Available via license: CC BY 4.0
Content may be subject to copyright.
International Journal of Network Security & Its Applications (IJNSA) Vol.8, No.6, November 2016
DOI: 10.5121/ijnsa.2016.8501 1
STEGANALYSIS ALGORITHM FOR PNG IMAGES
BASED ON FUZZY LOGIC TECHNIQUE
Jawaher alqahtani, Daniyal Alghazzawi 1 and Li Cheng 2
1Department of Information Systems, King Adbulaziz University, Jeddah, Saudi Arabia
2Xinjiang Technical Institute of Physics & Chemistry, Chinese Academy of Sciences,
China
ABSTRACT
Embedding a message in media files, also known as steganography, is a common approach to hide secret
information. It has been exploited by some criminals to confidentially exchange messages. As a counter-
measure, tools have been developed in order to detect hidden information form digital media such as text,
image, audio or video files. However the efficiency and performance of previous approaches still have
room for improvement. In this research, we focus on algorithm design for better efficiency of hidden
message detection from PNG files. We employ three classic AI approaches including neural network, fuzzy
logic, and genetic algorithm and evaluate their efficiency and performance in controlled experiments.
Finally we introduce our message detection system for PNG files based on LSB approach and present its
usability in different case scenarios.
KEYWORDS
Steganography, Steganalysis, Artificial Intelligence, fuzzy logic.
1. INTRODUCTION
In recent years, the rapid advancements in technologies and communication have generated an
increase in the number of cyber-crimes cases. The biggest obstacle facing digital forensic
examiners is the analysis and identification of hidden data in digital media (images, audio and
video).
In digital forensics there are many indications which point to the use of hidden data in digital
media in the planning for criminal activities [1]. Hiding information refers to the process of
inserting and embedding information in digital content, such as, image, audio or video, without
drawing attention to the change. This process is called Steganography, while trying to detect the
hidden information is called steganalysis.
Every day there is a new steganography tool to hide data; however, steganalysis algorithms have
difficulty in detecting the hidden data because the majority of it is based on rule-based
techniques. In addition, the majority of steganography tools focus on hiding data inside images.
The researcher found that there is little research on detecting hidden data in the PNG format.
Therefore, there is a vulnerability to discover the hidden data on the PNG format.
Many crimes have been committed undetected by hiding and exchanging information between
criminals in a confidential manner. The process of hiding information in any carrier is called
Steganography. There are many algorithms that are used for this purpose and will be discussed in
International Journal of Network Security & Its Applications (IJNSA) Vol.8, No.6, November 2016
2
Section 1. In the opposite direction, there is a process known as steganalysis, which detects
hidden data and it will be discussed in Section 2. Steganography can hide any data in any media,
therefore, we will focus on the image as a carrier in Section 3. In Section 4, we will discuss AI
techniques that can be used to replace the traditional detection methods. Finally, the paper
introduces the proposed system for detecting the hidden data. In the end, we conclude that the
fuzzy logic system accomplished high performance in terms of classifying the clean and stego
images in PNG images.
2. OVERVIEW OF STEGANOGRAPHY
The term steganography consists of two greek words: stegano which mean hidden or cover and
grape which means writing [2]. Steganography was defined as "an ancient discipline which
usually refers to hiding information within information"[3]. The purpose of this concealment
process is to prevent a third party being aware of the existence of the information, thereby
protecting it from reading, change or destruction by the third party [4].
Steganography has been used for a long time. In 480 BC a Greek man named Demaratus sent a
message to the Spartans warning of possible invasion by Xerxes. Demaratus scraped the wax off
a wooden table, wrote on the wood what Xerxes wanted to do and then cover the message with
wax [5].
Many steganography techniques were used in World War II. For example, the United States
Marines used Navajo "codetalkers." In addition, invisible ink was used to write secret messages
while the paper appeared blank to the human eye. Extracting the secret message was done by
heating a liquid such as vinegar or milk [2]. Steganography has grown rapidly and attracted more
attention since ancient times. Various steganography techniques have been created and developed
to hide information in different carriers [5].
The basic stenography model contains the cover (also known as a carrier), password and message
(see Figure 1). The cover is a file used to hide the presence of information and the password or
stego key ensures that only the recipient will be able to extract the hidden information. The stego
key used to insert (embed) and extract a secret message on the cover. The message is the
embedded information. The result of inserting the message into the cover file is called a stego
file[6, 7].
