Content uploaded by Ala' M. Alhusban
Author content
All content in this area was uploaded by Ala' M. Alhusban on Nov 04, 2017
Content may be subject to copyright.
International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 2, April 2017
DOI:10.5121/ijcsit.2017.9209 99
A
M
ELIORATED
K
ASHIDA
-
BASED
A
PPROACH FOR
A
RABIC
T
EXT
S
TEGANOGRAPHY
Ala'a M. Alhusban
and Jehad Q. Odeh Alnihoud
Computer Science Dept, Al al-Bayt University, Mafraq, Jordan
A
BSTRACT
Steganography is an art of hiding a secret message within some cover media such as: images, audios,
videos, and texts.
Many algorithms have been proposed for Arabic text steganography exploiting some
features of Arabic language such as: diacritics, extension character (kashida), and pointed letters. In this
research we propose a new method to enhance
a kashida-based methods for text steganography. In which
each existing kashida can hide two bits instead of only one bit. In addition, security measures is increased
through embedding the secret bits into the cover text by two different ways since the cover text is divided
into two blocks; each one of them is being treated in different way. Moreover, the original kashida in the
cover text
is ignored by the extractor.
A system is designed to achieve the embedding as well as the
extracting with high degree of security through authentication operation used in its interface. The proposed
approach is tested and compared with the most related kashida-based approaches in terms of capacity and
the results are promising. Furthermore, it overcomes the limitations of other approaches, maintain a
reasonable increases in the files size, and enhances security measures.
K
EYWORDS
Arabic text, Steganography, Kashida, Pointed characters, Zero-width character.
1.
I
NTRODUCTION
In general text is considered as one of the hardest cover media type, due to the limited options it
has, these limitations related to the nature of the direct appearance of the letters in their original
representations [14].
Many techniques have been introduced to hide a secret message in English language. However,
fewer methods have been proposed for Steganography in Arabic language. These methods have
mainly focused on some features of Arabic language such as diacritics, extensions (kashidas),
pointed and un-pointed letters and other techniques such as line shifting as well as spaces between
letters [5].
Kashida-based methods are concerned with the extension letter named kashida, which can be
added to the majority of Arabic letters, depending on their positions in the word. Such extension
cannot be added before the first letter of the word or after the last letter of the word, but it can
rather be added between two letters. Moreover, there are certain characters that do not accept
adding kashida after them: ( ﺇ،ﺃ،ء،ﺍﺁ،ﻭ،ﺯ،ﺭ،ﺉ،ﺅ، ), even if they come in the middle of the word [4].
There are two main types of Steganography; linguistic steganography and technical
steganography. Linguistic steganography refers to any form of steganography that uses language
in the cover. There are several examples of linguistic steganography. One of these is called open
codes, in which the readable text can contain certain secret text which can be in certain places in
the text, or can be hidden vertically or in reversed position. Technical steganography is the
International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 2, April 2017
100
method of steganography which uses a tool, a device or a method to conceal the message. Unlike
Linguistic steganography, dealing with the written words is not a must in Technical
Steganography [1]. Steganography process consists of two phases: embedding (hiding the secret
message) and extracting (retrieving the secret message) [11].
2.
L
ITERATURE
R
EVIEW
In [2], Gutub and Fattani proposed adding an extension to a pointed letter to represent a secret bit
of value (1) and adding an extension to a non-pointed letter to represent a secret bit of value (0).
By applying this method, you may add kashidas before or after the letters. Both options of adding
extensions before and after letters can be used within the same document but in different
paragraphs or lines. This method has some problems such as the large number of kashidas
required to represent the secret message, since a kashida represents only one bit, so that the size
of the document will increase significantly. Furthermore, extensive use of kashidas affects
security measure.
In [3], Shahreza and M. Shahreza used a method that is structured as follows; If the representative
forms of "Lam" (its Unicode is 0644) and "Alef" (its Unicode is 0627) letters are used in writing
"La" word, the text viewer sees the "La" word in its special form (" ﻟـﺎ "). In this case, we conclude
that bit (1) is hidden. However, by writing "La" in the normal form instead of inserting the Arabic
extension character between "Lam" (its Unicode is FEEO) and "Alef" (its Unicode is FE8E)
characters, it is written by using the code of the correct shape of each character ("ﺎﻠ"). We,
therefore, conclude that bit (0) is hidden. This method is so limited because it is implemented
only within the "La" word only.
In [4], Al-Nazer and Gutub proposed a steganography method that utilizes the extension character
(kashida) in Arabic language has been built to hide a secret message. The main goal is to
maximize the capacity by utilizing all possible locations for kashida in Arabic letters. A kashida is
put where it is applicable and the bit representation of the secret message has a value of (1). In
other words, if the extendable character must carry a secret bit of value (1), the kashida is
inserted. Otherwise, it is not inserted. Unfortunately, this method uses a kashida to hold a single
bit every time.
