This paper proposes a 2-tier data hiding technique that involves Exploiting Modification Direction (EMD) method and Huffman Coding. Firstly, a secret message of an arbitrary plain text is encrypted, compressed and transformed into a stream of bits. Subsequently, the bits are converted into secret digits by using the Huffman dictionary table. Secondly, a cover image is segmented into groups of n pixels and each group is embedded with one secret digit by modifying one gray-scale value at most to hide the secret digit in (2n+1)-ary notational system. The experimental results have shown that both PSNR and payload of the proposed method are higher than that of well-known methods namely, OPAP, EMD and Opt EMD.

1. ACEEE Int. J. on Information Technology, Vol. 02, No. 02, April 2012
A 2-tier Data Hiding Technique Using Exploiting
Modification Direction Method and Huffman Coding
Ali m-Ahmad1a, Ghazali Bin Sulong1b, Mohd. Shafry B. Mohd. Rahim1c, Saparudin2
Department of Computer Graphic & Multimedia1,
Faculty of Computer Science and Information Systems,
Universiti Teknologi Malaysia, 81310 Skudai, Johor, Malaysia.
Email:{ aa5555sobel@yahoo.com; bghazali@spaceutm.edu.my; cshafry@utm.my}
Fakultas llmu Komputer2, Universitas Sriwijaya,
Kampus Unsri-lndralaya, IO, Sumatera Selatan, Indonesia.
Email: saparudin1204@yahoo.com
Abstract—This paper proposes a 2-tier data hiding technique enemy is allowed to detect, intercept and modify messages
that involves Exploiting Modification Direction (EMD) method without being able to violate certain security premises
and Huffman Coding. Firstly, a secret message of an arbitrary guaranteed by a cryptosystem, the aim of steganography is
plain text is encrypted, compressed and transformed into a to hide secret information inside other harmless messages to
stream of bits. Subsequently, the bits are converted into secret
ensure nobody except the receiver can detect that there is a
digits by using the Huffman dictionary table. Secondly, a
cover image is segmented into groups of n pixels and each
second message present.
group is embedded with one secret digit by modifying one The steganography system consists of two procedures
gray-scale value at most to hide the secret digit in (2n+1)-ary which are: the embedder and a detector. The embedder takes
notational system. The experimental results have shown that two inputs. One is the payload to embed the secret message,
both PSNR and payload of the proposed method are higher and the other is the cover Work in which to embed the
than that of well-known methods namely, OPAP, EMD and payload. The output of the embedder is typically transmitted
Opt EMD. or recorded. Later, that Work is presented as an input to the
detector. Most detectors try to determine whether a payload
Index Terms—data hiding, steganography, Exploiting
is present, and if so, output the message encoded by it [1].
Modification Direction (EMD) method, Huffman coding, data
Thus stenography is used to protect secret information by
protection, cryptography.
embedding secret messages in the host media (cover image),
for instance: text, images, audio or video.
I. INTRODUCTION
The need for methods that provides efficient work on the II. BACKGROUND OF THE PROBLEM
protection of data and private property of individuals has
become very vital due to the huge growth of multimedia The basics of steganography are inaudibility, robustness
applications on networks. It is therefore important to create and data rate (payload). The required trade-offs between the
methods that provide security for the media from thieves and image quality and the payload, the steganography algorithm,
hackers to prevent them from tampering and misrepresenting must be at an acceptable level. In steganography, it is very
the data. hard to embed a large amount of data and preserve the high
Data protection consists of two techniques: image quality at the same time. Therefore, if it is required to
cryptography and data hiding. Cryptography means the have more payloads, the image quality will be low. Conversely,
provision of protection for data storage and data transfer a steganography algorithm is usually not efficient with high
while using a secret key. Encryption is still a successful way payload embedding [3].
to protect stored and transmitted data over a network. But, Steganography techniques are mostly divided in two
with the growing use of networks to send and receive data groups: spatial domain [8, 9, 11] and frequency domain [10,
on the global information network, it has become very difficult 12, 13, 14]. The former embeds messages directly into the
to maintain this data. Data hiding has two main approaches: image pixels. The Least Significant Bit (LSB) insertion method
steganography and digital watermarking. is probably the most well-known image steganography
technique. This method is easy to implement and has high
These two approaches have many techniques [2], one of
quality but, it is extremely vulnerable to attacks such as image
them is LSB steganography which is defined as the practice
manipulation (compression and low pass filtering).
of undetectably altering a Work to embed a secret message.