2.1. Steganography Vs. Cryptography
Information security has many disciplines such as steganography and cryptography (Figure 2). In
some cases confusion arises between steganography and cryptography. Cryptography scrambles
the information so that an unauthorised person can be aware of its existence but is unable to read
Cover Object
Message
Stego Key
Processing
Stego Object
Figure 1. Steganography process
International Journal of Network Security & Its Applications (IJNSA) Vol.8, No.6, November 2016
3
or understand it [7]. Steganography secures the information, so that an unauthorised reader does
not know if there is a message or not [7].
Figure 2. Information security sub-sections
The main differences between cryptography and steganography are summarised in Table 1. The
process of attempting to discover the presence of hidden information, and read, change or delete
bit is known as Steganalysis.
Table 1. Cryptography vs. Steganography
Steganography
Cryptography
Definition
Depend on hiding the message
existence
Depend on hiding the message
meaning
Purpose
Keep communication secure.
Provide protection for data
Visibility
Never
Always
Failure
When discover the presence of a
hidden message
When able to decrypt and read
the message
Concern
Embedding capacity and
detectability of cover object
Robustness against
deciphering.
Carrier
Any type of digital media
Depend on text as a carrier
Key
Optional, but provide more
security
Necessary
2.2. Steganography Classifications and Techniques
International Journal of Network Security & Its Applications (IJNSA) Vol.8, No.6, November 2016
4
The classification of steganography presented in [8] is based on many standards, as shown in
Figure 3. The spatial domain, [9] indicates the classifications of spatial domain image
steganography techniques. These techniques are: least significant bit steganography, Pixel Value
Differencing steganography, RGB based steganography, Mapping based steganography, Speared
spectrum steganography, Palette based steganography, Code based steganography and Collage
based steganography. In the transform domain, Discrete Cosine Transform (DCT) and Discrete
Wavelet Transform (DWT) are the most common techniques which are used for JPEG images
[10].
Figure 3. Steganography techniques
2.2.1. Spatial Domain Steganography Techniques
There are many techniques used for image steganography. According to [11] [11] the common
effective techniques in image steganography are: Least Significant Bits substitution, Palette
Modification and Blocking.
LSB substitution refers to the process of modifying the least significant bit (lsb) of the carrier
image's pixel. Palette modification replaces the unused colours within the carrier image's palette
with colours that contain and represent the hidden information. Blocking breaks the image into
blocks and uses the Discrete Cosine Transform (DCT). The block is divided into 64 DCT
coefficients that approximate colour and luminance.
The least significant bits substitution is a method of inserting information in the selected cover
image. The LSB modification is not usually detectable by the human eye because this technique
causes minor changes in the colours [12]. The LSB is the simplest method of inserting secret
information into the image. It is easy to understand and apply [13] . In December 2011, WetStone
declared that about 70% of steganography tools are based on the LSB algorithm [14].
The LSB is the most effective technique for BMP and PNG images since their compression is
lossless. In addition, grayscale GIF images can use the LSB algorithm while this algorithm would
change the whole colour palette of a coloured GIF image at the beginning of the modification
process.
The LSB method is based on substitution which can be done up to 4 least significant bits. It
depends on replacing the embedding information at the least significant bit of each pixel in the
cover images. The result of embedding is a stego image. This replacement does not change the
image quality of human perception [15]. It is the most common method of steganography. It is
simple and it can provide a high capacity by allowing information to be inserted into all of the
International Journal of Network Security & Its Applications (IJNSA) Vol.8, No.6, November 2016
5
pixels of the cover image, however, humans may notice when information is embedded in a
smooth area of the image [16-18].
However, with the PVD approach, there is need to reference and compare the original image with
the stego image. The inclusion of information in the cover image is done by replacing the
difference values calculated from values of two consecutive pixels of the cover image with
similar values in which bits of inserted information are included [17].
The PVD focuses on concealing information in the edge area to increase the security without
considering the possibility of increasing the amount of hidden information [18]. The LSB method
can provide more capacity while the PVD method can provide more security. These two
techniques, PVD and LSB, can provide a high degree of security and capacity when it used
together [9].
A new technique was proposed based on the security of PVD and capacity of LSB [18] This
approach increased the amount of hidden information by inserting more in the pixels in the outer
area and increased the security by inserting less in the smooth area.
