Steganography by multipoint Arabic letters

Abstract

Security methodologies are taken into consideration for many applications, where transferring sensitive data over network must be protected from any intermediate attacker. Privacy of data can be granted using encryption, by changing transmitted data into cipher form. Apart from encryption, hiding data represents another technique to transfer data without being noticeable by an attacker. This technique is called Steganography. In this paper, we will discuss the main concepts of Steganography and a carrier media that is used for this goal. Employing text as mask for other text represents the most difficult method that can be used to hide data. We will discuss some algorithms that use Arabic text. We then describe our doted space methodology to enhance data hiding.

Full text

Steganography by Multipoint Arabic Letters
Ammar Odeh, Aladdin Alzubi, Qassim Bani Hani, Khaled Elleithy
Department of Computer Science & Engineering,
University of Bridgeport
Bridgeport, CT 06604, USA
aodeh @bridgeport.edu
Abstract-Security methodologies are taken into consideration for
many applications, where transferring sensitive data over
network must be protected from any intermediate attacker.
Privacy of data can be granted using encryption, by changing
transmitted data into cipher form. Apart from encryption, hiding
data represents another technique to transfer data without being
noticeable by an attacker. This technique is called
Steganography. In this paper, we will discuss the main concepts
of Steganography and a carrier media that is used for this goal.
Employing text as mask for other text represents the most
difficult method that can be used to hide data. We will discuss
some algorithms that use Arabic text. We then describe our doted
space methodology to enhance data hiding.
Keywords: Steganography, carrier file, text steganograph, image
steganography, audio steganography, information hiding,
Persian/Arabic Text, steganalysis, stego medium, stego_key.
I. INTRODUCTION
A. Background
Steganography is a Greek word coming from cover text.
"Stegano" means hidden and “Graptos" means writing. In
steganography, the secure data will be embedded into another
object, so middle attacker can't catch it [1]. Invisible ink is an
example for Steganography using a readable message transfer
between source and destination. Everyone in the middle can
read the message without having any clue about the hidden
data. On other hand, authorized persons can read it depending
on the substances features [2][3].
Ancient Greeks used to shave the messenger head and then
wait until the hair grew back. That is when the message will be
sent to the destination [1]. Depending on this method, there are
two possibilities:
1. Message has arrived so the receiver can read the message
and recognize if message has changed or not.
2. If message did not arrive, it means the attacker has
detected the message.
B. Motivation
Steganography algorithms depend on three techniques to
embed the hidden data in the carrier files.
1. Substitution: Exchange a small part of the carrier file by
the hidden message where the middle attacker cannot observe
the changes on the carrier file. On the other hand, in choosing
a replacement process, it is very important to avoid any
suspicion. This means that it is important to select
insignificant parts from the carrier file and then replace them.
For instance, if the carrier file is an image (RGB), then the
least significant bit (LSB) can be used as the exchange bit [4].
2. Injection: By adding hidden data into the carrier file, the
file size will increase and this will increase the suspicion. So
the main goal to present techniques to add hidden data while
avoiding attacker suspicion [4].
3. Propagation: There is no need for a cover object. It
depends on using a generation engine fed by input (hidden
data) to produce and mimic a file (graphic or music or text
document ).
The Steganography process consists of three main
components as show in Figure 1.
Figure 1. General components of Steganography
Different types of cover media including image, sound,
video and text can be used in Steganograph, as shown in
Figure 2. Choosing carrier file is very sensitive where it plays
a key role to protect the embedded message. Successful
Steganography depends on avoiding suspicion. Steganalysis
will start checking the file. If there is any suspicion, this will
compromise the main goal of Steganography [3][4].
Figure 2. Stego Media
Text Steganography represents the most difficult type,
where there is generally lack of data redundancy in the text file
in comparison with other carrier files [5]. The existence of

