Public key steganography using lsb method with chaotic neural network

International INTERNATIONAL Journal of Computer JOURNAL Engineering OF and COMPUTER Technology (IJCET), ENGINEERING ISSN 0976-6367(Print),
&
ISSN 0976 - 6375(Online), Volume 5, Issue 9, September (2014), pp. 50-61 © IAEME
TECHNOLOGY (IJCET)
ISSN 0976 – 6367(Print)
ISSN 0976 – 6375(Online)
Volume 5, Issue 9, September (2014), pp. 50-61
© IAEME:

Journal Impact Factor (2014): 8.5328 (Calculated by GISI)

50

IJCET
© I A E M E
PUBLIC KEY STEGANOGRAPHY USING LSB METHOD WITH CHAOTIC
NEURAL NETWORK
Adel A. El-Zoghabi1, Amr H. Yassin2, Hany H. Hussien3
1Head, Department of Information Technology, Institute of Graduate Studies and Research,
Alexandria University
2Lecturer, Electronics Communication, Engineering Department, Alexandria Higher Institute of
Engineering and Technology
3PHD students, Department of Information Technology, Institute of Graduate Studies and Research,
Alexandria University
ABSTRACT
The art of information hiding has been around approximately as long as the need for secret
communication. Steganography, the concealing of information, get up early on as an extremely
useful method for covert information transmission. Steganography is the art of hiding secret image
within a larger image, this is in contrast to cryptography, where the existence of the secret image is
not disguised, but the content cover image is well obscure. The goal of a steganographic method is to
minimize the visually apparent and statistical differences between the cover data and a stego-image
while maximizing the size of the payload. This paper explains about how a secret image can be
hidden into an image using least significant bit insertion method with chaotic map as public stego-key
insertion along with chaotic neural network for merging methods.
Keywords: Chaotic Map, Chaotic Neural Network, Data Hiding, Public key, Steganography.
1. INTRODUCTION
Several digital services require more consistent security in storage and transmission for
digital images transferring. Attributable to the rapid growth of the internet, especially in the digital
world, the security of digital images has become more important and attracted much attention. The
propagation of multimedia technology in our society has promoted the digital images to play a more
significant role than the traditional texts, which request earnest protection of users' privacy for all
applications. Public key encryption and steganography techniques of digital images are very essential
and should be used to avoid opponent attacks from unauthorized access [1] [2].

International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print),
51

Chaos theory is considered to be brand new between sciences, and has been showed to be
showed to be the most powerful fields in presence. Chaos become at the most interested research in
physical systems, which make it implemented in many fields such as computer science, robotics,
economics, and psychology. The chaos science able to absorbed the scientific information which
may change the science hub in the future. In chaos theory, the term “Chaos” is not an opposite of
absence of order but in fact have a very intelligent order in itself not quite obvious as ordered
systems [3] [4].Chaotic signals have several properties that make them attractive candidate for
communication, such as wideband spectrum, waveform does not accurately repeat itself, random-like
appearance [5].
In the steganography the secret information can’t be detectable, because, the secret data is
hidden in a cover file which has notrigger for eavesdropper's suspicion to discover [6]. Capacity,
robustness, undetectable, invisibility, security and resistance against attacks are the main criteria in
steganography. The steganography is a form of security through ambiguity, by hidden the write
message (stego – object) in another object (host file) [7].Also statistical properties of the file, before
and after information embedding should have minimum differences. The cover-object for
steganography could be image, video clip, text, music and etc.
Due to the redundancy of data in digital images, a little change in the information cannot be
detected by the naked eye, which makes the digital images a good covers for steganography. Several
methods are used for image steganography such as Discrete Cosine Transform (DCT), Discrete
Wavelet Transform (DWT), which known as transform domain, and Least Significant Bit (LSB),
which known as spatial domain. In the first two methods the pixels are transformed into coefficients,
and then the secret bits are embedded in these coefficients, while the third one, the secret bits are
embedded directly in LSB image pixels. In the transform domain, major robustness is provided
against changes and attacks, where special domain has high embedding capacity [7].
The least significant bit (LSB) insertion is one of the most widely used methods for
embedding a message in a digital image. Steganography involves hiding information so it appears
that no information is hidden at all. Therefore, it is expected that the person will not be able to
decrypt the information. An alteration of the least significant bit of the color value of some pixels in
an image will not change the quality of the image significantly. Therefore, a message can be sent
within an image using these bits [8].
There are three types of steganographic: pure steganography, secret key steganography and
public key steganography. A steganography system that doesn’t need any exchange information
between secret communications before sending the message is named pure steganography. This
system it’s not provide any security if the attacker know the embedding method. On the other hand in
secret key steganography, the system is similar to a symmetric cryptosystem, where the sender and
the receiver must have a secret key to share between them. Used in the embedding process. The
receiver can reverse the process and extract the secret message. Public key steganography is similar
to asymmetric cryptosystem, where two keys are needs to process. One of keys is private (secret) and
the other is public, which are store in a public database. The public key is used in the embedding
process. The secret key is used to rebuild the secret message [9].
Neural network can be used to design a real time data protection, and hiding schemes for its
complicated and time-varying structures. Due to the attractive properties of combination neural and
the sensitively to initial value conditions and parameters of chaotic maps, a chaos- based neural
network created a combination model called a chaotic neural network (CNN) which given a novel
and efficient way for data hiding [10].
This work presents a public key algorithm coupled with a chaotic neural network model to
embed online image data into LSBs image pixels with key, and de-embed with another key. This
algorithm improves efficiency and robustness that are not existed in many image steganography
algorithms. The rest of the paper network is organized as follows, in section 2 discusses background

