Image Steganography Using Mid Position Value Technique
S Mukherjee1, S Roy2, Goutam Sanyal3
National Institute of Technology Durgapur, Durgapur, 713209, India
Abstract: This paper presents a steganographic approach of concealing the secret data so as to facilitate secure
communication. Arnold transformation has been imposed on the chosen cover image in the first stage. This results
in the scrambling of the data bits, thereby disrupting the normal pixel orientation. Thereafter, Mid Position Value
(MPV) technique is implemented to embed data bits from the secret image within the scrambled cover. Lastly,
inverse Arnold transformation is applied on the above image. This results in a descrambling operation, i.e.
reverting back the normal orientation. Henceforth the stego is generated. All the experimental results analyze the
outcome of the full methodology. For this purpose, several quantitative and qualitative benchmark analysis
pertaining to this approach have been made. All the results show that the imperceptibility, i.e. non- detectability of
secret data is well maintained. Also the payload is high with negligible distortion in the image quality.
Keywords: steganography, mid position value (MPV), peak signal to noise ratio (PSNR), similarity measure.
Abbreviations
MPV Mid Position Value
PSNR Peak Signal to Noise Ratio
LSB Least Significant Bit
GLM Gray Level Modification
MSE Mean Square Error
1
1. Introduction
Security issue is rising alarm around this modern developing globe. In this context cryptography (Yamaguchi et al,
2015) was developed as an encryption (Vreugdenhil et al, 2009) technique to secure the contents of the confidential
information. But often this is not sufficient and it becomes necessary to keep the very existence hidden. The
purpose of steganography (Nagpal et al, 2015) is actually fencing the specified concern. In this paper we have
proposed a technique to sustain the steganographic target. A scrambling transformation has been used on the cover
image in the beginning. This is done with the incorporation of the Arnold Transform (Elayan et al, 2016) on the
cover or host. A randomized distribution of the image pixels is generated, which serves as a carrier for the
confidential data bits. The proposed Mid Position Value (MPV) methodology facilitates the insertion mechanism of
the data bits. MPV technique speculates on the concept of the middle position and its respective integer values of
all the dwelling pixels. This concept is used to generate the keys, which guide the insertion pattern. With the
execution of the MPV procedure, pixel data bits get embedded in the shuffled form of cover. Lastly, application of
inverse Arnold transform gives the required stego-image (Mukherjee et al, 2015). This paper is subdivided into
several sections. Section 2 contains a brief discussion of some of the related works. Section 3 discloses the main
proposed approach with all the necessary algorithms. Section 4 comprises of the experimental results and
discussion. Section 5 infers the conclusion along with the future work.
2. Related works
2
Data hiding by LSB (Joshi et al, 2013) is one of the most simple and traditional methods. This approach works by
hiding data in the least significant bits (LSB) of the pixels. Significant though it is but still some changes in the
image may impair the embedded data. The method of Pixel Value Differencing (PVD) (Sanchetti et al, 2012),
proposed by Wu and Tsai is another effective steganographic approach. It forms pixel blocks from the cover and
alters the pixel difference in each pair of blocks for data insertion. Greater the difference more is the alteration
made. Based on PVD, another method of tri way of pixel value differencing is proposed by Chang et al. (2008).
This new methodology performed better in terms of payload (Lan et al, 2000) and PSNR computations. Gray Level
Modification (GLM) proposed by (Potder et al, 2004). is another technique which maps data by remodeling the
gray level values of the pixels residing in the image. Specifically the gray level values of selected pixels are
compared with the secret data bits. Some techniques of combining the PVD and GLM methods with the aim of
increasing the payload have also been proposed. One such approach is put forward by (Safarpour et al, 2016).
Another effective approach is proposed by (Ahmad T et al, 2009). Here, the cover at first is split into some blocks.
Later based on certain conditions, data bits are masked in the block edges.
