Histogram-based multilayer reversible data hiding method for securing secret data

Bulletin of Electrical Engineering and Informatics
Vol. 8, No. 3, September 2019, pp. 1128~1134
ISSN: 2302-9285, DOI: 10.11591/eei.v8i3.1321  1128
Journal homepage: http://beei.org/index.php/EEI
Histogram-based multilayer reversible data hiding method for
securing secret data
Chaidir Chalaf Islamy, Tohari Ahmad
Department of Informatics, Institut Teknologi Sepuluh Nopember, Surabaya, Indonesia
Article Info ABSTRACT
Article history:
Received Sep 3, 2018
Revised Jan 1, 2019
Accepted Apr 5, 2019
In this modern age, data can be easily transferred within networks. This
condition has brought the data vulnerable; so they need protection at all
times. To minimize this threat, data hiding appears as one of the potential
methods to secure data. This protection is done by embedding the secret into
various types of data, such as an image. In this case, histogram shifting has
been proposed; however, the amount of secret and the respective stego image
are still challenging. In this research, we offer a method to improve its
performance by performing some steps, for example removing the shifting
process and employing multilayer embedding. Here, the embedding is done
directly to the peak of the histogram which has been generated by the cover.
The experimental results show that this proposed method has a better quality
of stego image than existing ones. So, it can be one of possible solutions to
protect sensitive data.
Keywords:
Data hiding
Data protection
Histogram
Information security
Copyright © 2019 Institute of Advanced Engineering and Science.
All rights reserved.
Corresponding Author:
Chaidir Chalaf Islamy,
Department of Informatics,
Institut Teknologi Sepuluh Nopember, Surabaya, Indonesia.
Email: chaidir31@gmail.com
1. INTRODUCTION
The use of digital media such as video, image, and audio plays a vital role in communication and
data transmission on the internet. However, this rapid development results in various problems, such as
insecurity of the transmitted data. The application of data protection is critical in many fields such as military
and medical data. In order to protect the data, some techniques have been presented. One of them is
cryptography that the protection is done by converting the data into an unrecognizable form.
The implementation of this method can be found in [1, 2]. Another method to secure data is steganography or
data hiding. In this algorithm, we can keep the data confidential without arousing suspicion that there is a
secret in it. Data hiding can be applied to various media, one of which is an image.
A suitable data hiding method should produce a stego image which is able to hold a large payload
and has a low distortion level [3-5]. The common problem in this research is how to balance those two
factors. According to [6], the more data being embedded in an image leads to decreasing the quality of the
stego data, which means that it has been much distorted. On the contrary, a stego image with less payload
produces less distortion. In this research, we would like to work on these problems by exploring the
histogram of the cover image. The bases and research of data hiding are provided in Section 2, while the
proposed method is in Section 3. Next, the experimental results and the respective analysis are presented in
Section 4; it is followed by the conclusion in Section 5.
2. DATA PROTECTION METHODS
Difference Expansion (DE) is one of the protection algorithms which is firstly proposed by Tian et
al. [7]. This method is the forerunner of the RDE (Reduced Difference Expansion) method [8]. Al Huti et