and related work in the field of ANN based steganography. In section 3, the proposed model is
produced. In section 4, the performance and security analyses are described in detail. Finally section
5, concludes the whole work.
52
2. RELATED WORK

Steganography has many useful applications especially in the forefront of the current security
techniques by the gorgeous evolution in computational power, and the increase in security
consciousness.
Nameer N. proposed a new steganographic algorithm based neural network for hiding
information [11]. The model hides a large amount of information into color BMP image. The
adaptive image filtering and adaptive non-uniform image segmentation with bits replacement are
been used on the appropriate pixels, which are selected randomly using a new concept defined by
main cases with sub cases for each byte in one pixel. The proposed model has four high security
layers to make it difficult to break. The learning algorithm has been used on the fourth layer of
security through neural network to increase the difficulties of the statistical attacks. The algorithm
can embed efficiently a large amount of information that has been reached to 75% of the image size
(replace 18 bits for each pixel as a maximum) with high quality of the output.
Aboul Ella H. introduced an efficient approach for protecting the ownership by hiding the iris
data based on neural network [12]. The model hide the iris data into a digital image for
authentication purposes based on inspired spiking neural networks, called pulse coupled neural
network (PCNN) which first applied to increase the contrast of the human iris image and adjust the
intensity with the median filter. The PCNN segmentation algorithm is used to determine the
boundaries of the human iris image by locating the pupillary boundary and limbus boundary of the
human iris for further processing. The PCNN model is a two-dimensional neural network, each
neuron in the network corresponds to one pixel in an input image, receiving its corresponding pixel’s
color information as an external stimulus. The PCNN model is comprised of four parts that form the
basis of the neuron. The first part is the feeding receptive field that receives the feeding inputs, the
second part is the linking receptive field that receives the linking inputs from the neighbor neurons,
the third part is modulation field, which the linking input added a constant positive bias, then it is
multiplied by the feeding input, and the last part is a pulse generator that consists of an output pulse
generator and a threshold spike generator. Experimental results clearly indicate that the model
represent the authentication of the ownership very accurately.
Imran Khan proposed an original stenographic method based on neural network. The
proposed model hides the information message into cover image. The model divided the cover image
into blocks of non-overlapping pixels which generates a set of edged regions using pixel value
differencing (PVD) method. The total message calculated and store in the first pixel of the stego-image.
The model applied Xor back propagation neural network learning model for the embedding
process with the cover image and the key. The adjusted weights and the diagonal coefficient values
selected pixels act as the key. Experimental results show that the proposed method hides information
in edged regions and maintains a better visual display of stego image than the traditional methods
[13].
Arun Rana introduced a new high capacity image steganography method based on kohonen
neural network and wavelet contrast. The proposed method adopts the character that human eyes are
not sensible to the dark and texture block to embed the secret information into blocks. The kohonen
network is trained according to the absolute contrast sensitivity of pixels present in cover image
through classify the pixels in different classes of sensitivity which act as a secret key. The training of
the neural network determines the amount of information carried by individual pixels. Data
embedding is performed in less sensitive pixels by LSB substitution method, which replaces the least

significant bits of cover image with secret information that would be embedded. The Optimal Pixel
Adjustment Process (OPAP) implemented to obtain an optimal mapping function to reduce the
difference error between the cover and the stego-image, therefore improving the hiding capacity with
low distortions. Compared with DCT, the proposed method hides much more information and
maintains a better visual quality of stego-image. One the other hand, the secret key can be detect
[14].
53