3. Proposed work and algorithm
This section encircles the proposed approach in image medium. All the steps of embedding and extraction are
shown in Fig. 1 (a) and (b) respectively. In the sender side, the cover image is first scrambled by applying Arnold
Transformation over it. A chaotic representation of the cover is the resultant output. This scrambled representation
serves as a layer of security since the original pixel positions are scuffled before embedding. Bits from the secret
image are then inserted in the above transformed image. For this, we incorporate the proposed Mid Position Value
(MPV) technique. This methodology uses the concept of the middle position and its respective values for each of
the residing pixels. Further, resting on the specified ground, the key values are computed. Finally, private data bits
are embedded following a certain fashion on insertion (given in Table 1). Inverse Arnold Transformation gives
back the actual stego-image. In the receiving side, all the steps are executed in correct sequence so as to retrieve the
masked image.
The flow diagrams of the processes are shown next in fig. 1.
3
(a) (b)
Fig. 1. (a) The process in the Sender’s side (Embedding); (b) The process in the Receiver’s side (Extraction).
3. 1 Embedding
This stage efficiently masks the confidential data within a seemingly unimportant chosen cover or host.
• Input a cover
• Apply Arnold Transform on it
• Implement Mid Position Value (MPV) technique to embed data
• Incorporate inverse Arnold Transform to generate the Stego
Mid Position Value (MPV) technique (Sender)
a) Take an input image (say ‘img1’)
b) Trace the pixel positions (say ‘p(i,j)’) of ‘img1’ using eqn. 1
p(i,j)=(i-1)*m+j (1)
c) For each ‘p(i,j)’, estimate the number of digits (say ‘total_digits’) in it
d) Calculate the middle digit position (say ‘mid_position’) by the following eqn. 2
mid _ position= (total _ digits / 2) +1 (2)
e) Capture the residing integer value (say ‘mid_value’) in ‘mid_position’ of ‘p(i,j)’
f) Obtain key1 from eqn. 3
key1= p(i, j)mid _ value (3)
g) If key1 exceeds the total no of pixels (say ‘T’), then
4
key1 = key1%T (4)
h) Obtain a decimal (say ‘4_bit_dec’) with the last_4_bits of the pixel value at p(i,j) (i.e. ‘pix(key)’)
i) Perform key2=key1%(4_bit_dec_pix(key))+1 (5)
j) Insertion takes place in p(i,j) according to Table 1
Table 1. Conditional Table of Insertion/Extraction Strategies.
Estimated key1 Estimated key2 Insertion/Extraction Technique
Even (key1%2=0) Even (key2%2=0) Direct embedding/extraction of two bit secret data
Odd (key2%2!=0) Reverse embedding/extraction of two bit secret
Odd (key1%2!=0) Even (key2%2=0) Direct two bit complementary embedding/extraction
Odd (key2%2!=0) Reverse two bit complementary embedding/extraction
3. 2 Extraction
This stage extracts the masked data bits from the stego so as to retrieve the hidden image
Input the stego-image
Apply Arnold Transformation on it
Implement Mid Position Value (MPV) receiving technique to extract data bits
Finally, the secret image is retrieved
Mid Position Value (MPV) receiving technique
a) Take a stego image (say ‘img2’)
b) Repeat steps ‘b’ to ‘i’ from Mid Position Value (MPV) technique (Sender)
c) Extraction of bits takes place in p(i,j) according to Table 1
d) All the extracted data bits fill up individual eight bit pixel arrays, generating the inserted secret image
5
4. Experimental Results and Discussion
Cover Image Scrambled Cover Secret Image Intermediate Stego Stego Image
kImage Image
Fig. 2. The sequential process of the Embedding Phase
4. 1 Embedding Capacity
Embedding capacity (may also be referred to as payload) (Kanan et al, 2014) is the maximum limit up to which
data can be embedded within the carrier with no significant distortion.
6
Table 2. Comparison of Embedding Capacity
Image Image size PVD GLM Ahmed T et al. Safarpour et al. MPV
Lena 128x128 ** 2048 2493 3906 4096
256x256 ** 8192 10007 15500 16384
512x512 50960 32768 40017 58861 65536
Pepper 128x128 ** 2048 2443 3906 4096
256x256 35.00 37.20 47.50 38.97 45.66
512x512 40.97 34.00 52.50 39.12 45.84
(** For PVD method, all the images that were used, are of size 512x512)
Table 2 portrays the comparative study of payload or embedding capacity. Out of all the methods compared, the
MPV approach has the maximum capacity due to its 2 bit insertion strategy. Therefore, this serves to be an
advantage in the field of steganography.