In [5], Al-Haidari. et. al. developed an approach, in which the number of kashidas in one word is
delimited by some equations to improve security. This method enhances the security over the
previous kashida -based methods, but it still stores one bit in each kashida.
In [6], Gutub. et. al. use a secret key to generate random kashida characters added to words where
a secret message is then embedded in the words as a watermarking code. The first addition of
random kashidas is designed for a confusion purpose to ensure security. After that, selected
kashidas are embedded based on the needed secret data to form the watermarking process in order
to save its applications. This method uses extra kashidas to ensure security. Nonetheless, this
maximizes the size of the cover text needed to hide the secret message.
In [7], Mahfoodh. et. al. presented method added one kashida to represent the secret bit (0) and
two consecutive kashidas when the hidden bit is (1). The number of kashidas needed using this
approach is huge. Besides the noticeable increasing in the file size, the extensive use of kashidas
affects the security.
Odeh and Elleithy [8], proposed a letter connectivity and extension to hide one bit approach.
Moreover, they used zero-width letter, which is a letter used to connect two letters with a small
International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 2, April 2017
101
effect on the shape of the two connected letters and no effect on the meaning of the word, to hide
two bits in each connective character. Table 1 shows the coding deployed in this method.
Table 1. Coding of method
Extension
Zero Width Code Letter effect
No No 00 No effect
Yes
No
01
Extension
No Yes 10 Zero Width
Yes Yes 11 Extension + Zero
Width
This method merges two techniques to achieve the goal of hiding a secret message in some cover
media prior to the encoding process. The first technique is to add a kashida and the second is to
add a letter named 'zero-width letter', which has its own Unicode symbol (U+200D). This
character causes the previous letter to appear as a connected letter. In this study, we develop a
new method in which a letter can hide up to two bits by using only kashidas instead of using
kashidas technique as well as zero-width technique.
In [9], Alginahi. et. al. inserted kashidas before a specific list of characters. If a kashida is
inserted before any of these characters, it means that there is a secret bit (1), and if there is one of
these characters without a kashida before it, this means that there is a secret bit (0). This method
gives better security than others, but it provides less capacity due to the addition of kashidas
before a specific number of characters as well as the representation of only one bit at a time.
In [10], Odeh. et. al. applied four scenarios randomly in order to improve data privacy. These four
scenarios are. First scenario, adding Kashida after pointed letters to hide one, otherwise, hide
zero. Second scenario, adding Kashida after non-pointed letters to hide one, otherwise, hide zero.
Third scenario, adding Kashida after letters to hide one, otherwise, hide zero. Fourth scenario,
adding Kashida after letters to hide zero, otherwise, hide one. Then, random selection is deployed
to select one of these scenarios. However, each one of these scenarios can hide one bit at a time
(either 1 or 0) in each extendable character.
In [12], Alginahi. et. al. encoded the original text depending on a predefined key. A kashida is
placed after a set of characters considering whether the letter is assumed to have high frequency
of recurrence in general or not. Each letter hides only one bit at a time.
In [15], Jabri and Ibrahim converted the secret message to an encrypted bits set. Then, they
exploited the spaces and the extendable characters to hide the encrypted bits. If the secret bit to be
embedded is one and the current character in the cover text is an extendable character, this letter
is extended by a kashida, and no kashida is added if the secret bit to be embedded is zero. In the
same way, if the current character in the cover text is a space between two words and the secret
bit to be embedded is one, an additional space is added after the original one, and no additional
space is added if the secret bit to be embedded is zero. This method uses the extendable characters
as well as the spaces between letters to hide one bit at a time.
International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 2, April 2017
102
3.
T
HE
P
ROPOSED
A
PPROACH
3.1.
E
MBEDDING
P
HASE
The proposed method depends on the nature of the letter before and after the kashida instead of
the nature of just one of the surrounding letters as used in the previous methods.
In Arabic language, 15 letters out of 28 alphabet letters are pointed by one, two or three points.
Hence, we can use the pointed letters to hide some secret bits and the unpointed letters to hide
some other bits. In this study, the nature of the letter, if it is a pointed letter or an unpointed letter,
is exploited to hide some secret bits in an extension (kashida) between two letters. It is used as a
formatting technique in order to lengthen a specific letter without having any change on the
meaning of the word. For example, the word "ﻥﺎﻛ" consists of three letters ("ﻥ","ﺍ","ﻙ"). If we add
a kashida after the first letter, the word will be:"ﻥﺎـﻛ". The word still has the same meaning it had
before adding the kashida despite the fact that it consists now of four letters ("ﻥ","ﺍ","ـ","ﻙ").
