=Paper=
{{Paper
|id=Vol-3826/short27
|storemode=property
|title=Enhancing the steganographic resistance of hidden information to active attacks (short paper)
|pdfUrl=https://ceur-ws.org/Vol-3826/short27.pdf
|volume=Vol-3826
|authors=Yurii Yaremchuk,Olha Saliieva,Vasyl Karpinets,Andrii Nikolaienko,Nataliia Kunanets
|dblpUrl=https://dblp.org/rec/conf/cpits/YaremchukSKNK24
}}
==Enhancing the steganographic resistance of hidden information to active attacks (short paper)==
https://ceur-ws.org/Vol-3826/short27.pdf
Enhancing the steganographic resistance of hidden
information to active attacks ⋆
Yurii Yaremchuk1,*,†, Olha Saliieva1,†, Vasyl Karpinets1,†, Andrii Nikolaienko1,†
and Nataliia Kunanets2,†
1
Vinnytsia National Technical University, 95 Khmelnytsky Highway str., 21000 Vinnytsia, Ukraine
2
Lviv Polytechnic National University, 12 Bandera str., 79013 Lviv, Ukraine
Abstract
With the rapid development of modern information technologies and systems, the number and complexity
of threats aimed at breaching the security of confidential information transmitted secretly in multimedia
files is increasing. At the same time, active attacks aimed at steganographic information protection systems
are of great importance, as they differ from passive attacks in that the attacker not only tries to detect
hidden information but can also modify the image to remove or distort it. Given the importance of this
issue, the paper proposes to increase the resistance of hidden information in images to color gamut-
changing attacks. To achieve this goal, the hiding algorithm has been improved based on the use of matrix
filters to find the most suitable image areas and hide bits of information in them by correlating the average
brightness of segments in blocks. The proposed algorithm includes many variable parameters, which makes
it possible to adapt it to a wide range of images. In addition, flexible parameter settings enable improved
information embedding and extraction accuracy. To validate the effectiveness of the enhanced algorithm,
color-change attacks were conducted on an image containing embedded hidden information. Various
steganalysis methods were also applied, including visual steganalysis, which reveals the least significant bit
of the image; RS steganalysis, which estimates the approximate size of the hidden data; and a method
analyzing the distribution of image elements on a plane to detect patterns, structures, or anomalies within
the image. The obtained results demonstrate the robustness of the proposed method of information hiding
to the selected steganalysis algorithms and to color gamut-changing attacks.
Keywords
steganography, concealment of information, steganalysis, active attack, color gamut 1
1. Introduction information using matrix filters and test it to prove its
resistance to steganalysis and color change attacks.
With the rapid advancement of technology and In recent years, numerous studies have been conducted
communications, it is becoming increasingly important to on methods of hiding data in images based on the use of
protect confidential information from unauthorized access. algorithms for finding acceptable zones for hiding
After all, transmitting a secret message over an unprotected information. For example, in their study [2], the authors
network channel poses a serious threat to its security. developed the Bald Eagle Search Optimal Pixel Selection
Among the methods for solving this problem are with Chaotic Encryption (BESOPS-CE) steganographic
steganographic methods, which involve the open transport image method based on the method of searching for optimal
of a secret message embedded in a container in advance [1]. pixels with chaotic encryption. The presented method
However, there are many attacks aimed at breaking into effectively hides the secret image in encrypted form under
steganographic systems, including attacks that focus on the cover image. In this paper [3], a steganographic scheme
changing the color gamut in images, which can lead to the based on graph wavelet transformation using graph signal
disclosure of hidden information, violation of its processing (GSP) is proposed, which improves the visual
confidentiality and integrity. Therefore, this paper presents quality of a stego image. The authors of the paper [4]
a study related to increasing the resistance of hidden presented a new steganographic approach in which the
information in images to these attacks. Thus, in this paper, algorithm of an extreme learning machine is modified to
based on the analysis of known steganographic methods, it create a supervised mathematical model. This algorithm is
is necessary to improve the algorithm for hiding first trained on a part of the image and then tested in
regression mode, which allows choosing the optimal place
CPITS-II 2024: Workshop on Cybersecurity Providing in Information 0000-0002-6303-7703 (Y. Yaremchuk);
and Telecommunication Systems II, October 26, 2024, Kyiv, Ukraine 0000-0003-2388-7321 (O. Saliieva);
∗
Corresponding author. 0000-0001-8148-2002 (V. Karpinets);
†
These authors contributed equally. 0009-0007-0178-9444 (A. Nikolaienko);
yurevyar@vntu.net (Y. Yaremchuk); 0000-0003-3007-2462 (N. Kunanets)
salieva8257@gmail.com (O. Saliieva); © 2024 Copyright for this paper by its authors. Use permitted under
Creative Commons License Attribution 4.0 International (CC BY 4.0).