Contrary to the above method, frequency domain
Also, watermarking is defined as the practice of imperceptibly
technique embeds messages in the frequency coefficients of
altering a Work to embed a message about that Work [1].
images by using a frequency-oriented mechanism such as
Steganography is also defined as the art and science of
Discrete Cosine Transform (DCT) [17], Discrete Fourier
communicating in a way which hides the existence of the
Transform (DFT), Discrete Wavelet Transform (DWT) [17],
communication. In contrast to Cryptography, where the
and so on. Generally, embedding in the low frequency can
© 2012 ACEEE 46
DOI: 01.IJIT.02.02.34

2. ACEEE Int. J. on Information Technology, Vol. 02, No. 02, April 2012
withstand many attacks but provide clear effect on the media.
In contrast, embedding in the high frequency provide less
impact on the quality of images but creates low robustness
to different attacks [16]. Where n indicates no. of pixels, g1, g2…, gn represent values
Now the question is: How could the payload is increased of pixel within each group, s refers to index of the image, d
while maintaining high image quality? indicates value of secret digit.
If f = d no change is needed , if f # d and s lesser or equal to
III. RELATED WORK n, g(s) is increase by 1.If f ‘“ d and s is greater than n, g(2n+1-
) is decrease by 1.
The LSB is considered one of the most common methods s
For example, let n=4, d=3 and the original pixel-group be [99,
[6, 7, 8]. This approach deals directly with a cover-image,
110, 120, 130]. Then f = (99 * 1 + 110 * 2 + 120 * 3 + 130 * 4 )
after hiding a secret image within it. In general, bit-mapped
mod 9 = 2, s = (d – f ) mod 9 = 1 which is less than n. Thus, g(s)
images are commonly used. Every image consists of a set of
= g(1) + 1 = 100.
pixels, and every pixel represents one colour. The values of a
Second example: let d=9 and the same pixel-group is
gray-scale image range from 0 to 255. If the value of the pixel
employed. Then s = (d – f ) mod 9 = 7 which is bigger than n.
is equal to 0, it signifies darkness, and when is equal to 255,
thus, g(2n+1-s) = g(2) – 1 = 109.
it indicates lightness. Therefore, a gray level image can be
The embedded data can be extracted using the following
adjusted by adjusting these values. At least 8 bits are required
extraction function of stego-pixel-group (4).
to represent these values, and the binary system stores them
from bits a1, a2… a8. The LSB substitution changes the last bit
(a 1) to make imperceptible modification that cannot be
Ref. [5] proposes the relationship between the value of n and
detected by the human vision system. For instance, if the
amount of payload that minimizes cover image distortion.
value of a pixel is 100, and we want to embed a 1, the pixel
Here, the value of n is computed prior to embedding using
value becomes 101. Human vision cannot recognize this
(5).
difference. However, the LSB can easily embed secret data
into an image with an impalpable effect on the image. The
pixel values after embedding can be computed using (1).
where Is means the total number of pixels of the cover image,
x’i=xi - xi mod 2k +mi (1) n indicates the amount of pixels for each group and p
Where xi, and x’i refer to the pixel before and after embedding represents the payload to be embedded.
respectively, k refers to the number of bits which going to
embed and mi refers to the value of secret message. Other IV. PROPOSED METHOD
researchers used Intermediate Significant Bit (ISB) planes
The main aim of this method is to hide a large amount of
[18, 19, 20, 21, 22, 23] to overcome the LSB drawbacks. Here,
secret message in high secrecy and the same time preserves
higher bit-plane is selected to embed the secret messages.
the cover image’s quality. Here, a new 2-tier data hiding
Thus, it reduces stego-image quality but increases
technique is proposed that involves the EMD method and
robustness. The higher the bit-plane is chosen the better
the lossless Huffman coding. At first, the secret data is
robustness would be but the poorer image quality would
encrypted, compressed and transformed into a stream of bits.
become and vice versa [24].
Then the bits are converted into secret digits by using the
In the Optimal Pixel Adjustment Process (OPAP) proposed
Huffman dictionary table. Subsequently, the cover image is
by [6], this algorithm is effectively reduced distortion. The
segmented into groups of n pixels and each group is
OPAP makes use of the last n bits to embed data and at the
embedded with one secret digit by modifying one gray-scale
same time toggles the n+ 1 bit while comparing the toggle
value at most to hide the secret digit in (2n+1)-ary notational
with least distortion. For example, bit stream “0011” if the
embedding is”00"; the result is either “0000” or “0100 “. It is system.
obvious that changing “0011” to “0100” will cause less dis- Step1: Construct a table containing individual letters and
tortion in comparison to “0000”. Although this technique
their frequency numbers.
may reduce distortion, it still weak when value of n is more
Step2: Sort the letters in an ascending order according to
than two bits. Thus, the technique cannot handle higher im-
their frequency numbers.
age quality.
Step 3: Add the first two frequency numbers and then re-
Ref. [4], proposed the Exploiting Modification Direction
arrange the table again.