The embedding process starts with dividing the cover text into two blocks in which each block is
being dealt with using a special table. The dividing process is achieved by counting the number of
words of the cover text (m words), then dividing the number of words by two. After that, we take
the integer number resulted (n). The words from 1 to n are then being treated using a special table,
while the words from n+1 to m are being treated using another special table. In addition, the
secret message is divided into small blocks. Each one of them consists of two bits. Table 2, shows
the coding used for the first block.
Table 2. Embedding Code of the First Block
Secret
Bits
The Letter before
Kashida
The Letter
after Kashida
00 Pointed Pointed
01
Pointed
Unpointed
10 Unpointed Unpointed
11 Unpointed Pointed
While for the second block, Table 3 is deployed.
Table 3.Embedding Code of the Second Block
For instance, to hide the secret message: "110100010011" in the cover text:
" ﺭﻻﻭﺩ ﺕﺍﺭﺎﻴﻠﻣ ﺓﺮﺸﻋ ﻮﺤﻧ ﻥﺩﺭﻷﺍ ﻲﻓ ﺔﻳﺩﻮﻌﺴﻟﺍ ﺕﺍﺭﺎﻤﺜﺘﺳﻻﺍ ﺖﻐﻠﺑ ", we divide the secret message into blocks,
each two bits by their own:
Secret
Bits
The Letter before
Kashida
The Letter
after
Kashida
00 Unpointed Unpointed
01 Unpointed Pointed
10 Pointed Pointed
11 Pointed Unpointed
International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 2, April 2017
103
Table 4. Secret Message
110100010011
11 01 00 01 00 11
The cover text is divided into two blocks by dividing the number of the words of the cover text by
two:
Table 5. Cover Text
ﺭﻻﻭﺩ ﺕﺍﺭﺎﻴﻠﻣ ﺓﺮﺸﻋ ﻮﺤﻧ ﻥﺩﺭﻷﺍ ﻲﻓ ﺔﻳﺩﻮﻌﺴﻟﺍ ﺕﺍﺭﺎﻤﺜﺘﺳﻻﺍ ﺖﻐﻠﺑ (cover text)
ﺭﻻﻭﺩ ﺕﺍﺭﺎﻴﻠﻣ ﺓﺮﺸﻋ ﻮﺤﻧ ﻥﺩﺭﻷﺍ
(block no.2)
ﻲﻓ ﺔﻳﺩﻮﻌﺴﻟﺍ ﺕﺍﺭﺎﻤﺜﺘﺳﻻﺍ ﺖﻐﻠﺑ
(block no.1)
The embedding process starts from right to left by taking the first two bits which are "11", and
then we search for an un-pointed letter followed by a pointed letter in the first block of the cover
text, to put a kashida between them, since "11"means that there must be a kashida between an un-
pointed letter and a pointed letter (according to the table specified for the first block). The first
case we find is that the letter "ﻝ" is followed by the letter "ﻍ", and thus we insert a kashida
between the two letters and move on to the next two bits which are "00". We then search for a
pointed letter followed by another pointed letter starting from the letter which is placed after the
kashida we have inserted "ﻍ". The first un-pointed letter followed by a pointed letter we find here
is the letter "ﻍ" followed by the letter "ﺕ". Then, we continue to the next two bits "01" and search
for a pointed letter followed by an un-pointed letter. Here, we find that the letter "ﺙ" is followed
by the letter "ﻡ", so we insert a kashida between them. After that, we take the next two bits which
are "00", and search for a pointed letter followed by another pointed letter in the cover text to put
a kashida between them. The first case we find is that the letter "ﻱ" is followed by the letter "ﺓ",
so we insert a kashida between the two letters and move on to the next two bits which are "01".
We now search for a pointed letter followed by an un-pointed letter. While we search, we reach
the end of the first block without finding a pointed letter followed by an un-pointed letter, so we
move to the second block. According to the table specified for the second block, "01" means that
there must be a kashida between an un-pointed letter and a pointed letter. The first case we find is
the letter "ﻉ" and the letter "ﺵ", so we put a kashida between them. After that, we continue to the
last two bits "11" and search for a pointed letter followed by an un-pointed letter. Here, we find
that the letter "ﺵ" is followed by the letter "ﺭ". Consequently, we insert a kashida between them.
At this point, the embedding process comes to an end since all the secret bits have been embedded
within the cover text.