4. 2 MSE and PSNR
Mean squared error, i.e. MSE (Mukherjee et al, 2017) can be calculated from the eqn. 6, where C is the cover
(consisting MxN pixels) and S is its generated stego.
M N
1
MSE= ∑ ∑
(M∗N ) i=1 j=1
[C ( ij )−S (ij)]
2
(6)
Peak Signal to Noise Ratio (PSNR) [15] is estimated from the obtained MSE values in accordance with eqn. 7.
10 log 10 2552
PSNR=
MSE
(7)
Table 3. Comparison of PSNR values
Image Image size PVD GLM Ahmed T et Safarpour et MPV
al. al.
Lena 128x128 36.20 30.50 44.30 41.02 44.55
256x256 35.00 33.20 46.80 40.80 44.67
512x512 41.79 35.50 55.00 40.05 44.88
7
Pepper 128x128 38.70 38.00 43.50 40.96 44.97
256x256 35.00 37.20 47.50 38.97 45.66
512x512 40.97 34.00 52.50 39.12 45.84
The above Table (Table 3) displays the PSNR values obtained for various methods with respect to various
images. The MPV methodology gives moderate PSNR values which signify that the imperceptibility is maintained.
For some cases the method proposed by Ahmad T et al. gives better PSNR results than MPV approach. But in most
of the cases, MPV outperforms the others.
. 4. 3 Similarity Measure
The function used for estimating the similarity measure (Ash et al, 2015) is computed in eqn. 8, where ‘ci’ is the
cover pixels, ‘si’ is the stego pixel, c and s are the mean values of the cover and stego respectively.
r=
∑ (c i−c)(si−s)
√ ∑ (c i−c )2 √∑ (si −s )2
(8)
Shown next in Figs. 3, 4 and 5 are the comparison of similarity measure for various methods with respect to
different images
Fig. 3. Comparison of Similarity Measure w.r.t Lena Image.
8
Fig. 4. Comparison of Similarity Measure w.r.t. Baboon Image.
Fig. 5. Comparison of Similarity Measure w.r.t. Pepper Image.
From the above figures, we can clearly see that for all the cases, the values obtained from the MPV technique lies
very close to 1, as compared to all others. This proves that the quality of the image is well maintained, i.e. the cover
images and their stegos are highly similar.
5. Conclusion and Future work
Steganography facilitates security for several legitimate purposes during communication (Yuksel et al, 2009). In
this paper, we have proposed a steganographic approach in image medium which masks the secret data bits that we
want to communicate without any third party intervention. Application of Arnold Transform on the host image
layers a level of security in the beginning of the procedure itself. The MPV technique follows a conditional strategy
while embedding of secret data bits. Thus, the overall security (Mukherjee et al, 2015) is endorsed. This
methodology promotes high embedding capacity of the carrier image (Singh et al, 2017). The experimental results
9
affirm that the generated stego is highly imperceptible (Mukherjee et al, 2015). Therefore, it does not attract the
attention of unwanted sources. In future, this work can be extended to accommodate more secret data bits
(Solemani et al, 2017) within the carrier medium. Alongside, there must be no chances of introducing any kind of
distortion (Vreugdenhil et al, 2009) within the generated stego.
References
A. Almohammad and G. Ghinea, "Stego image quality and the reliability of PSNR," in 2010 2nd International
Conference on Image Processing Theory, Tools and Applications, 2010, pp. 215-220: IEEE.
A. T. Al-Taani, A. M. J. I. J. o. C. Al-Issa. 2009 Information, "A novel steganographic method for gray-level
images,", Systems Science, and Engineering, vol. 3, no. 1.
S. Ash, S. Mukherjee, and G. Sanyal. 2015, "A dwt based steganographic method using prime first mapping
(pfm)," in 2015 Second International Conference on Advances in Computing and Communication Engineering, pp.