Bulletin of Electr Eng and Inf ISSN: 2302-9285 
Histogram-based multilayer reversible data hiding method for… (Chaidir Chalaf Islamy)
1129
al. [9] improve the capacity of payload data by investigating the size of the blocks of RDE. They expand the
size of pixel blocks to 4x4, intending to decrease the difference between pixels which brings higher capacity.
The experimental results show that there is an improvement regarding the capacity. Angreni and Ahmad [10]
seek to improve the quality of the image by avoiding the overflow and underflow problems in the RDE
method. As in other RDE-based methods, this process requires the difference of pixel values; however,
their algorithm employs a random value instead of a pixel for the paired in a block. Similarly, this difference
holds a bit for each virtual block. According to the experimental results, it is shown that their performance
is superior.
Another method that can be used to perform data embedding on the cover image is the interpolation.
In general, this method increases the size of the image resolution. For example, the method which is proposed
by Benhfid et al. [11] which develops bicubic interpolation methods, nearest neighbors, and bilinear. It has
been proved that their method is able to raise the capacity of data to store. In further research, the data hiding
method can be performed in the encryption domain. Data embedding is performed after the original image
has been encrypted, as shown in [4, 12, 13].
Another reliable approach is histogram shifting. The benefit of this method is that it has low image
distortion, so the quality of the stego image is relatively high. Firstly introduced by Ni et al. [14],
the histogram shifting method embeds secret data into the original image by manipulating the histogram.
Here, the pixel value in the histogram is shifted so that it creates a space that can be used as a place to embed
the secret. The process is done by scanning the pixel value in the original image to find the highest (the peak)
frequency and the lowest (the zero) pixel value. This condition determines the direction of the shifting:
left and right. If it is to the left, then the respective value is reduced; otherwise, it is added. Let the peak and
the zero points be and , respectively; and the secret bit is . The pixel value of the image before
shifting is and after shifting is with the position is and (see (1).
{ (1)
After the original image has been shifted, the process of embedding secret data starts. Secret data are
inserted in a space or pixel whose frequency is zero as a result of the shifting process. We take an example
that the peak pixel value is and the zero value is . The algorithm scans the image to find pixels whose
value is . If the secret bit is , then the value of the respective pixel is reduced to . This value
results in filling the space of 173 which previously is empty. However, if the secret bit is , then the
pixel value does not change and the process continues to find the next . This operation can be illustrated
in (2).
{ (2)
Pan et al. [15] improve the method in [14] by increasing the space to hide the secret data. They use
the peak point as the reference without the minimum point. Similarly, the process of shifting is done two
ways, left and right. Once the peak pixel value is found, the neighboring pixels of the left and right of the
peak are shifted. The shifting process produces two empty spaces to embed the secret data. Suppose the peak
is , the left neighbor of is and the right neighbor of is . We shift the left neighbor
of or to the left and right neighbor of or to the right. After that, the process scans
the image. If the pixel value is from to , then the respective value is added by one. If the pixel
value is then it is reduced by one as depicted in (3).
{ (3)
In the process of hiding the secret data, each bit is embedded in both empty spaces, and
. Again, scan the image, if a pixel with the value of is found and the secret bit is then the pixel
value of is added by one. If the pixel with the value of is found and the secret bit is , then the
pixel value is reduced by one as provided in (4). As in the previous example, if in the scanning process we
find the pixel with the value , then that pixel is added by one and the value has become . If the value
is then it is reduced by one to have .

 ISSN: 2302-9285
Bulletin of Electr Eng and Inf, Vol. 8, No. 3, September 2019 : 1128 – 1134
1130
{ (4)
Wu et al. [16] increase the brightness level of the original image by modifying its histogram.
This enhancement aims to improve both the quality of the stego and provide more spaces to hold the secret.
Chen et al. [17] revise the methods of Wu et al. [9] by determining the process of histogram shifting adaptive
to the distribution of histogram characteristics. Furthermore, they explore the pixel value ordering for
increasing the data capacity. Next, Prabowo and Ahmad [18] use Adaptive Pixel Value Grouping to protect
the secret data. Their method is also in the field of histogram-based embedding. Another example in applying
histogram modification using the residual histogram is shown in [19].
Those previously discussed histogram-based methods employ a shifting process to generate empty
space for embedding. In facts, this shifting process has an impact on the quality of the stego image. In this
paper, we enhance the stego image quality by further exploring the shifting of the histogram. Here,
the shifting is not done before the embedding process. So, the secret is directly inserted to additional
preprocessing of the cover image. In addition, we use multilayer embedding in the whole process.
3. RESEARCH METHOD
To protect the secret data, we use the same principle as the method of Ni et al. [14], that the
embedding is done in the specified histogram. In contrast to that algorithm, where secret data are embedded
in empty spaces, we embed them in the peak, without shifting pixel values. By assuming that the peak pixel
value of the original image is , and the right neighbor of is , we take an example as follows.
We have is ; so, is . Then, we scan the cover image pixel by pixel. If the pixel with the
value of is found and the secret to be embedded is 1, then the pixel value of is reduced
by one. The value of does not change if the secret data is . The illustration of data insertion is
provided in Figure 1 and the embedding process is elaborated in (5).
{ (5)
In order to extract the secret bits, we need information of the frequency of the original pixels. It is
because this proposed method does not provide a specific space for bit 1 as in [14] and [15]. The number of
bit 1 of the secret can be found by subtracting the frequency of the peak of the whole cover after being
embedded ( ) by that before embedding ). So, if the number of bit 1 of the secret is , then .
In the scanning process for extracting the secret, if we find a pixel while , then extract the secret of
bit 1; add the pixel value by 1, and substract by 1. In case we find in the scanning, bit 0 of the secret
is extracted. This process is provided in (6).
{ (6)
10
7
173
Pixel
Value
174 175 176 177
11
6
173
Pixel
Value
174 175 176 177
z(n)=1
11
6
z(n)=0
(a)
11
6
(a)

