- Email: [email protected]
Robust Method for Protecting Electronic Document on Waterway Transport with Steganographic Means by Embedding Digital Watermarks into Images
Available online at www.sciencedirect.com
ScienceDirect Procedia Engineering 178 (2017) 507 – 514
16thConference on Reliability and Statistics in Transportation and Communication, RelStat’2016, 19-22 October, 2016, Riga, Latvia
Robust Method for Protecting Electronic Document on Waterway Transport with Steganographic Means by Embedding Digital Watermarks into Images Maksim Bukharmetov, Anatoliy Nyrkov, Sergei Sokolov, Sergei Chernyi*, Vladimir Kuznetsov, David Mamunts Admiral Makarov State University of Maritime and Inland Shipping, Saint-Petersburg, Dvinskayast., 5/7,198035, Russia
Abstract Waterway transport, depending on the availability of safe controls, navigation and communication systems, is obliged to pay special attention to the technological updating of industry, the implementation of high-performance automation systems, the use of innovative technologies. To store and transfer such vast amounts of information relevant automation system is required, which allows making the entire procedure of processing the documents of various kinds safe. Thus, the development of information technologies in transport companies and, in particular, on ships is directly related to the establishment of information processing systems. The scope of this paper is the development and analysis of algorithms for implementation of the digital watermark (DWM) on the basis of the brightness modulation in blocks of graphic documents, allowing at the same time providing covert insertion of any sequence of a given amount of information and authentication of the image, in which Digital Watermark was incorporated. To enhance the robustness of embedded digital watermark attachments, it is required to apply the same transformations in compression of graphic documents in stegoalgorithm like in the compression algorithms for these files. © Published by Elsevier Ltd. Ltd. This is an open access article under the CC BY-NC-ND license © 2017 2017The TheAuthors. Authors. Published by Elsevier (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of the scientific committee of the 16th International Conference on Reliability and Statistics in Peer-review under responsibility of the scientific committee of the International Conference on Reliability and Statistics in Transportation and Communication. Transportation and Communication Keywords: steganography, water transport, data protection
* Corresponding author. E-mail address: [email protected]
1877-7058 © 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of the scientific committee of the International Conference on Reliability and Statistics in Transportation and Communication
doi:10.1016/j.proeng.2017.01.097
508
Maksim Bukharmetov et al. / Procedia Engineering 178 (2017) 507 – 514
1. Introduction The quest for maximum efficiency of water transport in the European waterways system, has led the Russian transport industry to the need of improving the safety and efficiency of transport connections (Nyrkov and Vikulin, 2011). Recently, in the field of water transport it is possible to observe significant changes aimed at improving the quality of transport processes, starting with the development of high-performance vessels’ traffic control algorithms and finishing the development of automated systems for monitoring the technical condition of offshore structures (Sutherland, 2016). For example, new capabilities for IT application in order to monitor the offshore structures include: • the collection and transmission of measurement results received from the primary devices by means of mobile devices, eliminating the process of computer processing; • collection of information obtained from the web cameras, to compare images and to analyze them further in order to assess the state of hydraulic structures; • “cloud computing” and the analysis of the results of measurements outside the offshore structures; • analysis of the data using a set of detection methods in the “raw” data previously unknown, but practically useful knowledge required to identify patterns of behaviour and constructing a model of the dynamics of change of properties; • creation of intelligent virtual models of hydraulic structures that represent real objects, completely described by software; • tracking the measurement results and results of forecasting the state of offshore structures in real time. At the same time, in view of escalating the dependence of water transport on information systems and services, threats to the information infrastructure are greater and, consequently, the protection of relevant information resources draws more attention. Decentralization of control of the information systems, due to the use of distributed computing