471-476: IEEE.
K.-C. Chang, C.-P. Chang, P. S. Huang, and T.-M. J. J. o. m. Tu. 2008, "A Novel Image Steganographic
Method Using Tri-way Pixel-Value Differencing," , Journal of multimedia, vol. 3, no. 2.
M. A. Elayan and M. O. Ahmad, 2016, "Digital watermarking scheme based on Arnold and anti-Arnold
transforms," in International Conference on Image and Signal Processing, pp. 317-327: Springer.
Joshi, R., Gagnani, L., & Pandey, S. (2013). Image steganography with LSB. International Journal of Advanced
Research in Computer Engineering & Technology (IJARCET), 2(1), 228-229.
Kanan, H. R., & Nazeri, B. (2014). “A novel image steganography scheme with high embedding capacity and
tunable visual image quality based on a genetic algorithm”. Expert systems with applications, 41(14), 6123-6130.
T.-H. Lan, M. F. Mansour, and A. H. Tewfik. 2000, "Robust high capacity data embedding," in Proceedings
2000 International Conference on Image Processing (Cat. No. 00CH37101) , vol. 1, pp. 581-584: IEEE.
S. Mukherjee, S. Ash, and G. Sanyal. 2015, "A novel image steganographic methodology by Power Modulus
Scrambling (PMS) with logistic mapping," in TENCON 2015-2015 IEEE Region 10 Conference, pp. 1-6: IEEE.
S. Mukherjee, S. Ash, and G. Sanyal. 2015, "A novel image steganographic approach by Pixel Position
Modulus Method (PPMM)," in 2015 2nd International Conference on Computing for Sustainable Global
Development (INDIACom), pp. 278-283: IEEE.
10
S. Mukherjee and G. Sanyal.2015, "A novel image steganographic technique using position power first mapping
(PPFM)," in 2015 IEEE International Conference on Research in Computational Intelligence and Communication
Networks (ICRCICN), pp. 406-410: IEEE.
Mukherjee, S., & Sanyal, G. (2017). Enhanced position power first mapping (PPFM) based image
steganography. International Journal of Computers and Applications, 39(2), 59-68.
K. D. Nagpal, D. J. I. J. o. R. Dabhade, I. T. i. 2015, "A Survey on Image Steganography & its Techniques in
Spatial & Frequency Domain,", Computing, and Communication, vol. 3, no. 2, pp. 776-779.
V. M. Potdar and E. Chang. 2004, "Grey level modification steganography for secret communication," in 2nd
IEEE International Conference on Industrial Informatics, 2004. INDIN'04. 2004, pp. 223-228: IEEE.
Safarpour, M., & Charmi, M. (2016). Capacity enlargement of the PVD steganography method using the GLM
technique. arXiv preprint arXiv:1601.00299.
Sanchetti, A. (2012). Pixel value differencing image steganography using secret key. International Journal of
Innovative Technology and Exploring Engineering, 2.
Singh, A. K., Kumar, B., Singh, S. K., Dave, M., Singh, V. K., Kumar, P., ... & Mohan, A. (2017). Guest
editorial: robust and secure data hiding techniques for telemedicine applications. Multimedia Tools and
Applications, 76(5), 7563-7573.
Soleymani, S. H., & Taherinia, A. H. (2017). High capacity image steganography on sparse message of scanned
document image (SMSDI). Multimedia Tools and Applications, 76(20), 20847-20867.
J. Vreugdenhil, K. Iverson, and R. S. Katti. 2009, "Image encyption using dynamic shuffling and XORing
processes," in 2009 IEEE International Symposium on Circuits and Systems , pp. 734-737: IEEE.
Yamaguchi, Y. (2015). Extended visual cryptography for continuous–tone images: effect of the optimum tone
mapping. International Journal of Information and Communication Technology, 7(1), 25-39.
Yuksel, M., Liu, X., & Erkip, E. (2009). A secure communication game with a relay helping the eavesdropper.
IEEE Transactions on Information Forensics and Security, 6(3), 818-830.
11