Hieu Van Dang presented a new intelligent, robust and adaptive data hiding technique for
color images based on the combination of discrete wavelet transform (DWT), human visual system
(HVS) model and general regression neural network (GRNN). First, the RGB image is converted to
YCrCb image, and then the luminance component Y is decomposed by DWT. Wavelet coefficients
are then analyzed by a HVS model to select suitable coefficients for embedding the watermark. The
watermark bits are embedded into the selected coefficients by training a GRNN. At the decoder, the
trained GRNN is used to recover the watermark from the watermarked image. Experimental results
show that the proposed approach is robust in achieving imperceptibility in watermarking [15].
In 2013, RakeshGiri proposed a multilayer perceptron algorithm neural network for data
hiding. The model used the substitution method for embedding data without visibility. The proposed
method uses a multi-layered perceptron has a synaptic link structure with neurons between the layers
but no synaptic link among the layer itself. The updated weight of the neural network with back
propagation learning method is used for the embedding process. Results are observed through
different Media file as cover-objects and indicate that the system provides confidentiality and
integrity to the data during communication through open channels [16].
3. THE PROPOSED ALGORITHM
The chaotic system is rich in importance for it characterizes such as sensitivity to change
initial conditions and parameters, ergodicity, random behavior, and unstable periodic orbits with long
periods.
The Chebyshev chaotic map used as a public key method for embedding and de-embedding
with LSB insertion technique, and used with neural network to produce a combination of CNN,
based on a binary sequence generated from the Chebyshev chaotic maps, which the weight of
neurons are set in every iteration for generation the public key. The Chebyshev map possesses the
semi group property which lead to TsTr= TrTs. Here the Chebyshev polynomial map Tn: RR of
degree p is defined using the following recurrent relation: The Chebyshev map is defined in Equation
(1) [17]:

Where, n 2, T0(x) = 1, T1(x) = x.
The public key algorithm model is composed of the following three parts, which are Key
generation, Embed process, De-embed process, as follow:
3.1. Key Generation
This public key model uses two kinds of keys, a public key, and a private key. The keys are
decided by the de-embedor.
• Alice (receiver), in order to generate the keys, does the following steps:
Step 1: Generates a large integer S.
Step 2: Selects a random number X[−1,1] and computes Ts(X).
Step 3: Alice sets her public key to (X,Ts(X)) and her private key to S.
(1)

54

• Bob (Sender), in order to generate the keys, does the following steps:
Step 1: Bob (Sender), Obtain Alice’s authentic public key (X,Ts(X)).
Step 2: Generates a large integer r.
Step 3: Computes Tr(X),Tr·s(X) = Tr(Ts(X))
Step 4: Bob use the computed Tr(Ts(X)) for encrypt any image sent to Alice and his private key to r.
The whole process is shown in (Fig. 1).
Fig. 1: Generate Keys
3.2. Embed Process
In the embed process, we used the least significant bit (in other words, the 8th bit) of some or
all of the bytes inside an image is changed to a bit of the secret image. Digital images are mainly of
two types, 24 bit images and 8 bit images. In 24 bit images we can embed three bits of information in
each pixel, one in each LSB position of the three eight bit values. Increasing or decreasing the value
by changing the LSB does not change the appearance of the image; much so the resultant stego-image
looks almost same as the cover image. In 8 bit images, one bit of information can be hidden,
which will be done in our model based on the public key model generation through the following
steps:
Step 1: Reshape the keyTr(Ts(x)).
Step 2: Create a chaotic bit sequences bTrs(0), bTrs (1), bTrs (2) from the chaotic sequences Tr(1),
Tr(2), Tr(3), ….. Tr(m) through the generation scheme b(m), b(m), …… b(m-1), which is the binary
representation of Tr(m) for m=1,2,….M.
Step 3: Calculated the weight for WTrs as follows:
For i=0 to M, and J=0 to M

! #

!
# $
%

End
Step 4: Compute the secret image by merge each bit of the original secret image (O_img) with the
weight of WTrs as follows:
'(')!*' +!* ,
End
Step 5: Read the cover image.
Step 6: Determine the size and the length of cover image.
Step 7:Determine the public stego-key (ste_keyr)of n used in LSB embedding process based the
computed Tr(X),Tr·s(X) = Tr(Ts(X)) , which determine the number of LSB bits to be substituted in
the secrets image to be embedding into the cover image.
Step 8: Compute the stegnographic image (S_img) by insert the secret image in each first component
of next pixels of the cover image based on public stego-key, as follows:
!* )-.-/'01 )2) 3'(')01 )'*-4567 89
End
Step 9: Sends the stegnographic image S = (Tr(x),S_img) to the receiver Alice.
55
The embed process can be shown in (Fig. 2).