1131
173
Pixel
Value
174 175 176 177 173
Pixel
Value
174 175 176 177
11
6
173
Pixel
Value
174 175 176 177
z(n)=0
(b)
(a)
11
6
173
Pixel
Value
174 175 176 177
(b)
Figure 1. An example of the hiding process, (a) Embedding with =1, (b) Embedding with =0
3.1. Multilayer embedding
To increase the size of data that can be hidden in the original image we implement the concept of
multilayer embedding. Here, the resulted stego image of the embedding process (layer) is split into non-
overlapping blocks. This method can secure the embedded image, so the public may not be aware of the
resulted stego image [15].
In the first layer, we divide the original image into blocks of , with is the half of the width
and height of the image. The data embedding process is applied to each block using (7) where is the
image of each block and is the block of the embedded secret data. Then, the blocks are merged to obtain
an image with the original size. In the next layer, we separate the image that has been embedded in the first
layer. In this layer, the block size is . Each block is also embedded with secret data like the process
in the first layer and then merged to have one image. In the third layer, the same process is carried out as in
the previous layers; however, the block size is . Overall, this multilayer embedding is shown in
Figure 2.
Cover Image Blocks of size
S x S
Embed
Secret Data
Embed
Secret Data
Blocks of size
S x (2S)
Blocks of size
(2S) x S
Embed
Secret Data
Layer 1 Layer 2 Layer 3
Figure 2. Multilayer embedding
3.2. Embedding process
Because we use multilayer embedding, we split each layer into non-overlapping blocks, where the
process of embedding data is done in those blocks. We do not need the minimum point because no pixel
value should be shifted. Let the peak value of a block be and its right neighbor be . If in the
scanning process a pixel whose value is is found and the secret bit is 0, then the value of the
respective pixel does not change. On the contrary, if then it is reduced by 1. By assuming that is
the pixel value of the original image at position is and is the pixel value of the stego image at
position , that process can be described in (7).
{ (7)

 ISSN: 2302-9285
1132
3.3. Extraction process on image blocks
As previously explained, before performing the data embedding, the peak value of the original
image is stored for the extraction process. Supposed the peak value of a block in the cover image is and
that of stego image is . Equivalent to the processing the whole cover image, in this block level, the value
, is obtained from the difference between and .
In [14] and [15], the number of bit 1 of the secret is already known because both methods provide an
empty space for the bit 1. In this proposed method, we do not allocate it, so the difference of and is
required. In general, this process can be described in (8).
{ (8)
4. RESULTS AND ANALYSIS
For the experiment, we use the images of ''Baboon'', ''Lena'', “Pepper” which taken from [20] and
medical images: “Abdominal”, “Chest” and “Hand” from [21]. All those images are 512x512 pixels in size.
The hardware we use is a PC with Ryzen 5 1400 CPU, 8GB RAM, Nvidia GTX 1050Ti GPU and 1TB
7200RPM storage media.
In this evaluation, we compare our method with [15] and [14], regarding the quality of stego image
and the size of secret data that can be put into the original image. In each layer, we test the amount of data
which can be embedded in the original image for both our and the compared methods whose results are
provided in Tables 1, 2 and 3.
Table 1. PSNR value of stego images (layer 1)
Cover Image
PSNR Payload
Size
Proposed Ni et al. [14] Pan et al. [15]
Pepper 67,5668 50,6892 48,2243 2 KB
2 KB
2 KB
10 KB
10 KB
20 KB
Baboon 68,8118 52,3366 48,2006
Lena 66,7737 50,1295 48,2438
Abdominal 66,0796 59,6239 49,7144
Chest 65,2988 51,1278 50,3845
Hand 62,9465 48,8758 48,5629
Cover Image
PSNR Payload
Size
Pepper 66,3338 45,8150 43,0364 4 KB
4 KB
4 KB
10 KB
10 KB
20 KB
Baboon 65,4134 46,3737 42,2979
Lena 65,7189 46,0483 43,7989
Abdominal 49,1067 48,7216 42,6768
Chest 50,5732 45,0127 43,3652
Hand 48,9795 42,3795 41,8346
Cover Image
PSNR Payload
Size
Pepper 66,5355 42,8267 40,0172 4 KB
4 KB
4 KB
10 KB
10 KB
20 KB
Baboon 67,0826 43,7545 39,8353
Lena 64,8506 42,9889 40,2543
Abdominal 49,0925 48,6078 39,7364
Chest 50,5479 42,6293 40,2899
Hand 48,9387 39,2916 38,7742
4.1. Quality of stego image
The results of the experiment in Tables 1, 2 and 3 show the respective quality of the stego images
that is measured by Peak Signal to Noise Ratio (PSNR) from Layer 1 to Layer 3, respectively. As predicted,
increasing the number of layers results in decreasing the quality. However, the number of bits raises since
more layers hold more data. It is found that in our proposed method, the drop is not as significant as that of
Ni et al. [14] and Pan et al. [15], especially for general images.