such redundancy can help increase the capacity of hidden data
size. Furthermore, text Steganography depends on the
language, as each language has its own unique characteristics
which is completely different from other languages. For
example, the letter shape in English language does not depend
on its position in the word, while Persian/Arabic letters have
different forms depending on letter positioning [6].
In our new proposed algorithm, we hide text inside text by
employing Arabic language and applying a random algorithm
to distribute the hidden bits inside the message. The main
reasons for choosing the Arabic language are:
1. The proposed algorithm will depends on multi dotted
points letters. Therefore, the algorithm must employ a
language that has as many as possible dotted letters. For
example, the Arabic Language has 5 multipoint letters and
Persian/Farsi language has 8 letters [7], while English does not
have any.
2. Wealth availability of electronic textual information.
3. There is little research on other languages compared to
English.
4. The approach can be extended to other languages like Urdu
and Kurdish.
C. Main Contribution and Paper Organization
An efficient algorithm is presented in this paper. The main
idea is to use multiple point characters in Arabic which
enables us to hide more than two bits per letter. The rationale
behind this approach is that most of the previous algorithms
reported in literature hide one bit for one letter. Furthermore,
we will merge our algorithm with vertical shifting point
algorithm to increase the size of the hidden file. The size of the
carrier will be constant without any change. After we add the
data, we convert the file into image to avoid the retyping
problem.
The rest of this paper is organized as follows. In section II,
we discuss some text Steganography techniques as well as
their advantages and disadvantages. Our multipoint hidden
algorithm is discussed in section III. In section IV, we present
experimental results of our algorithm. Finally, conclusions are
presented in section V.
II. PRIOR WORK
Text Steganography is divided into two categories. The
first one is the semantic method, and the second is the
formatting method, as shown in Figure 2. In this Section we
will briefly explain some Steganography examples. In Table I,
we present a simple comparison between semantic and
formatting methods.
Table I. Comparison between text Steganography methods
Semantic Method Format Method
Amount of hidden
data
Small amount More than semantic
Flaws Sentence meaning notice from OCR or
retyping
Steganography criteria will depend on the amount of data
that can be hidden and the main problem facing the method.
We describe ten algorithms that hide data inside text
documents. The last two algorithms deal with Arabic and
Persian languages.
1. Word Synonym
Word Synonym is also called semantic method and it
depends on replacing some words by their synonym. See
Table II. This technique will convey data without making any
suspicion. It is limited in terms of that fact that hidden data
will be small relative to other methods. Moreover, it may
change the sentence meaning [7][10][12].
2. Punctuation
This method uses punctuation like (.)(;) to represent hidden
text. For example "NY, CT, and NJ" is similar to "NY, CT and
NJ" where the comma before the “and” represents 1, and the
other represents 0. The amount of hidden data in this method
is very small compared to the amount of cover media.
Inconsistence use of punctuation will be noticeable from
Steganoanalysis point of view [9].
Table II. Using Word Synonym
Word Synonym
Big Large
Find Observe
Familiar Popular
Dissertation Thesis
Chilly Cool
3. Line Shifting
Line shifting means to vertically shift the line a little bit to
hide information to create a unique shape of the text.
Unfortunately, line shifting can be detected by a character
recognition program. Moreover retyping removes all hidden
data [7][10].
In Figure 3, we present an example regarding line shifting
where the vertical shifting is very small (1/300 inch). This is
not noticeable by the human eye.
Figure 3. Line shifting; second line is shifted up 1/300 inch
[10].
4. Word Shifting
In this method, changing spaces between words enables us
to hide information. Word shifting is noticeable by OCR
through detecting space sequence between words [7][10].
5. SMS Abbreviations
Recently most SMS messages use abbreviations for
simplicity and security while used in different applications
such as internet chatting, email, and mobile messaging. The
main advantage of this method is to speed typing, reducing the