(EMD) to reduce distortion. EMD made use of n pixels as a
Step 4: Repeat step 3 until a single frequency number is
group to embed secret digits in a (2n+ 1)-ary notational
achieved.
system. During embedding stage, it requires to increase or
Step 5: Construct Huffman tree by assigning each pair of
decrease one from the value of a particular pixel within the
branches with (0,1) for all branches of the tree.
group using (2) and (3). Step 6: Rewrite the letters according to the above
Huffman tree.
© 2012 ACEEE 47
DOI: 01.IJIT.02.02. 34

3. ACEEE Int. J. on Information Technology, Vol. 02, No. 02, April 2012
Step 7: Construct the final table containing all the secret Step2: If f ‘“ d then s = d – f mod(2n+1)
letters with their codes. Step3: If s < n then increase g(s) by 1
Example, let the message be thisis Step4: if s > n then decrease g(2n+1-s) by 1
Step5: Repeat 1 to 4 until all the digits of the secret
message are processed.
Figure 2. Flow chart for embedding process
C. Extraction of The Secret Codes
The stego-image is received and (4) is applied to extract
the codes. Then these codes are converted into binaries using
Huffman dictionary and finally the original message of plain
text is deciphered using Huffman decoding. Fig. 3 depicts
the extraction method.
Extraction algorithm
Input: Stego-image
Output: Secret data
Step1: For i=1:n
d=(sum(gi’ * i)) mod (2n+1)
End
Step2: Convert (d) to binary.
Step3: Decode this binary number to original letter by
using Huffman decoding.
Step4: Repeat 1-3 until end of text.
Figure 1. Huffman tree
Figure 3.Flow chart for extraction process
B. EMD Method
This method hides each digit of the Huffman codes into V. EXPERIMENTAL RESULTS
the segmented cover image by increasing or decreasing by
one gray-scale value at most for each group of n pixels [4]. An experiment is conducted using four standard gray-
Here, the value of n is crucial to determine the right size of the scale images of 512x512 pixels as cover images shown in Fig.
group. Therefore, (5) is used to calculate n which resulted in 4 namely, Baboon, Airplane, Lena and Tiffany. Furthermore,
only minimum distortion for the cover image. Then (2) is an arbitrary plain text is used as the secret message. To
used to transform each n-pixel group into a reference value f. evaluate the quality of stego-image, the following peak signal-
Embedding algorithm to-noise ratio (PSNR) is employed:
Input: cover image (g1, g2, …..,gn), digits from Huffman
encoding d.
Output: stego-image
Where MSE refers to the Mean Square Error between two
Step1: For i=1:n
pixels.
f=(sum(gi * i)) mod (2n+1)
end
© 2012 ACEEE 48
DOI: 01.IJIT.02.02. 34

4. ACEEE Int. J. on Information Technology, Vol. 02, No. 02, April 2012
Where r and c refer to the number of rows and columns
respectively for the cover image and stego-image.
As shown in Table 1, for the purpose of a benchmark test
between the proposed technique and various well-known
methods namely, OPAP, EMD and Opt EMD, two different
sizes of payloads are used. Thus, the experimental results are
divided into two parts: The 1st. part utilizes full payload,
whereas the 2nd. part uses half of it. The results have revealed
that our proposed method is superior than the rest of the
techniques in terms of both PSNR and payload viz. 60.31 dB
and 36000 byte, respectively. The difference in terms of PSNR
is quite significant between the proposed method and the 2nd
best method which is around 5 dB. As for the 2nd. part of the
results, although the payload is reduced to half, it is obvious
that the performance of the proposed method is maintained
in relations to both PSNR and payload. Therefore, the
proposed method is capable of embedding large amount of
secret messages without sacrificing image quality.
TABLE I.
THE RESULTS OF THE PROPOSED METHOD, OPAP, EMD AND OPT . EMD
Figure 4. Test images (a)~(d)
CONCLOSIONS
In this paper, a 2-tier data hiding technique using Huffman
coding and EMD method is proposed. The secret messages
are encrypted, compressed and embedded in cover images
with high secrecy without compromising the quality of the
stego-images. Experimental results have revealed that the
proposed method has successfully outweighed the well-
known methods in both PSNR and payload. The experiments
would be more meaningful if attacks were performed to
measure the robustness; however this is beyond our scope
of study at this juncture.
© 2012 ACEEE 49
DOI: 01.IJIT.02.02.34

5. ACEEE Int. J. on Information Technology, Vol. 02, No. 02, April 2012
REFERENCES [14] N. M. Nasrabadi, and R.A. King,” Image coding using vector
quantization: a review,” IEEE Trans Commun, vol. 36(8), pp.
[1] I. J. Cox, M. L. Miller, J. A. Bloom, J. Fridrich, and T. Kalker 957–971, 1988.