- Original Kashida Case:
In order to distinguish between the original kashida in the text and the added kashidasin the
embedding process, an additional kashida is added after any original kashida in the sentence. For
example, if the cover text is" ﺐﺼﻌﺘﻟﺍ ﺬﺒﻧﻭ ﺢﻣﺎﺴﺘﻟﺍ ﻰﻟﺇ ﻮﻋﺪﻳﻭ ،ﺱﺎﻨﻟﺍ ﻦﻴﺑ ﻡﻼﺴﻠﻟ ﻡﻼﺳﻹﺍ ﻱﺩﺎـﻨﻳ ", and there is an
original kashida in the word "ﻱﺩﺎـﻨﻳ" between the letters "ﻥ" and "ﺍ", the system, in this case, adds
another kashida after this kashida in order to notify the extractor to ignore it when extracting the
hidden bits. If we enter this sentence as the cover text and enter the secret message "01", "01"
means that there must be a kashida between a pointed letter and an un-pointed letter, according to
the table specified for the first block. The first case the extractor finds is the letter "ﻥ" followed by
the letter "ﺍ" in the first word ("ﻱﺩﺎـﻨﻳ"), but since there is an original kashida between them, it adds
another kashida after the original one to inform the extractor that this is an original kashida, then
continues searching for a pointed letter followed by an un-pointed letter. It finds the letter "ﻥ"
International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 2, April 2017
104
followed by the letter "ﺍ" in the fifth word ("ﺱﺎﻨﻟﺍ"), so it adds a kashida between them. Since there
are no more bits in the secret message, the embedding process ends.
- Embedding Algorithm:
The embedding process deals with four groups of letters:
- Pointed letters that accept a kashida after them (PL1):
)ﺏ ،ﺥ ،ﺽ ،ﻅ ،ﻍ ،ﻑ ،ﻥ ،ﺕ ،ﻕ ،ﻱ ،ﺙ ،ﺵ(
- Pointed letters that accept a kashida before them (PL2):
)ﺥ ،ﺫ ،ﺯ ،ﺽ ،ﻅ ،ﻍ ،ﻑ ،ﻥ ،ﺕ ،ﻕ ،ﻱ ،ﺙ ،ﺵ ،ﺓﺏ ،ﺝ ،(
- Un-pointed letters that accept a kashida after them (UPL1):
)ﺡ ،ﻡ ،ﺱ ،ﻩ ،ﻝ ،ﻙ ،ﻉ ،ﻁ ،ﺹ(
- Un-pointed letters that accept a kashida before them (UPL2):
)ﺁ ،ﺍ ،ﺃ ،ﺡ ،ﺩ ،ﺭ ،ﻉ ،ﻁ ،ﺹ ،ﻭ ،ﻩ ،ﺱ ،ﻡ ،ﻝ ،ﻙ(
Figure 1, shows the proposed embedding algorithm.
Emb_Alg
Input : Cover Text (CT), Secret Message (SM)
Output: Stego Text (ST)
1) Let i:= -1, c:= -1, m = No. of words in CT
2) If m:= 1
Block 1 := one word
Block2 := empty
Else
Block 1 := words(1 To n = trun(m/2))
Block2 := words (n+1 To m)
3) i:= i + 2
Read 2 bits (b1, b2) SM, at positions i, i+1.
4) c:= c + 1
read 2 characters CT, at positions c, c+1.
If {
If {
If {
- insert (-) after x1
- If (i := length (SM) || c:= length (CT) then EXIT
Else go to (3) } }
Else go to (4)
Else if {
If {
-insert (-) after x1
- If (i := length (SM) || c:= length (CT) then EXIT
Else go to (3) } }
Else go to (4)
Else if {
If {
-insert (-) after x1
- If (i := length (SM) || c:= length (CT) then EXIT
International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 2, April 2017
105
Else go to (3) } }
Else go to (4)
Else
If {
-insert (-) after x1
- If (i := length (SM) || c:= length (CT) then EXIT
Else go to (3) }
Else go to (4) }// End of Block 1
Else // x1 && x2 belongs to block 2
If {
If {
- insert (-) after x1
- If (i := length (SM) || c:= length (CT) then EXIT
Else go to (3) } }
Else go to (4)
Else if {
If {
-insert (-) after x1
- If (i := length (SM) || c:= length (CT) then EXIT
Else go to (3) } }
Else go to (4)
Else if {
If {
-insert (-) after x1
- If (i := length (SM) || c:= length (CT) then EXIT
Else go to (3) } }
Else go to (4)
Else
If {
-insert (-) after x1
- If (i := length (SM) || c:= length (CT) then EXIT
Else go to (3) }
Else go to (4) }// End of Block 2
Figure 1. Embedding Algorithm
3.2.