1133
For medical images, the decrease of ours is slightly higher than the other two methods.
The comparison of the embedded medical image shows that our proposed method is still superior in
preserving stego image quality. However, our method has decreased dramatically in the second layer. Then,
in the third layer, the quality of stego images of our method has decreased but not as significant as that from
layer one to layer two. In general, our generated stego images have higher PSNR value than others. This
experiment shows that the method we propose is able to keep the quality of the stego image effectively.
In all three layers, our proposed method is still able to achieve more than 60 dB of PSNR for general
images (Baboon, Lena and Pepper) and more than 45 dB on medical images. In Ni et al. [14] and
Pan et al. [15], before the data embedding, there is a process of shifting the pixel values of the original image.
This process has a significant impact on stego image quality. As previously discussed, in this proposed
method, we do not shift the original pixels. An example of the cover image and its stego images are provided
in Figure 3.
(a) (b) (c) (d)
(e) (f) (g) (h)
Figure 3. Cover and stego images, (a) Original Baboon image, (b) Layer 3 embedding with the proposed
method, (c) Layer 3 embedding with Ni et al. [14], (d) Layer 3 embedding with Pan et al. [15],
(e) Original Abdominal Image, (f) Layer 3 embedding with the proposed method,
(g) Layer 3 embedding with Ni et al. [14] (h) Layer 3 embedding with Pan et al. [15]
4.2. Capacity of stego image
The data capacity that can be accommodated in the original image is shown in Table 4. The results
we obtained at the maximum capacity test show that the method we proposed has the lowest capacity in all of
the original images we tested. In Pepper, Baboon and Lena, the method of Pan et al. [15] have the highest
capacity. It is because their method uses the peak value of the left and right neighbors of the image as a space
for embedding secret data. Whereas in medical images, the method of Ni et al. [14] is able to hold the most
secret data. In the medical image, the value of the peak is relatively high and its difference from the
neighboring pixel is significant. On the other hand, our method and Pan et al. [15] do not use the peak pixel
value as the location to store the secret data. So, the capacity that can be accommodated is lower.
Table 4. Total Capacity of the embedded secret data from layer 1 to layer 3
Cover Image
Capacity (Bit(s))
Pepper 6037 10295 13920
Baboon 6440 8519 13942
Lena 7047 11953 18642
Abdominal 9966 198906 13040
Chest 11302 276511 16090
Hand 30314 39959 54839

