binary of eight bits and split into four MSBs and four to show the randomness of the digital image as well.
LSBs. The information MSBs is embedded into up- Entropy is defined as
per part of carrier signal and LSBs are in its lower n
part. The chaotic maps engaged in proposed algo- H =− p(xi ) log2 p(xi ) (13)
rithm define the exact positions in upper and lower i=1
part for embedding of information bits, that is, TD-
where p(xi ) is the probability of random variable x at
ERCS map defines the row number, NCA map defines
ith index.
the column number, and the logistic map defines the
frame number. After selecting the specific pixel of car- 4.3 Homogeneity
rier image for embedding information, it is converted
into binary and split into MSBs and LSBs. Its LSBs The homogeneity analysis processes the closeness of
are then replaced with respective information bits. Af- the distribution in the gray level cooccurrence matrix
ter embedding each symbol of information, upper and (GLCM) to the GLCM diagonal. The GLCM shows
lower parts of the carrier joins and makes the stegano- the measurements of combinations of pixel brightness
graphic image. The LSBs are replaced of the carrier values or gray levels in tabular form. The frequency
image and are considered as a loss of information or of the patterns of gray levels can be inferred from the
regarded as an addition of noise in it. For extraction GLCM table. The homogeneity can be determined as
of information, reverse methodology is applied. Fig-
p(i, j )
ure 8(a) shows the original image and Fig. 8(b) shows Hom = (14)
1 + |i − j |
the steganographic image. i,j
814 A. Anees et al.

Fig. 7 Flowchart of proposed chaotic steganographic algorithm

A technique for digital steganography using chaotic maps 815

Fig. 8 (a) Original image.

(b) Steganographic image
achieved by applying
proposed steganographic
algorithm on original
baboon image

Table 1 Statistical security analysis for red frame of original Table 3 Statistical security analysis for blue frame of original
and steganographic images and steganographic images

Statistical analysis Original Steganographic Statistical analysis Original Steganographic

Correlation 0.8582 0.8582 Correlation 0.8698 0.8698

Entropy 7.6840 7.6840 Entropy 7.7072 7.7072
Homogeneity 0.7781 0.7781 Homogeneity 0.7754 0.7754
Contrast 0.6420 0.6421 Contrast 0.6598 0.6599
Energy 0.0760 0.0760 Energy 0.0771 0.0771

Table 2 Statistical security analysis for green frame of original 4.5 Energy
and steganographic images

Statistical analysis Original Steganographic

The energy of the image gives the sum of squared val-
ues of gray pixels of a digital image defined as
Correlation 0.7827 0.7827
Entropy 7.3887 7.3887 E= p(i, j )2 (16)
Homogeneity 0.7817 0.7817 i,j

Contrast 0.6449 0.6449

4.6 MSE
Energy 0.0958 0.0958

Mean Squared Error (MSE) is used to measure the dif-

ference between two digital images. It can be defined
4.4 Contrast as
n
The contrast analysis of the image enables the viewer 1 
Xi − Xi∗
2
MSE = (17)
to vividly identify the objects in texture of an image. n
i=1
The contrast values for chaotic algorithm give bigger,
and thus better values. The contrast of an image is 4.7 PSNR
given as the following.
Pick Signal to Noise Ratio (PSNR) has the same func-
C= |i − j |2 p(i, j ) (15) tion as MSE, but it takes the signal strength and di-
i,j vided it by noise strength or the difference between
816 A. Anees et al.

Table 4 Statistical security analysis of MSE and PSNR 5 Conclusion

Frame MSE PSNR
The proposed technique is strong as measured of hid-
Red 0.0441 55.4126 ing the information in a digital image. The stegano-
Green 0.0108 55.4126 graphic image is almost identical to the original image
Blue 0.0025 55.4126 making very hard for a third party to deduce the in-
formation from it. This effect can be verified by the
different statistical analysis done in this paper, which
the images (MSE), thus gives the better comparative shows almost the same values for both original and
statistical analysis. It is given as steganographic images. Furthermore, at the same time,
MAX2i the anticipated technique is robust against differential
PSNR = 10. log10 (18) attacks as well.
MSE
The above established statistical analysis not only help
in determining the strength of steganographic algo- References
rithm, but also can assists the attacker during its at-
tacks. Some of the possible differential attacks are the 1. Johnson, N.F., Jajodia, S.: Exploring steganography: seeing
following. the unseen. Computer 31(2), 26–34 (1998)
2. Judge, J.C.: Steganography: past, present, future. SANS
4.8 Known cover attack Institute publication. http://www.sans.org/reading-room/
whitepapers/stenganography/552.php (2001)
The original cover image, C is known. The attacker 3. Provos, N., Honeyman, P.: Hide and seek: an introduction
to steganography. IEEE Secur. Priv. 1(3), 32–44 (2003)
then compares C with the steago image, S, to deter- 4. Sadkhan, S.B.: Cryptography: current status and future
mine the difference between these two images. In this trends. In: Proceedings of IEEE, International Conference
case, the attacker takes well used of statistical analy- on Information & Communication Technologies: From
sis stated above to get out the information hidden in S. Theory to Applications, Damascus, Syria, pp. 417–418
(2004)
However, the statistical analysis done for the proposed
5. Moulin, P., Koetter, R.: Data-hiding codes. Proc. IEEE
technique is far better for this type of attack. 93(12), 2083–2126 (2005)
6. Yang, T., Wu, C.W., Chua, L.O.: Cryptography based on
4.9 Known message attack chaotic systems. IEEE Trans. Circuits Syst. I 44(5), 469–
472 (1997)
The original message M is known. With only M 7. Verhulst, P.F.: Recherches mathématiques sur la loi
and S, though it can help the attacker in future attacks, d’accroissement de la population. Nouv. Mem. Acad. R.
Sci. B.-Lett. Brux. 18, 1–41 (1845)
but it is very difficult because the message embedded
8. May, R.M.: Biological populations with nonoverlapping
at random positions is defined by the chaotic map and generations, stable points, stable cycles, and chaos. Science
can be changed by slightly varying the initial values of 186, 645–647 (1974)
it. 9. May, R.M.: Biological populations obeying difference
equations, stable points, stable cycles, and chaos. J. Theor.
4.10 Known stego attack Biol. 51(2), 511–524 (1975)
10. May, R.M.: Simple mathematical models with very com-
plicated dynamics. Nature 261, 459–467 (1976)
The steganographic algorithm or stego-function, ϕ is
11. Gao, H., Zhang, Y., Liang, S., Li, D.: A new chaotic algo-
known. In this case, the work for the attacker is very rithm for image encryption. Chaos Solitons Fractals 29(2),
easy to attain the hidden message. 393–399 (2006)
12. Sheng, L.-Y., Sun, K.-H., Li, C.-B.: Study of a discrete
4.11 Stego-only attack chaotic system based on tangent-delay for elliptic reflecting
cavity and its properties. Acta Phys. Sin. 53(9), 2871–2876
Only the stego image, S, is known and taken as a result (2004)
13. Sheng, L.-Y., Cao, L.-l., Sun, K.-h., Wen, J.: Pseudo-
of interception between the source and destination; to
random number generator based on TD-ERCS chaos and
access the hidden message only from S is far difficult its statistic characteristics analysis. Acta Phys. Sin. 54(9),
as compared to the above mentioned attacks. 4031–4037 (2005)