systems, also contributes to the increase in the number of threats and reduction the level of information security (Song et al., 2000). The information security means the security of the stored and circulating information, and information infrastructure that supports it, preventing both accidental and intentional misrepresentation, which can cause enormous damage. Until recently, the information and the information resources have been subject to the protection. Today, however, the interaction between a human and an information resource becomes the object of the protection in most cases (Nyrkov et al., 2015). Considering the above, communication channels are required to be protected in modern systems, as well as data processing and storage systems with legitimate users access to the system information (Konakhovich, 2014). The purpose of this paper is to design the stenographic algorithms for information protection by embedding hidden information messages, allowing uniquely produce data authentication via secure communication channel. 2. Analysis of the steganographic methods of information protection In order to compare steganographic information protection techniques, let us introduce the following quality assessment indicators of their performance, divided them into the following groups of characteristics (Zhou, 2013; Qiang et al., 2016): • Indicators showing strength of the embedding method to detect the fact of hiding the data in the container. In the absence of the original container, data indicators may be determined by expert assessments. Otherwise, embedding method evaluation may be performed by means of quantitative, difference and correlation parameters. For example, peak signal-to-noise ratio may be used as such an indicator, showing the degree of difference of the original container from steganogram. This relationship is expressed in dB and is calculated from the following formula:
509
Maksim Bukharmetov et al. / Procedia Engineering 178 (2017) 507 – 514
PSNR =
m ⋅ n ⋅ max ( I i , j )
∑(I i, j
i, j
− Ci , j )
2
2
,
(1)
where m·n – is a container size; Ii,j – is a pixel of the initial image-container; Ci,j – is a steganogram pixel (image with the embedded data). • Indicators of capacity. Under the bandwidth method of concealing data, we mean the maximum amount of information that can be integrated into a single container element, provided that the requirements of the stability and secrecy were observed (Miano, 2005). This value for embedding steganographic method may be calculated based on the following formula:
⎛ τ2 ⎞ w = 0.5 ⋅ log2 ⎜1 + 2 w 2 ⎟ , ⎝ τ I +τ N ⎠
(2)
where τ w2 – is the capacity of the embedded message; τ I2 – is the capacity of the image-container; τ N2 – is the capacity of the noise while compressing. Table 1 lists the most popular and effective algorithms for steganography and shows resistance to various types of active attacks, based on indicators of quality assessment of the above (Jafar, 2016). From these data, it can be concluded that the following algorithms: Pitas, Cox, Barni, Meerwald have had the best indicators of resistance of the built-in data to active attacks on the container. 2.1. Algorithm Bruyndonckx Digital watermark (DWM) is a bit string. Implementation is carried out by modifying the brightness of the block of 8x8 pixels. The embedding process is made in three stages: 1. Classification or separation of the pixels within the block into two groups with approximately uniform luminosities. 2. Splitting of each group into categories, defined by a given grid. 3. Modification of the average luminance values for each category in each group. 2.2. Algorithm Pitas DWM is a two-dimensional array of bits with the size of the image, and the number of its units equals the number of zeros. There are several versions of the algorithm proposed by Pitas. Initially it was proposed to embed the DWM bit into each pixel of the image, but then it was wisely decided to use blocks of 2*2 or 3*3 pixels in size for this purpose. It makes the algorithm more robust to compression or filtering. The DWM is added to the image. In the case of using blocks for implementing a DWM detector calculates the average brightness value of the block. Hence, there is a possibility of non-uniform implementation of the DWM in pixels, that is, the value of const. Thus, it is possible to obtain the DWM optimized according to the criterion of robustness to JPEG compression algorithm procedure. For this purpose, the “capacity” of each pixel is calculated in the block of 8*8 elements. Then the DWM is implemented in accordance with the calculated capacity. This optimization is done once and for all, and 43 obtained mask is applied to any image. 2.3. Algorithm Koch It differs in that it encodes one bit of information in blocks of 8x8 pixels, but it is not resistant to the rotation of the image.