2.2.2. Transform Domain Steganography Techniques
Discrete Cosine Transform steganography (DCT) is a technique used to transform domain images
like JPEG. It can transform an image from the spatial to frequency domain. DCT separates the
JPEG image into parts according to their importance. The image is separated into low frequency,
middle frequency and high frequency components. The high frequency components are removed
through noise attacks and compression while the most important visual parts are in the low
frequency. The secret information is inserted by modifying the coefficients of the middle
frequency without affecting image visibility [10].
Discrete Wavelet Transform steganography (DWT) is also used for JPEG images. It was
introduced as highly flexible and efficient method of signals. A signal energy concentrates into
wavelet coefficients. Wavelets can convert an image to a series of wavelets which can be stored
in efficient way more than blocks of pixels [19].
3. STEGANALYSIS
Hiding information has been the objective of many researchers and scientists over the years as has
its discovery or steganalysis [20]. Steganalysis is the process of detecting information hidden
from a third party. Steganalysis aims to collect evidence about the existence of hidden messages.
Steganalysis is used in computer forensics and criminal activities on the internet [21].
According to [22], steganalysis algorithms of image can be divided into: specific algorithms and
blind (generic) algorithms. Blind (generic or universal) algorithms work independently of the
algorithms used in steganography and require a high degree of statistical analysis and complex
computations. They provide a good and improved result [20]. In the specific steganalysis, the
algorithm used depend on the image format and the algorithm used in steganography [20, 22].
The audio steganalysis algorithms take advantage of variations in the features of the audio signal
after embedding the secret message. The algorithm detects the gap phase, amplitude variations
and changes in the non-perceptual and perceptual audio quality metrics after embedding
messages. The video steganalysis algorithms which use temporal redundancies are more effective
than those that depend on spatial redundancies [22].
International Journal of Network Security & Its Applications (IJNSA) Vol.8, No.6, November 2016
6
[23] Presents a classification of steganalysis based on performance capacity. Steganalysis applied
to individual images is small scale steganalysis, whereas steganalysis performed on a large
number of images is large scale steganalysis. It is better to use small scale analysis when applying
it to social media where users can send many images daily. Universal steganalysis is used in large
scale steganalysis for more efficiency.
3.1. Steganalysis Techniques
Steganalysis has two main types of techniques: Statistical analysis and Visual analysis [24, 25].
Visual analysis depends on human eyes. In visual analysis, human abilities are used to detect the
presence of hidden information and secret communication. Visual inspection by eye can
distinguish and succeed in detecting hidden messages, especially when secret information is
embedded in the smooth area of an image [24, 25].
Statistical analysis is usually used to detect steganography. It can identify tiny modifications in
the statistical properties of the image caused by steganography [24, 25].
In addition, Steganalysis techniques can be classified as belonging to the spatial domain or
transform domain as in steganography [26]. A new steganalysis technique proposed by [1] uses
Colour Model Conversion to distinguish between the stego image and the original image. This
technique was tested on the stego-image generated by SIG tool (Stego-Image Generator) which
use the least significant bit algorithm.
In order to detect hidden data using LSB steganography, [6] presents the following methods of
steganalysis:
• Chi-squared which is the earliest method used to reveal LSB hidden in jpeg coefficients.
• Binary Similarity Measures.
• Regular/Singular (RS) scheme.
• Sample pair analysis.
The RS scheme and sample pair analysis are regarded as powerful LSB detection methods [6].
3.2. Steganalysis Tools
There are many steganalysis tools available to the public. Some of these tools designed to detect
certain steganography algorithms, consider this as a weakness of the steganalysis tool [27]. An
Experiment was done by [27] to compare the efficacy and response-time of StegalyzerAS and
StegalyzerSS software. The researchers use JPEG and PNG images as the cover file and create
the steganography by using open source applications like SecrtLayer, Openpuff, OpenStego, Steg,
steganographyStudio and GhostHost. The results of detection are represented in Table2
Table 2. StegalyzerAS & StegalyzerSS detection
Steganalysis tools
StegalyzerAS
StegalyzerSS
Steganography
tool
SecretLayer
Openpuff
OpenStego
Steg
SteganographyStudio
GhostHost
International Journal of Network Security & Its Applications (IJNSA) Vol.8, No.6, November 2016
7
We have done an experiment to measure steganalysis tool detection. Our experiment was done by
using three steganography tools (ImgStegano, Openpuff, OpenStego) to hide a secret message
into PNG images. To detect the secret message, we have been using five steganalysis tools. The
result shows that there is no tool can detect the hidden message embedding by all three
steganography tools (see Table 3).