karpinets@gmail.com (V. Karpinets);
andrey.nikolaienko.0@gmail.com (A. Nikolaienko);
nek.lviv@gmail.com (N. Kunanets)
CEUR
Workshop
ceur-ws.org
ISSN 1613-0073
350
Proceedings
to embed a message with the best values of the predicted The next stage of the algorithm is to divide the resulting matrix
evaluation metrics. In this research [5], a new method was 𝑀 and image 𝑃 into blocks of size 𝑛 х 𝑛, with the number of
proposed to increase the possibility of embedding secret blocks being 𝑟. For each block of the matrix 𝑀 a set of pixel
information in an image based on the edge region. The new indices is defined by 𝑁 , exceeding the threshold value. This
approach combines the Kenny and Pruitt edge detection process begins by calculating the average value between the
methods using a binary operation, and the secret message is blocks:
hidden using the least significant bit (LSB) method. 𝑟𝑚 (𝑥, 𝑦) + 𝑔𝑚 (𝑥, 𝑦)
An interesting idea is that of the authors of that paper ℎ (𝑥, 𝑦) = , (3)
2
[6], who proposed an algorithm for encrypting private 𝑖 = 1 … 𝑟, 𝑥 = 1 … 𝑛, 𝑦 = 1 … 𝑛,
images in combination with a new tent multi-dynamic where ℎ is ith block with the average value between the pixels
piecewise connected mapping lattice (TMDPCML) for of the blocks; 𝑟𝑚 is ith block from the red channel of the matrix
spatiotemporal chaotic systems. This algorithm extracts 𝑀; 𝑔𝑚 is ith block from the green channel of the matrix 𝑀; 𝑛
private image information and uses distributed nonlinear is the size of the block.
diffusion. Study [7] presents a method for detecting Next, the set of indices is defined as follows 𝑁 :
steganographic changes in images using convolutional 𝑁 = {(𝑥, 𝑦)|ℎ (𝑥, 𝑦) > 𝑇}, (4)
neural networks. In [8], a new stenographic swarm where 𝑇 is threshold value.
optimization technique with encryption for digital image
𝑁 = {(𝑥 , 𝑦 ), (𝑥 , 𝑦 ), … , (𝑥 , 𝑦 )}, 𝑘 = 1 … 𝑛 ;
protection, called CSOES-DIS, was developed. The proposed
𝑛 is the length of the ith variable 𝑁 .
model applies the method of double chaotic digital image
The next step is to split 𝑁 into indices corresponding to
encryption, and then the embedding process is
the lower and upper parts of the block:
implemented. The authors of this paper [9] proposed a new 𝑛
algorithm for encrypting and hiding triple images by 𝐷 = {𝑁 [𝑗]|𝑁 [𝑗][1] > , 𝑗 = 1 … 𝑛 }, (5)
2
combining a 2D chaotic system, compressive sensing (CS), 𝑛
and 3D discrete cosine transform (DCT). 𝑇 = {𝑁 [𝑗]|𝑁 [𝑗][1] ≤ , 𝑗 = 1 … 𝑛 }, (6)
2
Data hiding methods that use areas of rapid changes in where 𝐷 = {(𝑥 , 𝑦 ), (𝑥 , 𝑦 ), … , (𝑥 , 𝑦 )}, 𝑘 = 1 … 𝑑 —is
an image to hide information have many advantages, the set of lowercase indices of the part of the block from 𝑁 ;
including high capacity, adaptability to different types of 𝑇 = {(𝑥 , 𝑦 ), (𝑥 , 𝑦 ), … , (𝑥 , 𝑦 )}, 𝑘 = 1 … 𝑡 is the set of
images and data formats, and the ability to use both the
upper indices of the part of the block from 𝑁 ; 𝑁 [𝑗] is value
spatial and frequency characteristics of an image. However,
𝑥 , 𝑦 from іth block, and 𝑁 [𝑗] [1] is value of 𝑥 from the set
these methods are not very resistant to color gamut attacks,
which limits their effectiveness. In this regard, the paper of indices from 𝑁 іth block.