message’s length and manipulated keyboard limitation
character [13].
Other algorithms use numbers to convey specific
information. As mentioned above, SMS abbreviation can be
used in specific applications while using in others creates
suspicion of any entity that monitors the ongoing transmission.
Table III. Some SMS Abbreviations
Abbreviation Meaning
ADR Address
ABT About
URW You are welcome
ILY I love you
EOL End of lecture
AYS Are you serious?
6. Text Abbreviations
Text abbreviation is similar to SMS abbreviation, where a
dictionary is created for each word abbreviation and its
meaning. The dictionary is published between the
communication parties. Abbreviation represents one method to
hide data. For example if you send (see) it means (do you
understand) [13].
7. HTML Spam Text
This method depends on HTML pages, where their tags
and their members are insensitive. For example <BR> equal to
<Br>, and the same as <br> and <bR>. The hidden data
depends on upper case or lower case letters to embedd 0 or 1.
8. TeX Ligatures
In TeX ligatures, some special groups of letters can be
joined together to create a single glyph as shown in Figure 4.
The algorithm finds available ligatures in the text to hide a
single bit in each one. For example, if we want to hide 1 we
write fi to f {}i which creates some space between f and i.
Otherwise, we encode 0 [5].
The same algorithm can be applied to Arabic character "ﻻ"
or " ل". This algorithm has two problems. The first problem is
that file size increases when we apply extension in our text.
The second problem is that if the ORC notices the font change,
it can detect the decoding hidden message [6][5].
9. Arabic Diacritics
Arabic language uses different marks. The main reason to
use these symbols is to distinguish between words that have
same letters. It depends on Arabic Diacritics (Harakat), where
diacritics are optional. Most of Arabic novels can be read
without Diacritics which depends on the language’s grammar.
The most occurrence is Fatha " َ " which will be used to
encode 1 otherwise encode 0.Our new algorithm will enhance
the reuse of cover media. Furthermore, the carrier file size
might be reduced depending on the hidden message. On the
other hand, when ORC detects the same message with
different diacritics, it might conclude that there is a hidden
data. In addition, retyping will remove the embedded message
[8].
Figure 4 .Join between characters [5]
Table IV. Some Letters with mark and their Pronunciation
Pronunciation Letter with
Haraka
Haraka
Do ُد Dama
De ِد Kasra
Da َد Fatha
10. Vertical Displacement of the Points
This algorithm achieves excellent performance as it is
applied on pointed (dotted) letters. Other languages such as
English language have only two dotted letters; {i, j}; and thus
limits the application of this algorithm. Alternatively, some
languages such as Arabic and Persian have many pointed
letters which make them fit better for this technique.
Arabic and Persian languages have many pointed
characters. Arabic has 26 letters where 13 of them are pointed,
and Persian has 32 letters where 22 of them are pointed. In this
new algorithm, we encode 1 to shift up the point, otherwise
encode 0. This method can encode a huge number of bits, and
need a strong OCR to recognize the changes. Meanwhile,
retyping will remove the entire message [7].
Figure5. Vertical shifting point [7]

III.PROPOSED ALGORITHM
Pointed letters represents one of the important
characteristics of Arabic and Persian languages. Table II
classifies Arabic letter with respect to the number of points.
Table V Arabic letters with respect to the number of points
Letter Number of points
ا,ح,د,ر,ص,ط,ع,ك,ل,ﻩ,و 0
ب,ج,خ,ذ,ز,ض,ظ,غ,ف,ن 1
ت,ق,ي 2
ث,ش 3
Our algorithm hides data in multipoint Arabic/Persian
letters like (ث, tha). In Arabic language, there are five multi-
pointed letters, and in Persian there are eight. Each character
can be used to hide 2 bits to determine the shifting and
distance between letter points. Table III represents the relation
between letter, shifting, and encoding
Table VI represents relation of shifting and distance to
letter format
Point
shifting
distance code Letter effects
0 0 00 No change
0 1 01 Only distance between
point increase
1 0 10 Only little upper
shifting
1 1 11 Upper shifting and
increase distance
between point
A. Pseudo Code and Flow Chart
In this subsection we present the pseudo-code and
flowchart of the proposed Arabic multipoint steganography
algorithm. The flowchart of the algorithm is shown in Figure
7. The pseudo-code follows:
1. Enter the text and hidden file and its size
2. Search for multipoint letters
3. Hide size of embedded data at the beginning
4. For I= start to EOF
IF hidden data ="00" then call Nochange();
Else if hidden data= "01" then call distance ();
Else if hiddendata ="10" then call shifting ();
Else if hiddendata="11" then call
distance_shifting();
Else random call for any one // for padding purpose
End for
5. Convert file to image file and send to other side
6. End
As can be seen in the above pseudo-code, in step 5, the file
must be converting to image file for transfer. The receiver will
scan the image file and find out the multipoint letters and then
classify the function applied on it.
B. Advantages and Disadvantages
Our multipoint algorithm has many advantages, as one
character can hide 2 bits compared to other algorithms that can
hide 1 bit per letter. This implies that the amount of hidden
data can be duplicated. Furthermore, the number of changed
characters for a given message, which leads to decrease which
leads to less suspicion. Moreover, the file size will have fewer
changes, since the number of changes in characters format is
less.
On the other hand, any retyping process removes all the
hidden data, as the hidden data depends on the file format. The
consistent format used in the system might raise the level of
suspicion of an attacker.
IV. EXPERIMENTAL RESULTS
A. Multipoint algorithm
Our algorithm depends on multipoint letters to include
hidden data. For this reason, we test different websites that
contain text and picture.
We will run two files in parallel for carrier file and hidden
data file. This process we will use two bits for each letter by
applying the distance_shifting algorithm.
Figure 6. Parallel operation carrier file and hidden data
As shown in Figure 6, after we merge the hidden data, we
convert it to image file to prevent the issues caused by
retyping. After that, we can compress the file to produce the
compressed image. The receiver decompresses the image and
then extracts the hidden data.
B. Experimental Results of the Multipoint Algorithm
Table VII Capacity of webpage for different Arabic website
# Page Name
Page
Size
Character
# 2 point
or more
Capacity
Ratio
(Bit/
Kilobyte)
1 aljazeera.net
23.8
KB
1245
105
2 daralhayat.com
15.4
KB
968
126
3 salahws.com
10.3
KB
535
104
4 holyquran.net/tadabur
13.8
KB
516
75
5 khayma.com
21.8
499
46