510
Maksim Bukharmetov et al. / Procedia Engineering 178 (2017) 507 – 514
2.4. Algorithm Cox The DWM is a sequence of pseudorandom numbers, distributed according to the Gaussian law, 1000 numbers long. To modify, 1000 of the largest coefficients of discrete cosine transform (DCT) are selected. The advantage of the algorithm is that by selecting the most significant coefficients, the watermark is more robust under compression and for other types of signal processing. Algorithm Barni uses the discrete cosine transform (DCT) of the entire marked image, being the advanced modification of the Cox’s method. Unlike the latter method, it uses the blind scheme of the detection of the embedded data and extracts the embedded information. Both algorithms provide for the introduction of information in a few AC-coefficients of the DCT for the entire protected image, so this method would not be resistant to framing. Table 1. Resistance of the stenographic algorithms to active attacks. Title of the method
Coherence of the recovery
Resistance to the geometric attacks
Resistance to compression
Resistance to the mean so statistical steganalysis
Bruyndonckx
+
-
-
+
Pitas
+
-
+
+
Koch
+
-
-
-
Cox
+
+
+
-
Barni
-
-
+
+
Wang
+
-
+
-
Meerwald
-
+
+
-
Media containers carrying lossy data compression are the most popular among the steganographic methods (Abuadbba, 2015). This is because the compression attack will not affect the embedded data. This robustness is provided by applying the same procedures as in the data compression algorithms. Consider the JPEG compression algorithm in more detail (Glumov, 2011; Na, 2008). The first stage is characterized by the transformation of the colour space. Habitual RGB colour space, with the components responsible for the red, green, and blue components of the colour Unlike the colour model RGB, the space YCbCr is a model brightness-colourfulness. This transformation is explained by the fact that the human eye is more sensitive to luminance signal than to colour. In this colour space, the luminance component is assessed as is a weighted average value of components of red, green and blue colours and is calculated by the following formula:
Y = kr R + kgG + kb B ,
(3)
where k is a relevant weight factor. The rest colour components may be calculated via the differences between components of luminance and colour components:
Cb = B − Y ; Cr = R − Y ;
(4)
Cg = G − Y . In this case, instead of the RGB three-dimensional space, we obtain a new four-dimensional one. We assume that the sum of the components Cb + Cr + Cgis a constant that gives the right to keep only two of the three chromatic components. The third component, if necessary, can always be calculated based on two others. Generally, Cb and Cr are used as the two desired colour components.
511
Maksim Bukharmetov et al. / Procedia Engineering 178 (2017) 507 – 514
Unlike luminance, chromatic components may be represented with a smaller resolution as the nervous system of the human eye is more sensitive to brightness than to colour. This fact can significantly reduce the amount of information stored for the transmission of chromatic components. This graphic document will not undergo significant deterioration in the quality of display colours. Thus, the compression algorithm executes at first the transformation of the original colour space to the luminance-chromaticity, compresses, and then performs the image reconstruction process, transforming the graphic document back to the RGB colour model. Direct (5) and reverse (6) conversion is calculated using the following formulas:
⎧ ⎪Y = k R + (1 − k − k )G + k B; r b r b ⎪ ⎪ 0,5 ( B − Y ); ⎨ Cb = − kb 1 ⎪ ⎪ 0,5 ( R − Y ). ⎪Cr = 1 − kr ⎩
(5)
⎧ 1 − kr ⎪ R = Y + 0,5 Cr ; ⎪ ⎪ 2kb (1 − kb ) 2k (1 − kr ) Cb − r Cr ; ⎨G = Y − 1 − kb − k r 1 − kb − k r ⎪ ⎪ 1 − kb Cb . ⎪B = Y + 0,5 ⎩
(6)
At the second phase, the luminance and chromatic components undergo the segmentation process, breaking into blocks of 8x8 pixels. Then for each colour space component YCbCr direct discrete cosine transformation is made:
DCT (i, j ) =
N −1 N −1 1 ⎡ (2 x + 1)iπ ⎤ ⎡ (2 x + 1) jπ ⎤ C (i )C ( j ) ∑∑ p( x, y )cos ⎢ cos ⎢ ⎥ ⎥⎦ ; 2N ⎣ 2N ⎦ ⎣ 2N x =0 y =0
⎧ 1 ⎪ , x = 0; C ( x) = ⎨ 2 ⎪1, x > 0. ⎩
(7)
Discrete cosine transformation is a spectral conversion allowing you to submit an image as a sum of sine waves of different amplitude and frequency of variations in brightness and shade. A significant part of the algorithms carries out a process of embedding of a given information sequence into the graphic document at this stage. It should be noted that JPEG is primarily a convenient format for graphic documents with smooth colour transitions and noises, such as digital photos (Kwok, 2005). In Figure 1 shows a plot of all the pairs of adjacent pixels in the graphic document. The x-axis represents the value of the first pixel, the vertical axis –of the second one. If not to transit to the second stage of the compression algorithm (Chernyi, 2016), and to use the dependence of brightness values of the surrounding pixels, it is possible to produce embedding of the predetermined information sequence into this setting. This will speed up the integration process, because it is no use going down to a lower level of the compression algorithm and computing the discrete cosine transformation (DCT) for all blocks in the graphic document.