proposes an improvement that uses luminance correlation The last stage of preparation before embedding the
and matrix filters to effectively hide information in images information is to calculate the average brightness of the pixels
while providing a high level of protection. in the blue channel of the image 𝑃 for the pixels with the lower
and upper block indices:
2. Main body 𝑝𝑑 =
∑ 𝑏𝑝 (𝐷 [𝑗])
, (7)
𝑑
Most algorithms that use areas of rapid changes in the
image to hide information use the method of hiding ∑ 𝑏𝑝 (𝑇 [𝑗])
𝑝𝑡 = , (8)
information in one or two last bits of pixels of different 𝑡
channels. These methods are called LSB and 2-LSB according where 𝑝𝑑 is the average brightness of the lower part of the і-
to the number of least significant bits [10]. However, these th block by indices 𝐷 ; 𝑝𝑡 is the average brightness of the
algorithms cannot withstand color gamut attacks, so it is upper part of the іth block by indices 𝑇 ; 𝐷 [𝑗] is indices
worth improving an algorithm that will use luminance
𝑥 , 𝑦 of the lower part of the іth block; 𝑇 [𝑗] is indices
correlation to embed bits of information in the image, which
will help to withstand color gamut attacks by using 𝑥 , 𝑦 the upper part of the іth block; 𝑏𝑝 is a block from
luminance in areas that are most difficult to change. the blue image channel P, accordingly 𝑏𝑝 (𝐷 [𝑗]) and
To increase the resistance to the above attacks, it is 𝑏𝑝 (𝑇 [𝑗]) are pixels from the top and bottom part of і-th
proposed to use matrix filters to detect areas of sharp block for the specified indices 𝑇 [𝑗] and 𝐷 [𝑗].
transitions, which are expressed in the red and green Thus, the preparation stage for hiding information in
channels, while information is hidden in the blue channel. the image takes place.
The proposed algorithm applies a filter using the Next, the embedding process is examined. Assuming a
discrete Laplace operator [11], which can be represented in sequence of bits W exists, to hide one bit of information, the
two forms: average brightness of the pixels at the calculated indices in
𝛻 𝑓 = 𝑓(𝑖 + 1, 𝑗) + 𝑓(𝑖 − 1, 𝑗) + the lower and upper parts of the blue channel block of the
(1) image must satisfy the conditions for embedding the bit.
+𝑓(𝑖, 𝑗 + 1) + 𝑓(𝑖, 𝑗 − 1) − 4 ∗ 𝑓(𝑖, 𝑗),
0 1 0 Moreover, the smaller the embedding dimension, the more
1 −4 1 , (2) intensively the pixel brightness will change, and vice versa.
0 1 0 Example of conditions for hiding a single bit:
where 𝑖, 𝑗 are indices in the cell of the selected pixel to 𝑖𝑓 𝑤 = 1, (0 ∗ 𝛿) + 𝜀 ≤ 𝑝𝑑 ≤ (0 + 1) ∗ 𝛿 − 𝜀 ∨,
determine ∇ 𝑓.
(2 ∗ 𝛿) + 𝜀 ≤ 𝑝𝑑 ≤ (2 + 1) ∗ 𝛿 − 𝜀 ∨ … ∨,
351
(𝑐 ∗ 𝛿) + 𝜀 ≤ 𝑝𝑑 ≤ (𝑐 + 1) ∗ 𝛿 − 𝜀 ∧, brightness for these parts of the block is correlated and the
brightness value is calculated. This process is repeated
(1 ∗ 𝛿) + 𝜀 ≤ 𝑝𝑡 ≤ (1 + 1) ∗ 𝛿 − 𝜀 ∨, cyclically until the desired values are selected and the condition
(3 ∗ 𝛿) + 𝜀 ≤ 𝑝𝑡 ≤ ((3 + 1) ∗ 𝛿) − 𝜀 ∨ … ∨, is fully met.
(𝑣 ∗ 𝛿) + 𝜀 ≤ 𝑝𝑡 ≤ (𝑣 + 1) ∗ 𝛿 − 𝜀 ,
𝑣 = 1, 3, 5 … 𝑞 − 1, 𝑐 = 0, 2, 4 … 𝑞 − 2, 𝛿 = ,
where 𝑞 is embedding dimension; 𝜀 is brightness change
coefficient (0 < 𝜀 < ).