Table 3. Result of testing steganalysis tools
Steganalysis tools
StegSpy
OpenStego
ImgStegano
StegoPNG
StegExpose
Steganography
Tool
ImgStegano
Openpuff
OpenStego
According to our experiment and [27], there is no tool can detect the hidden data in PNG image
embedded by any steganography tool.
4. IMAGE IN STEGANOGRAPHY AND STEGANALYSIS
Many digital mediums can be used to hide information, such as images, video and audio [15].
The image in the computer system is stored as an array of pixels, which can be coloured or
grayscale. An array of the coloured image is a combination of the three basic colours: Red, Green
and Blue. Each pixel in the coloured image is represented by 24 bits. In a grayscale image there
is only one channel and 8 bits used to represent each pixel.
The image can be manipulated in the transform domain (frequency domain) or spatial domain [9,
28]. The spatial domain directly manipulates the pixel values of an image whereas the transform
domain deals with the rate at which the values of the pixels are changing in the spatial domain.
The transform domain is more complex than the spatial domain [26, 29].
Images have a set of formats which differ in their structures and features. The most popular
formats are JPEG, GIF, BMP and PNG because of their broadband speeds, compatibility with
browsers and users' needs [15].
4.1. JPEG Images
JPEG stands for Joint Photographic Experts Group. It is the most common image format. It can
be manipulated under the transform domain. The common steganography tools that are available
to embed information into jpeg image are JSteg-Shell, JSteg, Outguess and JPhide [20]. A
steganalysis technique depends on the Discrete Cosine Transform (DCT) Proposed by [30] to
detect steganography in JPEG images. The F5 steganalysis algorithm is used to discover
information hidden in an image by an F5 steganography algorithm. After applying the F5
algorithm, the status of the image will change.
International Journal of Network Security & Its Applications (IJNSA) Vol.8, No.6, November 2016
8
Applying the F5 algorithm for steganalysis to compare the stego image and cover image uses the
DCT and analyses the histogram of both. The result of this comparison will show whether the
image is a stego image or not [30].
4.2.GIF Images
The Graphics Interchange Format (GIF) was introduced in 1987 by CompuServe. Due to its huge
support and portability, GIF is widely used on the World Wide Web [31]. Palette image
steganalysis is used for GIF images. The LSB technique changes the 24bit RGB value of a pixel
results [31].
4.3. BMP Images
BMP (known as bitmap image) is the native format of the windows platform. It acts as the parent
format of JPEG and GIF. The BMP format does not usually allow for the compression of images.
Raw image steganalysis is used for the BMP image and it has a lossless LSB plane. While
embedding over the lossless LSB plane results in casting over two grayscale values. Statistical
analysis over a BMP image shows the length of the hidden message[20].
4.4. PNG Images
PNG (Portable Network Graphics) was designed as an alternative to the GIF image format [12].
A PNG image is much larger than a GIF image and capable of sorting more transparency and
colour depth than GIF. It has more efficient compression techniques than GIF [20]. PNG supports
grayscale, true colour and indexed colours in addition to an optional alpha channel. It is robust,
provides simple detection of transmission errors and fulfils integrity checking.
The PNG structure always starts with a signature which contains eight bytes. The decimal values
of signature are: 137 80 78 71 13 10 26 10. A series of chunks follow the signature where
each chunk contains four parts: chunk type/name, length and cyclic redundancy code which each
contain four bytes. The fourth part is chuck data (length bytes).
According to the image format, [32] introduced a review of steganalysis techniques. To avoid the
loss of hidden data, JPEG use frequency domain steganalysis. The other formats (GIF, PNG, and
BMP) use spatial domain steganalysis.
The steganalysis techniques for grayscale images are easier than the techniques for coloured
images and can detect hidden information more accurately. Most steganalysis techniques are
applied to JPEG and BMP images and there is insufficient research on GIF and PNG images.
Hence, further research must be conducted on PNG steganalysis techniques. [15] concluded that
the PNG image is the most suitable format for LSB steganography when the focus is on the
amount of the information to be sent. The PNG image has the ability to hide a large amount of
information.