512
Maksim Bukharmetov et al. / Procedia Engineering 178 (2017) 507 – 514
Fig. 1. Dependence of pairs of adjacent pixels in the image.
The proposed method is based on algorithm QIM (Quantization Index Modulation) with image pixels brightness modulation (Glumov, 2011). Data of Table 1 show that the algorithm has a low hiding of the data embedded. This is because when embedding the predetermined digital sequence, the algorithm cannot be lowered to a level at which the DCT is calculated. Thereby the advantage of the QIM algorithm is also a disadvantage at the same time. 3. Calculations All digital image of DWM is black and white (binary), respectively, analogy could be drawn and it allows to set for the white bit equal to “0”, and for the black –“1”. Suppose that there is need to implement a bit equal to “0” in a graphic digital image. It is necessary at the beginning to calculate the difference between indicators of the brightness pixel blocks in the image-source file. For convenience, it considers the case without using hash functions, and pseudo-random distribution of blocks, we are going to build sequence of blocks by their appearance in the image. To calculate the brightness values we will use the following formula:
Y = 0, 3R + 0, 59G + 0,11B , Where R, G, B– red, green and blue, respectively. Let us take two neighboring blocks 8 × 8 pixels of the digitized image and calculate the average of brightness for each of them. To implement the bit with a value of “0” it is necessary to have less average brightness value in the first block than in second. It is necessary to redistribute brightness. Duplication of the embedded hidden information, based on a pseudo-random distribution of blocks in a graphic document enhances the security of a digital watermark from external influences on geometric stegocontainer. At the same time, block-check of the information sequence by finding the same key elements allows to identify possible areas for modification of the image. As an illustration of the proposed algorithm and its properties, we will build in to the image-container DW with size of 25 × 10 pixels, and then we will subject image with an embedded DWM a series of geometric attacks and compression of the JPEG format (Figure 3).
Maksim Bukharmetov et al. / Procedia Engineering 178 (2017) 507 – 514 а)
b)
Fig. 2. Averages of brightness of two neighboring blocks: (a) - Y = 162.978; (b) - Y = 161.468.
Because of cropping block boundaries integration has shifted, it results in a shift of DWM. A digital sign has also lost 10% of the information, though the original image-container has lost about 40%. At such enormous modifications of the image, the digital sign remained well readable, and it also showed possible options of the modified areas. These results show that the DWM, built-in with the use of the developed algorithm, is resistant to a set of geometric transformations of image-container and to compression by JPEG format.
Fig. 3. The result of the extraction of DWM from the modified image: (a) - the image-container and built-in DWM; (b) - modified image and the recovered DWM; (c) - image-container modification fields.
4. Conclusion In this work, the new algorithm of block embedding of DWM having the following advantages compared to the existing methods is presented. The algorithm does not require the use of extra procedures of pre-processing (such as encryption) of the digital watermark image. The algorithm does not require the use of fixed digital watermark to detect modifications, which provides digital watermark resistance to so-called “attack with a fixed DWM” (watermark template attack, deliberate attack of an experienced intruder making possible to modify images, even in the presence of a built DWM). The algorithm performs faster by utilizing brightness modulation rather than calculating matrixes of discrete cosine transformation.
513
514
Maksim Bukharmetov et al. / Procedia Engineering 178 (2017) 507 – 514
Using the redistribution of brightness as a basis for withholding the information gives unlimited possibilities to improve the algorithm. One can use different algorithms for brightness distribution within the block for tighter implementation bits of the digital watermark and for correction of boundaries of blocks with a pronounced redistribution of brightness. The most common steganography algorithms to protect data by embedding an arbitrary information sequence of a given volume into the image containers were studied. Based on the evaluation of hiding and capacity of the embedding algorithms the four most stable steganographic algorithms were identified. Reduction in hiding of the embedded data was demonstrated. The influence of the level of embedding the information into a graphic document was shown. When using steganography algorithms operating at a higher level of the data compression, hiding of the embedded information sequence decreases.