At the same time, the conditions for hiding the zero bit
will be as follows:
𝑖𝑓 𝑤 = 0, (0 ∗ 𝛿) + 𝜀 ≤ 𝑝𝑡 ≤ (0 + 1) ∗ 𝛿 − 𝜀 ∨,
(2 ∗ 𝛿) + 𝜀 ≤ 𝑝𝑡 ≤ (2 + 1) ∗ 𝛿 − 𝜀 ∨ … ∨,
(𝑐 ∗ 𝛿) + 𝜀 ≤ 𝑝𝑡 ≤ (𝑐 + 1) ∗ 𝛿 − 𝜀 ∧,
(1 ∗ 𝛿) + 𝜀 ≤ 𝑝𝑑 ≤ (1 + 1) ∗ 𝛿 − 𝜀 ∨,
(3 ∗ 𝛿) + 𝜀 ≤ 𝑝𝑑 ≤ ((3 + 1) ∗ 𝛿) − 𝜀 ∨ … ∨,
(𝑣 ∗ 𝛿) + 𝜀 ≤ 𝑝𝑑 ≤ (𝑣 + 1) ∗ 𝛿 − 𝜀 ,
𝑣 = 1, 3, 5 … 𝑞 − 1, 𝑐 = 0, 2, 4 … 𝑞 − 2, 𝛿 = .
If the condition for hiding a bit of information is not
met, the total brightness is correlated by increasing or
decreasing by a unit block of pixels whose indices belong to
the lower or upper threshold of the blue channel block. This
happens until the conditions for hiding the selected bit of
information are met. This is how information in the image
is hidden.
We will develop algorithms that implement an
improved method of hiding information.
Fig. 1 shows a general algorithm for embedding one bit
of information in one block. Figure 1: Flowchart with a general algorithm for hiding one
After determining the luminance values, a check is bit per block
performed to determine whether the bit has been embedded. If
this condition is not met, then the correlation value for the At the same time, the general algorithm for extracting a bit of
lower and upper parts of the block is calculated, then the information from a block is shown in Fig. 2.
Figure 2: Flowchart with a general algorithm for extracting a bit of information
352
3. Experimental research where 𝑛 is the number of equal bits in both messages; 𝑀 is
the size of the hidden message in bits.
The security of the proposed algorithm was analyzed by Value of 𝑚 indicates the success of extracting hidden
performing color change attacks on the image in which the information from the image.
hidden information was embedded. More than 100 different images were selected for the
To determine the effectiveness of the algorithm’s experiment, most of which are detailed full-color images
protection against such attacks, the values of 𝑚 , are with a sufficient number of contrast transitions and
calculated, reflecting the correspondence between the monotonous areas.
embedded hidden information 𝑤 and the extracted hidden Fig. 3 shows an example of one of the images after
information 𝑤 ∗ obtained: attacks that altered the color scheme: making it brighter,
𝑛
(9) more contrasted, and darker, closer to grayscale.
𝑚 = 100%,
𝑀
Figure 3: Images with hidden information after attacks: change of color scheme to brighter, more contrasty, and darker, to
shades of grey
Let’s examine the impact of color gamut attacks on images After analyzing the data obtained, it can be concluded that
with an embedded message (Table 1). when the color gamut shifts toward grey, the percentage of
lost data in the image increases; however, the algorithm
Table 1 remains quite resistant to this attack.
The result of calculating the value 𝑚 for an image that is Next, a visual method will be tested that shows the most
under attack significant bit of the image. The test result is shown in Fig. 4,
Attack 𝑚 with the results from the original image displayed on the
Change the color scheme to a brighter one 98,223 left and the modified image on the right.
Change the color scheme to a more contrasting and
92,164
darker one
Change the color scheme to shades of grey 84,857
Figure 4: Comparison of the least significant bits in the original and hidden images
After analyzing the data, it can be concluded that the least The following testing will be carried out using the RS
significant bits from the original image differ significantly steganalysis method, which is effective for detecting the
from those of the container image; however, the container presence of a message in an image and estimating the
does not exhibit significant distortions or artifacts that approximate size of the hidden data. The results of the
would indicate the presence of hidden data. method on the original image are shown in Fig. 5, and on
the modified image—in Fig. 6.
353
Figure 5: The result of steganalysis of the original image using the RS method
Figure 6: The result of steganalysis of a modified image using the RS method
Thus, minor changes in different image channels were parameters, another significant advantage of the improved
detected, but these values cannot indicate the presence or algorithm can be identified, namely the flexibility of
absence of hidden information in the container or the settings, which allows the algorithm to be adapted to
original. different types of images.