When we take a screenshot on the iPhone, android and windows systems, the image is
automatically saved as a PNG. PNG has many capabilities which make it superior to other image
formats (Table 4). PNG compression is fully lossless, and restoring or resaving an image will not
alter its quality unlike JPEG. PNG supports full transparency information unlike JPEG (for JPEG,
there is no transparency). It supports up to 16bit grayscale or 48bit truecolour and it is more
appropriate for images with few colours or images with a lot of sharp edges[33].
International Journal of Network Security & Its Applications (IJNSA) Vol.8, No.6, November 2016
9
Table 4. Comparison between Image formats
With LSB steganography, the PNG image is the most suitable format when the focus is on the
amount of information to be sent. The PNG image has the ability to hide a large amount of
information.
5. ARTIFICIAL INTELLIGENCE IN IMAGE STEGANALYSIS
A few studies have been conducted on identifying how more advanced techniques could be
applied and used to perform the intelligent processing of digital forensics. Automation approaches
have been adopted into digital forensic processes to increase the speed and decrease the time
required to identify a relevant evidence [34]. According to [35] it is necessary to enhance the use
of the resources available. Intelligent techniques and tools could enhance the investigation in
terms of efficiency and time when applied to digital investigations.
Most efforts focus on partially automating the processes of digital forensics to save time and
resources. According to [36], computational intelligence has a strong chance of being applied
successfully to a steganalytic system or steganasystem. Three major methods of computational
intelligence have been identified to be useful in steganalysis: Bayesian, neural network, and
genetic algorithm. These three techniques are applied in image steganalysis, audio steganalysis,
and video steganalysis. The result of [36] and other studies discussed in this paper, shows that the
application of computational intelligence methods has had a high effect on the performance of
steganalysis.
The Automated Forensic Examiner is an approach proposed by [34] which aims to solve the
problem of sorting and identifying relevant artefacts by applying artificial intelligence. A number
of techniques used by AFE such as visualization and competency measure to provide and present
an investigator with deeply understanding of cases.
The agent system will increase the speed and effectiveness and provide better results. [37]
suggested a steganalysis technique based on statistics to discover the hidden information and used
a Multi Agent System. MAS is capable of using all set agents to determine if the image contained
hidden information or not, where each agent has a specific task to do. The main techniques of
Artificial intelligence are: Neural Network, Fuzzy Logic and Genetic Algorithm.
5.1. Neural Network
Specific discovering provides highly accurate and more reliable results than general (universal)
discovering. Approaches based the Neural Network are studied in [38] to measure its
effectiveness in detecting hidden information. The most popular approaches are: visual detection,
which is based on first order statistics, dual statistics methods that use spatial correlations in
images and higher order statistics which is a general blind discovering.
JPEG
TIFF
GIF
PNG
Transparency
Palette image
Truecolor
Best compression
Web supporting
Animation
International Journal of Network Security & Its Applications (IJNSA) Vol.8, No.6, November 2016
10
Embed a data
Produce stego image
The high degree of redundancy present in the image is used for performing steganography. [39]
suggested a method based on the neural network and a random selection of pixels. This technique
shows how a selection of different portions of the images increased the chances of hiding
information in a more effective way. This technique reduces the chances of information detection.
A detection system using an Artificial Neural Network (ANN) presented by [40] to classify the
stego images and clean images. This system provide result with high accuracy rate because ANN
is a blind classifier. The results are very good and have a lower misclassification rate than other
ANNs.
5.2. Fuzzy logic
A multi-agent system for image steganalysis presented by [11]. This system is based on the
paradigm of the community of polygenic bees using fuzzy logic. The system is designed to detect
a file contain hidden data. The result presents accuracy rate of 89.37%, with 10.54% for false
negative and 10.63% for false positive.
Also, [11] test a decision tree using entire dataset without polygenic multi agent system fuzzy
clustering steganalysis. The result show accuracy rate of 72.45% with 26.92% for false negative
and 27.23% for false positive. An accuracy rate of 56% for JPHide and Seek with Adaboost and
0.8% with OAASVM presented by[41].
5.3. Genetic Algorithm
The steganalysis process can be improved using genetic algorithms. The new approach for
universal steganalysis proposed in [42] is based on the genetic algorithm (GA) and higher order
statistics. This approach increases the accuracy of universal detection.
6. THE PROPOSED SYSTEM
Two basic stages represent the proposed system building process. The first stage focus on create
a collection of stego images when texts and images are embedded into clean images as hidden
data (See Figure 4) . OpenStego and Steganography Studio are the selected steganography tools,
which support LSB (Least Significant Bit). The hiding ratio is between 11% and 20%. More
than 600 images of people, animals, food, landscapes, buildings and interiors are used. The
images file formats used PNG. Figure 5 and 6 show examples of the clean images and the hidden
data.