E
XTRACTING
P
HASE
In a similar way to the embedding process, the extracting of the secret message from the stego
text starts by dividing the message (stego text) into two blocks in which each block is being
considered using a special table. The dividing process is achieved by counting the number of
words of the cover text (m words). Then we divide the number of words by two to get the
integer number (n). The words from 1 to n are then being treated using a special table and the
words from n+1 to m are being treated using another special table. The extractor starts dealing
with the first block by searching from right to left for a kashida. If a kashida is found, the
extractor takes the letter before the kashida as well as the letter after it and looks in the
extracting table specified for the first block in order to extract the two bits which the founded
kashida represents. These two bits are inserted in the secret message obtained. Then, the
extractor continues searching for another kashida in the first block. If the extractor finds a
kashida, it looks for the nature of the letters surrounding the kashida and looks in the first
International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 2, April 2017
106
block extracting table to extract the two bits that the kashida represents. If the first block is
finished, the extractor moves to the second block and starts searching for kashidas in the same
way. If any kashida is found, the extractor looks at the table of the second block to interpret
what the kashida represents. For example, if we have the stego text: " ﻢﻜـﻴﻠﻋ ﻡﻼﺴـﻟﺍ ﻪﻠـﻟﺍ ﺔـﻤﺣﺭﻭ
ﻪـﺗﺎﻛﺮﺑﻭ", this sentence consists of five words. Each one of these words has one kashida (the
word ﻡﻼﺴـﻟﺍ" " has a kashida between the letters "ﻝ" and "ﺱ". The word "ﻢﻜـﻴﻠﻋ" has a kashida
between the letters "ﻱ" and "ﻙ". The word " "ﺔـﻤﺣﺭﻭ has a kashida between the letters "ﻡ" and
"ﺓ", and the word "ﻪﻠـﻟﺍ" has a kashida between the letters "ﻝ" and "ﻝ"). The decoder divides
this sentence into two blocks by dividing the number (5) by (2). The integer number it gets is
(2), so the first block consists of the first two words and the second block consists of the rest
words of the sentence.
The first block"ﻢﻜـﻴﻠﻋ ﻡﻼﺴـﻟﺍ" is treated using Table 6.
Table 6. Extracting Table of the First Block
Secret Bits The Letter before Kashida The Letter after Kashida
00
Pointed
Pointed
01 Pointed Un-pointed
10
Un
-
pointed
Un
-
Pointed
11 Un-pointed Pointed
The Extracting process starts from the right to the left searching for a kashida. If the kashida is
found and the surrounding letters of it belong to one of the two sets (pointed or un-pointed
letters), it looks for the two bits this kashida represents. The first kashida we find here is placed in
the first word which belongs to the first block ("ﻡﻼﺴـﻟﺍ") between the letter "ﻝ" and the letter "ﺱ",
which are an un-pointed letter and another un-pointed letter, so the extractor looks into the table
to find the two bits represented. These two bits are "10", so these bits are inserted in the secret
message extracted by the extractor. After that, we move to the left, searching for another kashida.
We will find another one in the second word ("ﻢﻜـﻴﻠﻋ") between the letter "ﻱ" and the letter" "ﻙ,
which are a pointed letter and an un-pointed letter, according to the table. This reflects the two
bits: "01". This is the last existing kashida in the first block. Consequently, the next kashidas
found will be extracted using Table 7, which specified for the second block:
Table 7. Extracting Table of the Second Block
Secret Bits The Letter before Kashida The Letter after Kashida
00 Un-pointed Un-pointed
01
Un
-
pointed
Pointed
10 Pointed Pointed
11 Pointed Un-pointed
The first kashida found in the second block is placed in the word "ﺔـﻤﺣﺭﻭ" between the letters"ﻡ"
and "ﺓ", which represents the two bits: "01". Since it is located between an un-pointed letter and a
pointed letter, the next kashida we find is the kashida in the word "ﻪﻠـﻟﺍ" between the letters "ﻝ"
and "ﻝ". We add the two bits:"00" to the secret message extracted because the kashida was found
between an un-pointed letter and another un-pointed letter. The last kashida in the second block is
placed in the word "ﻪـﺗﺎﻛﺮﺑﻭ" between the pointed letter "ﺕ" and the un-pointed letter "ﻩ", so we
add "11" to the secret message extracted. The secret message obtained is: "11 00 01 01 10".
Figure 2, shows the proposed extracting algorithm.