Let’s conduct additional testing of the algorithm, In addition, the study presents the algorithm of the
focusing on the analysis of the distribution of image program in the form of flowcharts, which is an important
elements on the plane (Fig. 7). step in testing the improved algorithm for resistance to
active attacks. The results of the testing indicate high
resistance to color gamut attacks on bright colors and
above-average resistance to grey gamut attacks. Also,
during the testing, resistance to steganalysis by visual
methods, RS method, and the method of analyzing the
distribution of the image on the plane was revealed.
Thus, based on the proposed improvement of the
algorithm, the steganographic resistance of hidden
information in images to active attacks was increased.
References
[1] S. Buchyk, et al., Improvement of Steganographic
Figure 7: The result of steganalysis using the method of Methods based on the Analysis of Image Color
analyzing the distribution of image elements on the plane Models, in: Workshop on Cybersecurity Providing in
Information and Telecommunication Systems, CPITS,
If the points are randomly distributed over the entire area, vol. 2923 (2021) 117–124.
the sequence elements are independent, which is typical for [2] A. Bahaddad, K. Almarhabi, S. Abdel-Khalek, Image
containers with embedded data. If the container is not filled, Steganography Technique based on Bald Eagle Search
then an uneven distribution of points on the field will be Optimal Pixel Selection with Chaotic Encryption,
observed. Since the container did not show a chaotic Alexandria Eng. J. 7(5) (2023) 41–54. doi:
arrangement of points, this method was unable to recognize 10.1016/j.aej.2023.05.051.
that the container contained hidden information. [3] V. K. Sharma, D. K. Srivastava, P. Mathur, Efficient
Thus, according to the results of the study, it can be Image Steganography using Graph Signal Processing,
assumed that the proposed method of hiding information is IET Image Process. 12(6) (2018) 1065–1071. doi:
resistant to the described steganalysis algorithms and to 10.1049/iet-ipr.2017.0965.
color change attacks. [4] H. A. Atee, et. al., Extreme Learning Machine based
Optimal Embedding Location Finder for Image
4. Conclusions Steganography, PLoS ONE 12(2) (2017). doi:
10.1371/journal.pone.0170329.
In this work, the information-hiding algorithm is improved
[5] N. A. Mohsin, H. A. Alameen, A Hybrid Method for
by using matrix filters to identify the most suitable areas in
Payload Enhancement in Image Steganography Based
the image and embed information bits into them. This is
on Edge Area Detection, Cybernetics and Information
achieved by varying the brightness of segments in image
Technologies 21(3) (2021) 97–107. doi: 10.2478/cait-
blocks, which are consistently changed to hide bits of
2021-0032.
information. The main feature of the proposed algorithm is
[6] X. Wang, J. Yang, A Privacy Image Encryption
its high resistance to color gamut attacks, achieved by
Algorithm based on Piecewise Coupled Map Lattice
embedding bits by changing the average brightness of the
selected hiding zones. Due to the large number of variable
354
with Multi Dynamic Coupling Coefficient, Inf. Sci. 569
(2021) 217–240. doi: 10.1016/j.ins.2021.04.013.
[7] D. Denysiuk, et. al., Method for Detecting
Steganographic Changes in Images using Machine
Learning, in: 13th International Conference on
Dependable Systems, Services and Technologies
(DESSERT) (2023) 1–6. doi:
10.1109/DESSERT61349.2023.10416453.
[8] A. A. Eshmawi, et al., Competitive Swarm
Optimization with Encryption based Steganography
for Digital Image Security, Comput. Mater. Contin.
72(2) (2022) 4173–4184. doi:
10.32604/cmc.2022.028008.
[9] X. Wang, C. Liu, D. Jiang, A Novel Triple-Image
Encryption and Hiding Algorithm based on Chaos,
Compressive Sensing and 3D DCT, Inf. Sci. 574 (2021)
505–527. doi: 10.1016/j.ins.2021.06.032.
[10] J. T. Tarigan, D. I. Putra, Implementation of
Steganography Modified Least Significant Bit using
the Columnar Transposition Cipher and Caesar
Cipher Algorithm in Image Insertion, Journal of
Physics: Conference Series 1898(1) (2021). doi:
10.1088/1742-6596/1898/1/012003.
[11] Discrete Laplacian, Medium (2024). URL:
https://hilbert-cantor.medium.com/discrete-
laplacian-8a5dde7ff001
355