References Abuadbba, A. and Khalil, I. (2015) Wavelet based steganographic technique to protect household confidential information and seal the transmitted smart grid readings. Information Systems, 53, 224–236. Chernyi, S. (2016) Use of Information Intelligent Components for the Analysis of Complex Processes of Marine Energy Systems.Transport and Telecommunication Journal, 17(3), 202–211. DOI:10.1515/ttj-2016-0018. Glumov, N.I. (2011) Algorithm for embedding of the semi-fragile digital water marks for the authentication tasks and hidden data transmission. Computer optics, 35(2), 262–264. Glumov N. I.and Mitekin V.A. (2011) Algorithm of embedding of digital water marks for tasks of authentication of images and the hidden information transfer. Computer optics, 2, 262–264. Jafar, I., Darabkh, K., Al-Zubi, R. and Saifan, R. (2016) An efficient reversible data hiding algorithm using two steganographic images. Signal Processing, 128, pp. 98–109. Konakhovich, G.F. (2014) Evaluating the effectiveness of methods of embedding steganographic information into the spectral area of the image. Automated control systems and automation devices, 168, 59–63. Kwok, T.andThao, N. (2005) An automatic electronic contract document signing system in a secure environment. Seventh IEEE International Conference on E-Commerce Technology (CEC'05), 497–502. Miano, G. (2005) Formats and algorithms of the image compression in action.TRIUMF, 330. Na, S. and Lee, S. (2008). Design of Security Mechanism for Electronic Document Repository System. 2008 International Conference on Convergence and Hybrid Information Technology. Nyrkov, A., Sokolov, S.and Belousov, A. (2015) Algorithmic Support of Optimization of Multicast Data Transmission in Networks with Dynamic Routing. Modern Applied Science, 9(5), 162–176. DOI: 10.5539/mas.v9n5p162. Nyrkov, A.P.andVikulin, P.V. (2011) Traffic Control Algorithm for vessels on crossing courses. Journal of the Water Communications University, 1, 100–105. Qiang, J., Chen, P., Ding, W., Xie, F. and Wu, X. (2016) Multi-document summarization using closed patterns. Knowledge-Based Systems, 99, 28–38. Song, W.W., Cheung, D.andTan, C.J. (2000) A semantic similarity approach to electronic document modeling and integration. Web Information Systems Engineering, 2000. Proceedings of the First International Conference, 1, 116-124. DOI:10.1109/WISE.2000.882382. Sutherland, C., Luxton-Reilly, A. and Plimmer, B. (2016) Freeform digital ink annotations in electronic documents: A systematic mapping study. Computers & Graphics, 55, 1–20. Zhou, J-J. (2013) Study on several confidentiality protection technologies for electronic document. Mechatronic Sciences, Electric Engineering and Computer, 2282–2285.
ScienceDirect Procedia Engineering 178 (2017) 507 – 514
16thConference on Reliability and Statistics in Transportation and Communication, RelStat’2016, 19-22 October, 2016, Riga, Latvia
Robust Method for Protecting Electronic Document on Waterway Transport with Steganographic Means by Embedding Digital Watermarks into Images Maksim Bukharmetov, Anatoliy Nyrkov, Sergei Sokolov, Sergei Chernyi*, Vladimir Kuznetsov, David Mamunts Admiral Makarov State University of Maritime and Inland Shipping, Saint-Petersburg, Dvinskayast., 5/7,198035, Russia
Abstract Waterway transport, depending on the availability of safe controls, navigation and communication systems, is obliged to pay special attention to the technological updating of industry, the implementation of high-performance automation systems, the use of innovative technologies. To store and transfer such vast amounts of information relevant automation system is required, which allows making the entire procedure of processing the documents of various kinds safe. Thus, the development of information technologies in transport companies and, in particular, on ships is directly related to the establishment of information processing systems. The scope of this paper is the development and analysis of algorithms for implementation of the digital watermark (DWM) on the basis of the brightness modulation in blocks of graphic documents, allowing at the same time providing covert insertion of any sequence of a given amount of information and authentication of the image, in which Digital Watermark was incorporated. To enhance the robustness of embedded digital watermark attachments, it is required to apply the same transformations in compression of graphic documents in stegoalgorithm like in the compression algorithms for these files. © Published by Elsevier Ltd. Ltd. This is an open access article under the CC BY-NC-ND license © 2017 2017The TheAuthors. Authors. Published by Elsevier (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of the scientific committee of the 16th International Conference on Reliability and Statistics in Peer-review under responsibility of the scientific committee of the International Conference on Reliability and Statistics in Transportation and Communication. Transportation and Communication Keywords: steganography, water transport, data protection