International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 2, April 2017
107
- Extracting Algorithm:
Input: Stego text, pointed characters list, un-pointed characters list
Output: Secret message
Abbreviations: pointed characters list: PL, un-pointed characters list: UPL, character:
char
1. Get stego text
2. If the number of the words of the cover text is 1
The first block of the cover text consists of the word number one
The second block is empty
Else
Divide the cover text into two blocks from right to left:
N:= number of the words of the cover text
J:= integer value results from dividing
The first block of the cover text: (words from 1 to J)
The second block of the cover text: (words from J+1 to N)
3. While a character in the first block exists (from right to left):
If the character is "_":
If the character before it PL
If the character after it PL
Add 00 to the secret message
Else if the character after it UPL
Add 01 to the secret message
Else if the character before it UPL
If the character after it UPL
Add 10 to the secret message
Else if the character after it PL
Add 11 to the secret message
4. While a character in the second block exists(from right to left):
If the character is "_":
If the character before it UPL
If the character after it UPL
Add 00 to the secret message
Else if the character after it PL
Add 01 to the secret message
Else if the character before it PL
If the character after it PL
Add 10 to the secret message
Else if the character after it UPL
Add 11 to the secret message
5. Output: secret message
Figure 2. Extracting Algorithm
3.3.
S
YSTEM
D
ESIGN
The implementation of the proposed method is represented by a system designed using c#. This
system is called "ATSKS", which refers to "Arabic Text Steganography using kashida System".
The system processes the operations of embedding as well as extracting, but before that there is a
International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 2, April 2017
108
hashing phase used to secure the system. If a user passes it correctly, they will enter the correct
system. Otherwise, they will enter a fake system. Three tables are used in this system; one table
for hashing and two for embedding as well as extracting.
4.
E
XPERIMENTAL
R
ESULTS
A
ND
D
ISCUSSION
4.1.
C
APACITY
The capacity ratio is computed by dividing the amount of hidden bytes over the size of cover text
media in bytes as in [5]. One of the major contributions of the proposed method introduced in this
research is to hide two bits using only one kashida depending on its position between two letters.
This method is supposed to increase the capacity of the hidden message since a kashida can hide
two bits. In addition, there is no need to merge two methods to hide two bits in one kashida. We
have used 10 different essays as cover text files of various sizes to hide secret data. Then capacity
ratio of each cover text file is computed. Table 8, shows the capacity ratio of the proposed
method. We have also calculated the maximum possible capacity (optimistic case).
Table 8. Capacity ratio of the proposed method
Essay
No.
Website Address Cover Text
Size
(Bytes)
No. of
Hidden
Bits
Capacity
Ratio
(%)
1 http://motaded.net/show-2567052.html 362 352 12.15
2 http://www.shorouknews.com/news/view.
aspx?cdate=04032016&id=944d3cee-
c368-4864-8420-8b5e6839fb65
674 664 12.31
3 http://www.alanbatnews.net/print.php?nid
=120388
924 896 12.12
4 http://assabeel.net/culture/item/161130 1036 1056 12.74
5 http://www.al-
ayyam.ps/ar_page.php?id=109f47edy278
874093Y109f47ed
1174 1192 12.70
6
http://www.alriyadh.com/1134698
1203
1224
12.72
7 http://www.elkhabar.com/press/article/10
1630
2061 2056 12.47
8 http://www.ahram.org.eg/NewsQ/483345.
aspx
2361 2408 12.74
9 http://www.alwatannews.net/NewsViewer
.aspx?ID=118537
5442 5544 12.73
10 http://www.alrai.com/article/771518.html 20435 20400 12.48
The capacity ratio of the proposed approach for each one of the essays doubles the capacity ratio
of most of the previous kashida approaches that hided one bit at a time. For example, the
extendable characters found in essay number 1 are 178. These characters are able to store 356 bits
at best using the proposed method, while they can store only 178 bits at best using the previous
kashida approaches which store 1 bit at each extendable location.
In contrast with [8], we do not need to add another character beside the kashida to represent two
bits. Each single existing kashida using the proposed method hides two bits by its own without
merging it with a zero-width character which is used in this study just to distinguish between the
original kashida and the kashida used to store secret bits. Table 9, shows the capacity ratio values
International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 2, April 2017
109
of three other methods using the10 essays in Table 8 as cover text files. We have also calculated
the maximum possible capacity (optimistic case):
Table 9. Capacity ratio of methods [9, 15]
Essay
No.
Cover
Text
Size
(bytes)
No. of
Hidden
Bits
Using
Method
[9]
Capacity
Ratio
Using
Method
[9] (%)
No. of
Hidden
Bits
Using
Method
[15]
Capacity
Ratio
Using
Method
[15] (%)
1 362 72 2.49 248 8.56
2 674 152 2.81 464 8.61
3 924 192 2.60 616 8.33
4
1036
232
2.80
720
8.69
5 1174 232 2.47 824 8.78
6 1203 216 2.24 856 8.89
7 2061 448 2.72 1488 9.02
8 2361 480 2.54 1472 7.79
9 5442 952 2.19 3824 8.78
10 20435 4712 2.88 14392 8.80
The average capacity ratio of method [9, 15] is 2.57%, and 8.63% respectively. The average
capacity ratio of the proposed method is 12.52%. It is obvious that the capacity ratio of the
proposed method outperforms the capacity ratio achieved by [9, 15].