KB
HD: Hidden Data
EOF: End of File
Figure 7. Flowchart of multipoint algorithm
As shown in Table VII, we used different Arabic news
websites to observe the data ratio that can be hidden in each
website.
C. Analysis of the Algorithm
Our algorithm is compared to the vertical shifting
algorithm in terms of the number of changing letters and the
number of bits that can be hidden. In Table VIII, the result of
testing one paragraph is shown. The total number of letters is
115 which indicates that 42% of it is pointed. On the other
hand, the number of multipoint letters is 29, which indicates
25%. The data that can be hidden in the multiple point
algorithm is more than the first one. So we calculate the
efficiency depending on Equation 1:
E= #of hidden characters / #of characters *100% (1)
where E is the efficiency.
So, the efficiency of multiple points is 50% while vertical
point shifting is 43%.
Table VIII Vertical Point shifting versus Multipoint
algorithm
فﺮﻌﺘﻟا ﻦﻜﻤﻳ ﻪﻧأ ﻦﻴﻴﻧﺎﻄﻳﺮﺒﻟا ﻦﻴﺜﺣﺎﺒﻟا ﻦﻣ ﺔﻋﻮﻤﺠﻣ ﺎهاﺮﺟأ ﺔﺜﻳﺪﺣ ﺔﺳارد ﺖﺘﺒﺛأ
ﻪﺗرﺎﻴﺳ عﻮﻧ ﻲﻓ ﻖﻴﻗﺪﺘﻟا لﻼﺧ ﻦﻣ صﺎﺨﺷﻷا ﺾﻌﺑ ﺔﻴﺼﺨﺷ ﺢﻴﺗﺎﻔﻣ ﻰﻠﻋ
ﺔﺻﺎﺨﻟا[13]
Number of letter Number of
hidden bit
Pointed letter 50 50
Multipoint letter 29 58
D. Merge with Vertical Point Shifting
In this subsection, we calculate the ratio when we
compound the vertical point shifting algorithm discussed in [7]
with the multipoint algorithm.
Figure 8. Merging vertical and multipoint algorithms
In Figure 8, we note the ability of the vertical point
algorithm to hide huge amount of data as compared to the
multipoint algorithm. On the other hand, an observer can
notice the vertical changes more than multipoint changes
where the number of changes will be more. Consequently, the