* Corresponding author. E-mail address: [email protected]
1877-7058 © 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of the scientific committee of the International Conference on Reliability and Statistics in Transportation and Communication
doi:10.1016/j.proeng.2017.01.097
508
Maksim Bukharmetov et al. / Procedia Engineering 178 (2017) 507 – 514
1. Introduction The quest for maximum efficiency of water transport in the European waterways system, has led the Russian transport industry to the need of improving the safety and efficiency of transport connections (Nyrkov and Vikulin, 2011). Recently, in the field of water transport it is possible to observe significant changes aimed at improving the quality of transport processes, starting with the development of high-performance vessels’ traffic control algorithms and finishing the development of automated systems for monitoring the technical condition of offshore structures (Sutherland, 2016). For example, new capabilities for IT application in order to monitor the offshore structures include: • the collection and transmission of measurement results received from the primary devices by means of mobile devices, eliminating the process of computer processing; • collection of information obtained from the web cameras, to compare images and to analyze them further in order to assess the state of hydraulic structures; • “cloud computing” and the analysis of the results of measurements outside the offshore structures; • analysis of the data using a set of detection methods in the “raw” data previously unknown, but practically useful knowledge required to identify patterns of behaviour and constructing a model of the dynamics of change of properties; • creation of intelligent virtual models of hydraulic structures that represent real objects, completely described by software; • tracking the measurement results and results of forecasting the state of offshore structures in real time. At the same time, in view of escalating the dependence of water transport on information systems and services, threats to the information infrastructure are greater and, consequently, the protection of relevant information resources draws more attention. Decentralization of control of the information systems, due to the use of distributed computing systems, also contributes to the increase in the number of threats and reduction the level of information security (Song et al., 2000). The information security means the security of the stored and circulating information, and information infrastructure that supports it, preventing both accidental and intentional misrepresentation, which can cause enormous damage. Until recently, the information and the information resources have been subject to the protection. Today, however, the interaction between a human and an information resource becomes the object of the protection in most cases (Nyrkov et al., 2015). Considering the above, communication channels are required to be protected in modern systems, as well as data processing and storage systems with legitimate users access to the system information (Konakhovich, 2014). The purpose of this paper is to design the stenographic algorithms for information protection by embedding hidden information messages, allowing uniquely produce data authentication via secure communication channel. 2. Analysis of the steganographic methods of information protection In order to compare steganographic information protection techniques, let us introduce the following quality assessment indicators of their performance, divided them into the following groups of characteristics (Zhou, 2013; Qiang et al., 2016): • Indicators showing strength of the embedding method to detect the fact of hiding the data in the container. In the absence of the original container, data indicators may be determined by expert assessments. Otherwise, embedding method evaluation may be performed by means of quantitative, difference and correlation parameters. For example, peak signal-to-noise ratio may be used as such an indicator, showing the degree of difference of the original container from steganogram. This relationship is expressed in dB and is calculated from the following formula:
509
Maksim Bukharmetov et al. / Procedia Engineering 178 (2017) 507 – 514
PSNR =
m ⋅ n ⋅ max ( I i , j )
∑(I i, j
i, j
− Ci , j )
2
2
,
(1)
where m·n – is a container size; Ii,j – is a pixel of the initial image-container; Ci,j – is a steganogram pixel (image with the embedded data). • Indicators of capacity. Under the bandwidth method of concealing data, we mean the maximum amount of information that can be integrated into a single container element, provided that the requirements of the stability and secrecy were observed (Miano, 2005). This value for embedding steganographic method may be calculated based on the following formula:
⎛ τ2 ⎞ w = 0.5 ⋅ log2 ⎜1 + 2 w 2 ⎟ , ⎝ τ I +τ N ⎠
(2)