Fig. 2: The Embed Process
3.3. De-embed Process
In the de-embed process, the received image is then fed to the proposed de-embed model
algorithm based on Chaotic Neural Network weights for retrieve the original secret image.
Alice, to extract the original secret image (O_img) from the stegnographic image (S_img), does the
following:
Step 1: Uses her private key S to compute Ts·r = Ts(Tr(x)).
Step 2: Determine the public stego-key (ste_keys) of n used in LSB embedding process based the
computed Ts·r(X) = Ts(Tr(X)).
Step 3: Reshape the keyTs(Tr(x)).

Step 4: Create a chaotic bit sequences bTsr(0), bTsr (1), bTsr (2) from the chaotic sequences Ts(1),
Ts(2), Ts(3), ….. Ts(m) through the generation scheme b(m), b(m), …… b(m-1), which is the binary
representation of Ts(m) for m=1,2,….M.
Step 5: Calculated the weight for WTsr as follows:

56

! #
!
# $
%
End
Step 6: Extract the embed image (E_img) from the stegnographic image (S_img) based on public
stego-key (ste_keys), as follows:
:!* )
;; )2) 301 8 )'*-4569
End
Step 7: Recover the original secret image by uncombined each bit of the original image (O_img)
from the weight of WTsras follows:
-*
='('()!*' :!*
End
The de-embed process can be shown in (Fig.3).
Fig. 3: De-embed Process
4. EXPERIMENT AND TEST RESULTS
A series of experiments have been conducted for comparing stego image with cover results
requires a measure of image quality. This section discusses the efficiency of the proposed public key
technique such as capacity, Mean-Squared Error, Peak Signal-to-Noise Ratio, and Root-Mean-
Squared Error.

We demonstrate the performance of our public key model in steganography technique and
compare it with F5, J-Stego models, with two standard 512 * 512 grayscale images, which are Lena,
and Boat images using some measures:
57
4.1. Cover image and Stego Image
The cover image and the stego image for the two standard 512 * 512 grayscale images are
shown in (Fig. 3).
In the exposed (Fig. 4) both original and the stego-image is shown. Stego-image looks like
the original image which does not show any distortion. Thus the stego-image will not attract
attention towards itself. So it can be transferred to the recipient without displaying information
within itself.
Fig. 4: Cover Stego image

4.2. Mean Square Error (MSE)
The MSE represents the cumulative squared error between the stego-image and the cover
image, the lower the value of MSE, the lower the error. The block calculates the mean-squared error
using the following Equation (2) [18] [19]:
(2)
Where Xij: The intensity value of the pixel in the cover image, X ij: The intensity value of the pixel in
the stego image, M*N: Size of an Image. The MSE value is shown in (Table 1) and (Fig. 5).

TABLE 1: Comparison of MSE values for various Embedding Algorithms
Cover Image Public key model J-Stego [20] F5 [21]
Lena 0.1148 1.4438 1.3039
Boat 0.1152 1.3503 1.3294

Fig. 5: Comparison of MSE values for various Embedding Algorithms
As shown form (Table 1) and (Fig. 5), the best embedding algorithms value MSE value is
the public key model, which implies that the algorithms is generates a stego-image that will be
immune to statistical steganalysis. It has been analyzed that Lena as a cover image have given the
best least MSE value.
58
4.3. Peak signal to noise ratio (PSNR)
Its compute the peak error signal to noise ratio between the stego-image and the cover image.
The ratio is used as a quality of measurement between the cover and a stego-image. The block
computes the PSNR using the following in Equation (3) [18] [22]:
In this Equation L, is the maximum value of the cover image. The higher of the PSNR value,
the better the quality of the reconstructed image. The PNSR value is shown in (Table 2) and (Fig. 6).
TABLE 2: Comparison of PSNR (DB) values for various Embedding Algorithms
Cover Image Public key model J-Stego F5
Lena 63.59 33.36 36.94
Boat 63.48 35.67 36.23
(3)

Public key steganography using lsb method with chaotic neural network

More Related Content

What's hot (18)

Similar to Public key steganography using lsb method with chaotic neural network (20)

More from IAEME Publication (20)

Recently uploaded (20)

Public key steganography using lsb method with chaotic neural network