merging both algorithms gives us better idea and flexibility in
hiding data and those that can be detected by an observer.
V. CONCLUSION
In this paper we introduce a new text Steganograph for
Arabic multipoint letters. The new algorithm deals with two
bits for each multipoint letter. We combine our strategy with
vertical point shifting [7] to improve the amount of hidden
data.
The retyping process is a challenging problem for similar
algorithms which removes all the hidden data. We solve this
challenge to mitigate any new font format changes by unifying
all data which leads a homogenous file. Finally, the result
reported by this implementation has outperformed similar
results reported in literature in terms of the hiding capacity and
the possibility of having such steganography mechanism used
in hiding information.
REFERENCES
[1] Aelphaeis Mangarae "Steganography FAQ," Zone-H.Org March
18th 2006.
[2] S. Dickman, "An Overview of Steganography," July 2007.
[3] V. Potdar, E. Chang. "Visibly Invisible: Ciphertext as a
Steganographic Carrier," Proceedings of the 4th International
Network Conference (INC2004), pp. 385–391, Plymouth, U.K.,
July 2004.
[4] M. Al-Husainy "Image Steganography by Mapping Pixels to
Letters," Science Publications, 2009.
[5] M. Shahreza, S. Shahreza, “Steganography in TeX Documents,”
Proceedings of Intelligent System and Knowledge Engineering,
ISKE 2008. 3rd International Conference, Nov. 2008.
[6] M. S. Shahreza, M. H. Shahreza, “An Improved Version of
Persian/Arabic Text Steganography Using "La" Word”
Proceedings of IEEE 6th National Conference on
Telecommunication Technologies, 2008.
[7] M. H. Shahreza, M. S. Shahreza, “A New Approach to
Persian/Arabic Text Steganography “
Proceedings of 5th
IEEE/ACIS International Conference on Computer and
Information Science, 2006.
[8] M. Aabed, S. Awaideh, A. Elshafei and A. Gutub “Arabic
Diacritics Based Steganograophy” Proceedings of IEEE
International Conference on Signal Processing and
Communications (ICSPC'07), 2007.
[9] W. Bender, D. Gruhl ,N. Morimoto ,A. Lu “Techniques for data
Hiding” Proceedings of IBM Systems Journal, Vol. 35, Nos 3&4,
1996.
[10] J. Brassil, , S. Low, N. Maxemchuk, L. Gorman, "Electronic
Marking and Identification Techniques to Discourage Document
Copying", in Proceedings of the 13th IEEE INFOCOM
Networking for Global Communications Conference, Oct.1995.
[11] K. Bennett, “Linguistic Steganography: Survey, Analysis, and
Robustness Concerns for Hiding Information in Text” Center for
Education and Research in Information Assurance and Security,
Purdue University, 2004.
[12] M. Nosrati , R. Karimi and, M. Hariri ,” An Introduction to
Steganography Methods” World Applied Programming, Vol. 1,
No. 3,pp. 191-195, Aug. 2011.
[13] M.H. Shirali-Shahreza, M. Shirali-Shahreza, "Text
Steganography in Chat" Proceedings of 3rd IEEE/IFIP
International Conference in Central Asia , Sept. 2007.
Ammar Odeh is a PhD. Student in University of
Bridgeport. He earned the M.S. degree in Computer
Science College of King Abdullah II School for
Information Technology (KASIT) at the University
of Jordan in Dec. 2005 and the B.Sc. in Computer
Science from the Hashemite University. He has
worked as a Lab Supervisor in Philadelphia
University (Jordan) and Lecturer in Philadelphia
University for the ICDL courses and as technical
support for online examinations for two years. He
served as a Lecturer at the IT, (ACS,CIS ,CS)
Department of Philadelphia University in Jordan,
and also worked at the Ministry of Higher
Education (Oman, Sur College of Applied Science)
for two years. Ammar joined the University of
Bridgeport as a PhD student of Computer Science
and Engineering in August 2011. His area of
concentration is reverse software engineering,
computer security, and wireless networks.
Specifically, he is working on the enhancement of
computer security for data transmission over
wireless networks. He is also actively involved in
academic community, outreach activities and
student recruiting and advising.
Qassim Bani Hani is Ph.D. candidate of computer
science and Engineering department in the
University of Bridgeport. His current research
interests include the design and development of
learning environment to support the learning about
heterogamous domain, collaborative discovery
learning and the development of mobile
applications to support mobile collaborative
learning (MCL), The congestion mechanism of
transmission of control protocol including various
existing variants, delivery of multimedia
applications. He completed his Bachelor degree in
computer science from Irbid National University in
2004 and Master degree in computer science from
Al-Balqa' Applied University in 2007. Qassim has
been directly involved in design and development of
mobile applications to support learning
environments to meet pedagogical needs of schools,
colleges, universities and various organizations.

Aladdin Alzubi received the B.Sc. in Software
Engineering from Philadelphia University, Amman,
Jordan in 2004, and the Master of Computer
Sciences from University Sians Malaysia –
Malaysia in 2006. In 2011 he joined University of
Bridgeport as Ph.D. student in computer science and
engineering at the University of Bridgeport,
Connecticut-USA. From 2000 to 2004.
Dr. Elleithy is the Associate Dean for Graduate
Studies in the School of Engineering at the
University of Bridgeport. He has research interests
are in the areas of network security, mobile
communications, and formal approaches for design
and verification. He has published more than one
hundred fifty research papers in international
journals and conferences in his areas of expertise.
Dr. Elleithy is the co-chair of the International Joint
Conferences on Computer, Information, and
Systems Sciences, and Engineering (CISSE). CISSE
is the first Engineering/Computing and Systems
Research E-Conference in the world to be
completely conducted online in real-time via the
internet and was successfully running for four years.
Dr. Elleithy is the editor or co-editor of 10 books
published by Springer for advances on Innovations
and Advanced Techniques in Systems, Computing
Sciences and Software.