where τ w2 – is the capacity of the embedded message; τ I2 – is the capacity of the image-container; τ N2 – is the capacity of the noise while compressing. Table 1 lists the most popular and effective algorithms for steganography and shows resistance to various types of active attacks, based on indicators of quality assessment of the above (Jafar, 2016). From these data, it can be concluded that the following algorithms: Pitas, Cox, Barni, Meerwald have had the best indicators of resistance of the built-in data to active attacks on the container. 2.1. Algorithm Bruyndonckx Digital watermark (DWM) is a bit string. Implementation is carried out by modifying the brightness of the block of 8x8 pixels. The embedding process is made in three stages: 1. Classification or separation of the pixels within the block into two groups with approximately uniform luminosities. 2. Splitting of each group into categories, defined by a given grid. 3. Modification of the average luminance values for each category in each group. 2.2. Algorithm Pitas DWM is a two-dimensional array of bits with the size of the image, and the number of its units equals the number of zeros. There are several versions of the algorithm proposed by Pitas. Initially it was proposed to embed the DWM bit into each pixel of the image, but then it was wisely decided to use blocks of 2*2 or 3*3 pixels in size for this purpose. It makes the algorithm more robust to compression or filtering. The DWM is added to the image. In the case of using blocks for implementing a DWM detector calculates the average brightness value of the block. Hence, there is a possibility of non-uniform implementation of the DWM in pixels, that is, the value of const. Thus, it is possible to obtain the DWM optimized according to the criterion of robustness to JPEG compression algorithm procedure. For this purpose, the “capacity” of each pixel is calculated in the block of 8*8 elements. Then the DWM is implemented in accordance with the calculated capacity. This optimization is done once and for all, and 43 obtained mask is applied to any image. 2.3. Algorithm Koch It differs in that it encodes one bit of information in blocks of 8x8 pixels, but it is not resistant to the rotation of the image.
510
Maksim Bukharmetov et al. / Procedia Engineering 178 (2017) 507 – 514
2.4. Algorithm Cox The DWM is a sequence of pseudorandom numbers, distributed according to the Gaussian law, 1000 numbers long. To modify, 1000 of the largest coefficients of discrete cosine transform (DCT) are selected. The advantage of the algorithm is that by selecting the most significant coefficients, the watermark is more robust under compression and for other types of signal processing. Algorithm Barni uses the discrete cosine transform (DCT) of the entire marked image, being the advanced modification of the Cox’s method. Unlike the latter method, it uses the blind scheme of the detection of the embedded data and extracts the embedded information. Both algorithms provide for the introduction of information in a few AC-coefficients of the DCT for the entire protected image, so this method would not be resistant to framing. Table 1. Resistance of the stenographic algorithms to active attacks. Title of the method
Coherence of the recovery
Resistance to the geometric attacks
Resistance to compression
Resistance to the mean so statistical steganalysis
Bruyndonckx
+
-
-
+
Pitas
+
-
+
+
Koch
+
-
-
-
Cox
+
+
+
-
Barni
-
-
+
+
Wang
+
-
+
-
Meerwald
-
+
+
-
Media containers carrying lossy data compression are the most popular among the steganographic methods (Abuadbba, 2015). This is because the compression attack will not affect the embedded data. This robustness is provided by applying the same procedures as in the data compression algorithms. Consider the JPEG compression algorithm in more detail (Glumov, 2011; Na, 2008). The first stage is characterized by the transformation of the colour space. Habitual RGB colour space, with the components responsible for the red, green, and blue components of the colour Unlike the colour model RGB, the space YCbCr is a model brightness-colourfulness. This transformation is explained by the fact that the human eye is more sensitive to luminance signal than to colour. In this colour space, the luminance component is assessed as is a weighted average value of components of red, green and blue colours and is calculated by the following formula:
Y = kr R + kgG + kb B ,
(3)
where k is a relevant weight factor. The rest colour components may be calculated via the differences between components of luminance and colour components:
Cb = B − Y ; Cr = R − Y ;
(4)
Cg = G − Y . In this case, instead of the RGB three-dimensional space, we obtain a new four-dimensional one. We assume that the sum of the components Cb + Cr + Cgis a constant that gives the right to keep only two of the three chromatic components. The third component, if necessary, can always be calculated based on two others. Generally, Cb and Cr are used as the two desired colour components.