Figure 4. The process of creating stego images
International Journal of Network Security & Its Applications (IJNSA) Vol.8, No.6, November 2016
11
Figure 5. Example of the clean images
Figure 6. Example of hidden data
In the second stage , the focus on evaluating images set (stego and clean images) to detect hidden
data by extracting a number of features from images and using fuzzy logic and neural network as
AI techniques. Three cases are created based on the number of features as shown in Table 5.
Table 5. Three cases with different number of features
Case
No. of Features
No. of Images
Case 1
5 features
600 images
Case 2
30 features
600 images
Case 3
50 features
600 images
The proposed system for detection is depend on extracting a set of image features from stego and
clean images and using fuzzy logic and neural network techniques to distinguish between clean
an stego images. To train the proposed system to detect a hidden data in the images, it was set up
a co-occurrence matrix for each image. Usefulness of this matrix is to provide valuable
information about the neighboring pixels position in an image. From the created matrixes a
number of features will be extracted. Different features will be selected for the three colour
channels (Red, Green and Blue). These image features include correlation, contrast, colours
variance, entropy and energy. Table 6 shows the five extracted features from clean image1 and its
stego copy named stego image1.
The following are the steps of features extraction from images by using Matlab software :
Using imread() function to read the image.
Using graycomatrix() function to create co-occurrence matrix for the image.
Using graycoprops() function to extract the correlation, homogeneity and entropy features
from the created matrix.
International Journal of Network Security & Its Applications (IJNSA) Vol.8, No.6, November 2016
12
Table 6. Values of the five features extracted from clean image1 and stego image1
Features name
Clean image 1
Stego image 1
Contrast
0.44698
0.44608
Correlation
0.91807
0.91823
Energy
0.11202
0.11208
Homogeneity
0.87812
0.87828
Entropy
7.56398
7.56370
For validation, the Fuzzy Logic (FL) is used to validate the results and the accuracies achieved by
the system during the training and the testing phases. Takagi–Sugeno type of fuzzy inference
method is used for this FL. The system divided the images into two sets. It used around 70% of
the images for training and 30% for testing. The function xlsread() used to import the extracted
features from Excel file and the function genfis2() is used to generate a fuzzy interface system.
The Multilayer Perceptron Neural Network (MLP) is also used to validate the results. The MLP
were run 6 times for accuracy; the test automatically divided the images into two sets. It used
around 70% of the images for training and 30% for testing. However, the MLP chose different
images with each run for the training and the testing. Therefore, the results are not the same for
each test.
7. DISCUSSION
As it known about fuzzy logic that it is more accurate methods of artificial intelligence
techniques, successive tests have proved that the detection ratio of the clean images and the stego
images increased respectively to the increased in the number of images and features.
We conclude that the fuzzy logic system achieved high overall accuracy during the testing phase
for each case comparing to Multilayer Perceptron Neural Network. It is noticeable that in each
test of MLP, the accuracy rate of detecting the stego images and the clean images was increased
slowly relatively in each case than previous one, despite the increase in the number of features in
each case as shown in figure 7. This reflects that the fuzzy logic system accomplished high
performance in terms of classifying the clean and stego images.
Figure 7. Accuracy rate of validation
8. FUTURE WORK
International Journal of Network Security & Its Applications (IJNSA) Vol.8, No.6, November 2016
13
As it has been previously mentioned in this paper, genetic algorithm will be used as a third
technique of AI to test the proposed system of detection. Genetic algorithm will be implemented
as a classifier to distinguish between stego and clean images. In addition, the accuracy of the
genetic algorithm test will be compared to the results of fuzzy logic and neural network.
9. CONCLUSIONS
This paper presented a briefly review for the steganography and steganalysis definitions,
categories, techniques and some related work. We concluded that the steganography field is
developing rapidly and new methods and techniques are evolving to detect what is concealed by
steganography techniques. One steganography technique is LSB steganography. The PNG image
is the most suitable format for the LSB technique when the focus is on the amount of information
to be sent. The PNG image has many capabilities which make it superior to other image formats
and it requires further research.
There have been several attempts to incorporate artificial intelligence techniques in the
steganalysis process. This combination is used to increase the accuracy and speed of detection.