We have calculated the results by taking all the extendable characters in the cover text. Then, we
divided them by 8 to get the number of secret bytes that we may hide within the cover text. After
that, we divided the size of secret message (bytes) by the size of cover text (bytes) to get the
capacity ratio. By contrast, in [15], they took all the extendable characters as well as the
whitespaces as secret message carriers. Then, they divided them directly by the total number of
characters and whitespaces. It is obvious that this way of calculation is incorrect and yield to
misleading results.
4.2.
E
MBEDDING
R
ATIO
Another measurement used to evaluate the proposed approach is the "Embedding Ratio(ER)", as
in [13] which is used to determine the total fitness of the hidden text that can be embedded in
cover text:
X 100% …..(1)
X 100% …..(2)
We have calculated the embedding ratio of the same essays:
International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 2, April 2017
110
Table 10. Embedding Ratio Values of the Proposed Method
Essay No. Cover Text Size
(bytes)
Number of
Embedded Bits
Embedding Ratio
(%)
1 362 352 10.84
2 674 664 10.96
3
924
896
10.81
4 1036 1056 11.30
5
1174
1192
11.26
6 1203 1224 11.28
7
2061
2056
11.09
8 2361 2408 11.30
9 5442 5544 11.30
10
20435
20400
11.09
These embedding ratio values exceed the values of the embedding ratio for the previous methods
which hide one bit at a time, because the proposed method uses less number of kashidas to embed
the same number of secret bits within the same cover text. The following table shows the
embedding ratio values of the same 10 essays used as cover text files which have been used in
Table 10 for two other methods:
Table 11. Embedding Ratio Values of Three Previous Methods
Essay
No.
Cover
Text Size
(bytes)
No. of
Embedded
Bits
Using
Method [9]
Embedding
Ratio
Using
Method [9]
(%)
No. of
Embedded
Bits
Using
Method [15]
Embeddi
ng Ratio
Using
Method
[15] (%)
1
362
72
2.43
248
7.88
2 674 152 2.74 464 7.92
3
924
192
2.53
616
7.69
4 1036 232 2.72 720 7.99
5 1174 232 2.41 824 8.06
6 1203 216 2.19 856 8.17
7 2061 448 2.65 1488 8.28
8
2361
480
2.83
147
2
7.23
9 5442 952 2.14 3824 8.07
10
2
0435
4712
2.80
14392
8.09
The average embedding ratio of method [9, 15] is 2.54%, and 7.94% respectively. While, the
proposed method achieves 11.12% as average embedding ratio. This was expected, since a
capacity ratio affects positively or negatively the embedding ratio.
5
.
C
ONCLUSION
A new method to hide secret message within Arabic texts has been proposed. This method
depends primarily on the nature of letters; whether they are pointed or un-pointed letters. We have
exploited this feature of Arabic text to add a kashida (-). Since there are two cases of each letter;
pointed or un-pointed, a table of four cases is used to add a kashida between two letters every
time to hide two bits in each kashida. The most common kashida-based methods hided just one
bit in each kashida or used a kashida as well as a zero-width character to hide two bits, while the
International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 2, April 2017
111
proposed method hided two bits using kashida without the need to add the zero-width character.
Adding zero-width character increases the file size dramatically, which affects the security
measures. By hiding two bits in each kashida, the capacity is remarkably increased as compared
with some of the well-known kashida-based approaches. The dependency on the nature of both of
the surrounding letters of the kashida as well as dividing the cover text into two blocks, each one
is being dealt with in a different way, increases the security of the proposed method. Since,
intruders are not able to expect the method of extraction. Furthermore, the original kashida case is
considered. Ignoring the original kashida in the cover text affects the accuracy of extraction
phase. A system of embedding the secret message within a cover text and extracting the secret
message from a stego text has been built. This system has a hashing phase in order to distinguish
between the authorized user and the un-authorized user, which may increase the robustness of the
system. In some cases, the capacity ratio of the proposed method may affected by the sequences
of the secret bits, and suitable appearances of the targeted kashida(-). This might be considered as
limitation that may yield to undesirable results in such cases. As a future work, we should
overcome this drawback by proposing a suitable method to make use of all kashidas irrespective
to the sequences of the secret bits.
R
EFERENCES
[1] G. Kipper, (2003) Invistigator's Guide to Steganography, Auerbach Publications, ISBN
9780849324338, October 27.
[2] Gutub, A. and M. M. Fattani, (2007) "A Novel Arabic Text Steganography Method Using Letter
Points and Extensions", International Journal of Computer, Electrical, Automation, Control and
Information Engineering Vol: 1, No: 3.