“Digital and watermarking”, ELSEVIER, USA, 2nd ed., pp. 4-9, [15] A. Munteanu, et al., “Wavelet image compression—the
2006. quadtree coding approach,” IEEE Trans Technol Biomed, vol. 3(3),
[2] A. A. Zaidan, et al. “Novel multi-cover steganography using pp. 176–185, 1999.
remote sensing image and general recursion neural cryptosystem”, [16] A. A. Shjul, and U. L. Kulkarni “secure skin tone based
International Journal of Physical Sciences, vol. 5(11), pp. 1776- steganography Using wavelet transform”, International Journal of
1786, 2010. computer theory and Engineering, vol. 3, pp. 16-22, February,
[3] C. Nedeljko, “Algorithm for Audio Watermarking and 2011.
Steganography,” Acta Universitatis Ouluensis. Series C., 2004. [17] B. Chen, S. Latifi, and J. Kanai, “Edge enhancement of remote
[4] X. Zhang, and S. Wang, “Efficient Steganographic Embedding image data in the DCT domain,” Image Vis Comput, vol. 17(12):
by Exploiting Modification Direction”, IEEE Communications pp. 913–921, 1999.
Letters, vol. 10(11), pp. 781-783, 2006. [18] A. M. Zeki, and A. A. Manaf, “A Novel Digital Watermarking
[5] K. Y. Lin, W. Hong, J. Chen, T. S. Chen and W. C. Chiang, Technique Based on ISB (Intermediate Significant Bit)”,
“Data hiding by Exploiting Modification Direction technique using International Journal of Information Technology, vol. 5, pp. 989-
Optimal Pixel grouping”, International Conference on Education 996, 2009.
Technology and Computer (ICETC), 2010, vol. 3, pp. 121-123. [19] Y. Dejun, Y. Rijing, Y. Yuhai, and X. Huijie, “Blind Digital
[6] E. Adelson., “Digital signal encoding and decoding apparatus”, Image Watermarking Technique Based On Intermediate Significant
US Patent, no. 4939515, 1990. Bit and Discrete Wavelet Transform”, Proc. of International
[7] L. F. Turner, “Digital data security system”, Patent IPN Conference on Computational Intelligence and Software Engineering,
WO 89/08915, 1989. CISE. IEEE Computer Soceity, 2009, vol. 15, pp. 1-4.
[8] W. Bender, D. Gruhl, M. Morimoto, and A. Lu, “Techniques [20] S. M. Perumal, and V. V. Kumar, “A Wavelet based Digital
for data hiding”, IBM Systems, vol. 35(3-4), pp. 313–336, 1996. Watermarking Method using Thresholds on Intermediate Bit
[9] C.K. Chan, L.M. Cheng, “Hiding data in images by simple Values”, International Journal of Computer Applications, Vol. 15(3),
LSB substitution”, Pattern Recognition, vol. 37(3), pp. 469–474, pp. 29-36, Feb 2011.
2004. [21] A. Habes, “Information Hiding in BMP image Implementation,
[10] H. Inoue, A. Miyazaki, and T. Katsura, “An Image Analysis and Evaluation”, Information Transmissions In Computer
Watermarking Method Based On the Wavelet Transform”, IEEE Networks, Tom. 6, No. 1, pp. 1-10, 2006.
Communications Letters, vol. 1, pp. 296-300. Aug 2002. [22] B. A. Mehemed, T. E. A. El-Tobely, M.M. Fahmy, M. E. L.
[11] C.C. Chang, J.Y. Hsiao, and C. S. Chan, “Finding optimal least Said Nasr, and M. H. A. El-Aziz, “Robust digital watermarking
significant- bit substitution in image hiding by dynamic based falling-off boundary in corners board-MSB-6 gray scale
programming strategy”, Pattern Recognition, vol. 36(7), pp. 1583– images,” International Journal of Computer Science and Network
1593. 2003. Security, Vol. 9(8), pp. 227-240, August 2009.
[12] W.C. Du, and W.J. Hsu, “Adaptive data hiding based on VQ [23] M. S. Emami, and G. B. Sulong, “A Statistical Method based
compressed Images,” IEEE Proc. of Vis Image Signal Process, 2003, on L2Norm Technique for EISB Information Watermarking
vol. 150(4), pp. 233–238. Scheme”, Proc. of International Conference on Future Information
[13] J. In, S. Hsiarani, and F. Kossentini, “On RD optimized Technology IPCSIT, Sep. 2011, Singapore, vol. 13, pp. 139-143.
progressive image coding using JPEG,” IEEE Trans Image Process, [24] K. Rabah, “Steganography — The Art of Hiding Data,”
vol. 8(11): pp. 1630–1638, 1999. Information Technology Journal, Vol. 3(3), pp. 245-269, 2004.
© 2012 ACEEE 50
DOI: 01.IJIT.02.02.34