The paper introduced the proposed detection system. OpenStego and Steganography Studio are
the selected steganography tools to create the stego images. the proposed system depend on
creating a co-occurrence matrix for images to extract set of image features. Fuzzy Logic (FL) and
The Multilayer Perceptron Neural Network (MLP) used to validate these features. The fuzzy
logic system achieved high overall accuracy.
ACKNOWLEDGEMENTS
I wish to express my sincere thanks to the supervisor Dr. Daniyal Alghazzawi, for supporting me
and providing me with guidance and all the necessary facilities for the completion of this work.
REFERENCES
[1] P. Thiyagarajan, G. Aghila, and V. P. Venkatesan, "Steganalysis Using Color Model Conversion,"
arXiv preprint arXiv:1206.2914, 2012.
[2] R. Poornima and R. Iswarya, "An Overview of digital image Steganography," International Journal of
Computer Science & Engineering Survey (IJCSES), vol. 4, pp. 23-31, 2013.
[3] S. Engle, "Current State of Steganography: Uses, Limits, & Implications," Retrieved March, vol. 12,
p. 2012, 2003.
[4] S. A. Laskar and K. Hemachandran, "A Review on Image Steganalysis techniques for Attacking
Steganography," in International Journal of Engineering Research and Technology, 2014.
[5] C. James, "Steganography Past, Present, Future,"
URL:http://www.sans.org/reaing_room/whitepapers/steganography/steganogaphy_past_present_futur
e _552. pdf ( 12.12. 2009).
[6] B. Boehm, "StegExpose-A Tool for Detecting LSB Steganography," arXiv preprint arXiv:1410.6656,
2014.
[7] M. M. Amin, M. Salleh, S. Ibrahim, M. R. Katmin, and M. Shamsuddin, "Information hiding using
steganography," in Telecommunication Technology, 2003. NCTT 2003 Proceedings. 4th National
Conference on, 2003, pp. 21-25.
[8] R. Amirtharajan, V. Rajesh, P. Archana, and J. Rayappan, "Pixel indicates, standard deviates: A way
for random image steganography," Res. J. Inform. Technol, vol. 5, pp. 383-392, 2013.
[9] G. Swain and S. K. Lenka, "Classification image steganography techniques in spatial domain: A
study," Int J Comput Sci Eng Tech, vol. 5, pp. 219-32, 2014.
[10] S. Goel, A. Rana, and M. Kaur, "A Review of Comparison Techniques of Image Steganography,"
Global Journal of Computer Science and Technology, vol. 13, 2013.
International Journal of Network Security & Its Applications (IJNSA) Vol.8, No.6, November 2016
14
[11] S. Azevedo, L. Gonçalves, and R. Rudson, Fuzzy Logic on a Polygenic Multi-Agent System for
Steganalysis of Digital Images: INTECH Open Access Publisher, 2012.
[12] W. W. Zin, "Message Embedding In PNG File Using LSB Steganographic Technique," International
Journal of Science and Research (IJSR) Volume, vol. 2, 2013.
[13] T. Qiao, F. Retraint, R. Cogranne, and C. Zitzmann, "Steganalysis of JSteg algorithm using
hypothesis testing theory," EURASIP Journal on Information Security, vol. 2015, pp. 1-16, 2015.
[14] J. Fridrich and J. Kodovský, "Steganalysis of LSB replacement using parity-aware features," in
Information Hiding, 2013, pp. 31-45.
[15] V. L. Reddy, A. Subramanyam, and P. C. Reddy, "Implementation of LSB steganography and its
evaluation for various file formats," Int. J. Advanced Networking and Applications, vol. 2, pp. 868-
872, 2011.
[16] H.-W. Tseng and H.-S. Leng, "A Steganographic Method Based on Pixel-Value Differencing and the
Perfect Square Number," Journal of Applied Mathematics, vol. 2013, 2013.
[17] D.-C. Wu and W.-H. Tsai, "A steganographic method for images by pixel-value differencing," Pattern
Recognition Letters, vol. 24, pp. 1613-1626, 2003.
[18] M. H. Mohamed, N. M. Al-Aidroos, and M. A. Bamatraf, "Innovative Multi-Level Secure
Steganographic Scheme based on Pixel Value Difference."
[19] D. Gupta and S. Choubey, "Discrete Wavelet Transform for Image Processing," International Journal
of Emerging Technology and Advanced Engineering, vol. 4, pp. 598-602, 2015.