[3] M. S. Shahreza and M. H. Shahreza, (2008) "An Improved Version of Persian/Arabic Text
Steganography Using "La" Word", Proceedings of IEEE 2008 6th National Conference on
Telecommunication Technologies and IEEE 2008 2nd Malaysian conference on Photonics, 26-27
August, Putrajaya, Malaysia, pp: 372-376.
[4] A. Al-Nazer, A. Gutub, (2009) "Exploit Kashida Adding to Arabic E-Text for High Capacity
Steganography", Proceedings of Third IEEE International Conference on Network and System
Security, Pages: 447-451. DOI: 10.1109/NSS.2009.21.
[5] F. Al-Haidari, A. Gutub, K. Al-Kahsah, and J. Hamodi, (2009) "Improving Security and Capacity for
Arabic Text Steganography Using 'Kashida' Extensions", Proceedings of IEEE/ACS International
Conference on Computer Systems and Applications.pp: 396-399.DOI:
10.1109/AICCSA.2009.5069355
[6] A. Gutub, F. Al-Haidari, K. Al-Kahsah, and J. Hamodi, (2010) "E-Text Watermarking: Utilizing
'Kashida' Extensions in Arabic Language Electronic Writing", Journal of Emerging Technologies in
Web Intelligence, Vol. 2, No. 1, Pages: 48-55. DOI: 10.4304/jetwi.2.1.48-55.
[7] A. Gutub, W. Al-Alwani, and A. Bin Mahfoodh, (2010) "Improved Method of Arabic Text
Steganography Using the Extension ‘Kashida’ Character", Bahria University Journal of Information
& Communication Technology Vol. 3, Issue 1, pp: 68-72.
[8] A. Odeh and K. Elleithy, (2012) "Steganography in Arabic Text Using Zero Width and Kashidha
Letters", International Journal of Computer Science & Information Technology (IJCSIT), Vol. 4,
Pages: 1-11. DOI: 10.5121/ijcsit.2012.4301.
[9] Y. M. Alginahi, M. N. Kabir and O. Tayan, (2013) "An Enhanced Kashida-Based Watermarking
Approach for Increased Protection in Arabic Text-Documents", Proceedings of International
Conference on Electronics, Computer and Computation (ICECCO). DOI:
10.1109/ICECCO.2013.6718288
[10] A. Odeh, K. Elleithy and M. Faezipour, (2013) "Steganography in Arabic Text Using Kashida
Variation Algorithm (KVA)", Proceedings of the 2013 IEEE Systems, Applications and Technology
Conference (LISAT). DOI: 10.1109/LISAT.2013.6578239.
[11] Mersal, S., Alhazmi S., R. Alamoudi and N. Almuzaini, (2014) "Arabic Text Steganography in
Smartphone", International Journal of Computer and Information Technology (ISSN: 2279 – 0764),
Vol. 03, No. 02, pp. 441-445.
International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 2, April 2017
112
[12]Y. Alginahi, M. Kabir and O. Tayan, (2014) "An Enhanced Kashida-Based Watermarking Approach
for Increased Protection in Arabic Text- Documents Based on Frequency Recurrence of Characters",
International Journal of Computer and Electrical Engineering, Vol.6, No. 5. DOI:
10.177706/ijcee.2014.v6.857.
[13] B. Osman, R. Din and M. R. Idrus, (2015) "Capacity Performance of Steganography Method in Text
Based Domain", ARPN Journal of Engineering and Applied Sciences, Vol. 10, No.
3.http://repo.uum.edu.my/id/eprint/14832.
[14] R. Saluja, K. Kanwal and S. Dahyia, (2014) "Review on Steganography for Hiding Data",
International Journal of Computer Science and Mobile Computing, Vol.3 No. 4, pp: 225-229. ISSN
2320–088X.
[15] R. Jabri and B. Ibrahim, (2016) "Capacity Improved Arabic Text Steganography Technique Utilizing
‘Kashida’ with Whitespaces", Proceedings of the Third International Conference on Mathematical
Sciences and Computer Engineering (ICMSCE2016), pp: 38-44, Langkawi, Malaysia.
A
UTHORS
Jehad Q. Odeh received his Ph.D. of Computer Science from University Putra Malaysia in
2004. Currently, Dr. Jehad is an Associate Professor at the Faculty of Information
Technology in Al al-bayt University, Jordan. His research areas include image retrieval and
indexing, image processing, and steganography. He has published a number of papers
related to these areas, in addition to supervision on many MSc theses in different computer
science fields.
Ala'a M. Alhusban received here MSc. of computer Science from Al al-Bayt university in
2017. Her research areas include text retrieval and indexing, and steganography.