 ISSN: 2302-9285
1134
Multilayer embedding plays a primary role in increasing the capacity of secret data. Apparently,
the increase in capacity also followed by some quality lost. However, this trade-off is still worthy because the
decrease is still in the tolerable range.
5. CONCLUSION
In this paper, we have proposed a variation of data hiding methods base on the histogram shifting.
Different from the previous research, we do not shift pixel values; instead, we directly embed the secret. It is
shown that this method is able to raise the quality of the stego image. In order to compensate the capacity, we
explore the multilayer embedding. Overall, there is a trade-off between the capacity and the quality. In facts,
we may determine whether the focus is on the capacity or quality. Additionally, we may specify the
balancing point between those two factors.mIn the future, this research can be further extended to have a
higher quality level. Applying histogram shifting only in the specific value, is a possible way to do.
Hopefully, this can maintain the capacity of the embedding space.
REFERENCES
[1] E. R. Arboleda, J. L. Balaba, and J. C. L. Espineli, “Chaotic rivest-shamir-adlerman algorithm with data encryption
standard scheduling,” Bulletin of Electrical Engineering and Informatics (BEEI), vol. 6, no. 3, pp. 219-227. 2017.
[2] A. O. Mulani and P. B. Mane, “Watermarking and cryptography based image authentication on reconfigurable
platform,” Bulletin of Electrical Engineering and Informatics (BEEI), vol. 6, no.2, pp. 181-187. 2017.
[3] C. Qin, W. Zhang, F. Cao, X. Zhang, and C.-C. Chang, “Separable reversible data hiding in encrypted images via
adaptive embedding strategy with block selection,” Signal Processing, vol. 153, pp. 109–122, 2018.
[4] M. Li and Y. Li, “Histogram shifting in encrypted images with public key cryptosystem for reversible data hiding,”
Signal Processing, vol. 130, pp. 190–196, 2017.
[5] C. Kim, D. Shin, L. Leng, and C.-N. Yang, “Separable reversible data hiding in encrypted halftone image,”
Displays, vol. 55, pp. 71–79, 2018.
[6] M. Manju and V. Kavitha, “Survey on reversible data hiding techniques,” Int. J. Secur. its Appl., vol. 8, no. 4,
pp. 297–306, 2014.
[7] Jun Tian, "Reversible data embedding using a difference expansion," in IEEE Transactions on Circuits and Systems
for Video Technology, vol. 13, no. 8, pp. 890-896, Aug. 2003.
[8] A. M. Alattar, "Reversible watermark using the difference expansion of a generalized integer transform," in IEEE
Transactions on Image Processing, vol. 13, no. 8, pp. 1147-1156, Aug. 2004.
[9] M. H. A. Al Huti, T. Ahmad, and S. Djanali, “Increasing the capacity of the secret data using DEpixels blocks and
adjusted RDE-based on grayscale images,” in Proceedings of 2015 International Conference on Information and
Communication Technology and Systems, ICTS 2015, pp. 225–230. 2016.
[10] D. S. Angreni and T. Ahmad, “Enhancing DE-based data hiding method by controlling the expansion,”
in Proceedings of 2016 4th International Conference on Cyber and IT Service Management, CITSM 2016, 2016.
[11] A. Benhfid, E. bachir Ameur, and Y. Taouil, “High capacity data hiding methods based on spline interpolation,”
in 2016 5th International Conference on Multimedia Computing and Systems (ICMCS), pp. 157–162. 2016.
[12] X. Wu, J. Weng, and W. Q. Yan, “Adopting secret sharing for reversible data hiding in encrypted images,” Signal
Processing, vol. 143, pp. 269–281, 2018.
[13] S. Agrawal and M. Kumar, “Mean value based reversible data hiding in encrypted images,” Optik (Stuttg).,
vol. 130, pp. 922–934, 2017.
[14] Zhicheng Ni, Yun-Qing Shi, N. Ansari and Wei Su, "Reversible data hiding," in IEEE Transactions on Circuits and
Systems for Video Technology, vol. 16, no. 3, pp. 354-362, March 2006.
[15] Z. Pan, S. Hu, X. Ma, and L. Wang, “Reversible data hiding based on local histogram shifting with multilayer
embedding,” J. Vis. Commun. Image Represent., vol. 31, pp. 64–74, 2015.
[16] H. Wu, J. Dugelay and Y. Shi, "Reversible Image Data Hiding with Contrast Enhancement," in IEEE Signal
Processing Letters, vol. 22, no. 1, pp. 81-85, Jan. 2015.
[17] H. Chen, J. Ni, W. Hong, and T. S. Chen, “Reversible data hiding with contrast enhancement using adaptive
histogram shifting and pixel value ordering,” Signal Process. Image Commun., vol. 46, pp. 1–16, 2016.
[18] H. Eko Prabowo and T. Ahmad, “Adaptive Pixel Value Grouping for Protecting Secret Data in Public Computer
NetworksNo Title,” J. Commun., vol. 13, no. 6, pp. 325–332, 2018.
[19] I. C. Chang, Y. C. Hu, W. L. Chen, and C. C. Lo, “High capacity reversible data hiding scheme based on residual
histogram shifting for block truncation coding,” Signal Processing, vol. 108, pp. 376–388, 2015.
[20] “SIPI Image Database.” [Online]. Available: http://sipi.usc.edu/database/database.php?volume=misc. [Accessed:
28-Apr-2018].
[21] “eMicrobes Digital Library.” [Online]. Available: https://www.idimages.org/. [Accessed: 28-Apr-2018].

Histogram-based multilayer reversible data hiding method for securing secret data

More Related Content

What's hot (19)

Similar to Histogram-based multilayer reversible data hiding method for securing secret data (20)

More from journalBEEI (20)

Recently uploaded (20)

Histogram-based multilayer reversible data hiding method for securing secret data