[20] R. an Amirtharajan and J. Rayappan, "Steganography—time to time: A review," Research Journal of
Information Technology, vol. 5, pp. 58-66, 2013.
[21] A. Ibrahim, "Steganalysis in computer forensics," in Australian Digital Forensics Conference, 2007,
p. 10.
[22] N. Meghanathan and L. Nayak, "Steganalysis algorithms for detecting the hidden information in
image, audio and video cover media," international journal of Network Security & Its application
(IJNSA), vol. 2, pp. 43-55, 2010.
[23] R. P. S. Sruthi Das N, "A Survey on Different Image Steganalysis Techniques," International Journal
of Modern Trends in Engineering and Research vol. 02, p. 5, 2015.
[24] K. Curran and J. Mc Devitt, "Image analysis for online dynamic steganography detection," Computer
and Information Science, vol. 1, p. p32, 2008.
[25] H. Wang and S. Wang, "Cyber warfare: steganography vs. steganalysis," Communications of the
ACM, vol. 47, pp. 76-82, 2004.
[26] S. O. Mundhada and V. Shandilya, "Spatial and Transformation Domain Techniques for Image
Enhancement," International Journal of Engineering Science and Innovative Technology (IJESIT),
vol. 1, pp. 213-216, 2012.
[27] J. Green, I. Levstein, C. R. J. Boggs, and T. Fenger, "Steganography Analysis: Efficacy and
Response-Time of Current Steganalysis Software," J Comput Sci, vol. 9, pp. 236-44, 2015.
[28] R. Kumar, K. K. Saini, and S. Chand, "A New Steganography Technique Using Snake Scan Ordering
Strategy," International Journal of Image, Graphics and Signal Processing (IJIGSP), vol. 5, p. 25,
2013.
[29] R. Singh and G. Chawla, "A Review on Image Steganography," International Journal of Advanced
Research in Computer Science and Software Engineering, vol. 4, pp. 686-689, 2014.
[30] S. Mehta, A. Maru, and P. K. Goel, "A Survey Paper on Steganalysis F5 Algorithm."
[31] A. Brown, "Graphics File Formats," ed, 2008.
[32] S. M. Badr, G. Ismaial, and A. H. Khalil, "A Review on Steganalysis Techniques: From Image
Format Point of View," International Journal of Computer Applications, vol. 102, 2014.
[33] T. Boutell, "PNG (Portable Network Graphics) Specification Version 1.0," 1997.
[34] M. Al Fahdi, N. Clarke, and S. Furnell, "Towards An Automated Forensic Examiner (AFE) Based
Upon Criminal Profiling & Artificial Intelligence," 2013.
[35] A. Irons and H. S. Lallie, "Digital Forensics to Intelligent Forensics," Future Internet, vol. 6, pp. 584-
596, 2014.
[36] R. Din and A. Samsudin, "Computational intelligence in steganalysis environment," 2008.
[37] A. I. Bouguerne, B. H. F. Merouani, and C. N. Kobsi, "Multi Resolution Decomposition For A
Passive Steganalysis Based On a Multi Agent System."
[38] N. Kobsi and H. F. Merouani, "Neural Network Based Image Steganalysis: A Comparative Study," in
Neural Networks for Signal Processing [1994] IV. Proceedings of the 1994 IEEE Workshop, 2007,
pp. 423-430.
International Journal of Network Security & Its Applications (IJNSA) Vol.8, No.6, November 2016
15
[39] I. Khan, "An Efficient Neural Network based Algorithm of Steganography for image," International
Journal of Computer Technology and Electronics Engineering (IJCTEE) Volume, vol. 1, pp. 63-67.
[40] C. Bergman and J. Davidson, "An artificial neural network for wavelet steganalysis," Final Report to
Midwest Forensics Resource Center, 2005.
[41] S. Bakhshandeh, J. R. Jamjah, and B. Z. Azami, "Blind Image Steganalysis based on local
information and human visual system," in Signal Processing, Image Processing and Pattern
Recognition, ed: Springer, 2009, pp. 201-208.
[42] X. Y. Yu and A. Wang, "An investigation of genetic algorithm on steganalysis techniques," in
Intelligent Information Hiding and Multimedia Signal Processing, 2009. IIH-MSP'09. Fifth
International Conference on, 2009, pp. 1118-1121.