511
Maksim Bukharmetov et al. / Procedia Engineering 178 (2017) 507 – 514
Unlike luminance, chromatic components may be represented with a smaller resolution as the nervous system of the human eye is more sensitive to brightness than to colour. This fact can significantly reduce the amount of information stored for the transmission of chromatic components. This graphic document will not undergo significant deterioration in the quality of display colours. Thus, the compression algorithm executes at first the transformation of the original colour space to the luminance-chromaticity, compresses, and then performs the image reconstruction process, transforming the graphic document back to the RGB colour model. Direct (5) and reverse (6) conversion is calculated using the following formulas:
⎧ ⎪Y = k R + (1 − k − k )G + k B; r b r b ⎪ ⎪ 0,5 ( B − Y ); ⎨ Cb = − kb 1 ⎪ ⎪ 0,5 ( R − Y ). ⎪Cr = 1 − kr ⎩
(5)
⎧ 1 − kr ⎪ R = Y + 0,5 Cr ; ⎪ ⎪ 2kb (1 − kb ) 2k (1 − kr ) Cb − r Cr ; ⎨G = Y − 1 − kb − k r 1 − kb − k r ⎪ ⎪ 1 − kb Cb . ⎪B = Y + 0,5 ⎩
(6)
At the second phase, the luminance and chromatic components undergo the segmentation process, breaking into blocks of 8x8 pixels. Then for each colour space component YCbCr direct discrete cosine transformation is made:
DCT (i, j ) =
N −1 N −1 1 ⎡ (2 x + 1)iπ ⎤ ⎡ (2 x + 1) jπ ⎤ C (i )C ( j ) ∑∑ p( x, y )cos ⎢ cos ⎢ ⎥ ⎥⎦ ; 2N ⎣ 2N ⎦ ⎣ 2N x =0 y =0
⎧ 1 ⎪ , x = 0; C ( x) = ⎨ 2 ⎪1, x > 0. ⎩
(7)
Discrete cosine transformation is a spectral conversion allowing you to submit an image as a sum of sine waves of different amplitude and frequency of variations in brightness and shade. A significant part of the algorithms carries out a process of embedding of a given information sequence into the graphic document at this stage. It should be noted that JPEG is primarily a convenient format for graphic documents with smooth colour transitions and noises, such as digital photos (Kwok, 2005). In Figure 1 shows a plot of all the pairs of adjacent pixels in the graphic document. The x-axis represents the value of the first pixel, the vertical axis –of the second one. If not to transit to the second stage of the compression algorithm (Chernyi, 2016), and to use the dependence of brightness values of the surrounding pixels, it is possible to produce embedding of the predetermined information sequence into this setting. This will speed up the integration process, because it is no use going down to a lower level of the compression algorithm and computing the discrete cosine transformation (DCT) for all blocks in the graphic document.
512
Maksim Bukharmetov et al. / Procedia Engineering 178 (2017) 507 – 514
Fig. 1. Dependence of pairs of adjacent pixels in the image.
The proposed method is based on algorithm QIM (Quantization Index Modulation) with image pixels brightness modulation (Glumov, 2011). Data of Table 1 show that the algorithm has a low hiding of the data embedded. This is because when embedding the predetermined digital sequence, the algorithm cannot be lowered to a level at which the DCT is calculated. Thereby the advantage of the QIM algorithm is also a disadvantage at the same time. 3. Calculations All digital image of DWM is black and white (binary), respectively, analogy could be drawn and it allows to set for the white bit equal to “0”, and for the black –“1”. Suppose that there is need to implement a bit equal to “0” in a graphic digital image. It is necessary at the beginning to calculate the difference between indicators of the brightness pixel blocks in the image-source file. For convenience, it considers the case without using hash functions, and pseudo-random distribution of blocks, we are going to build sequence of blocks by their appearance in the image. To calculate the brightness values we will use the following formula:
Y = 0, 3R + 0, 59G + 0,11B , Where R, G, B– red, green and blue, respectively. Let us take two neighboring blocks 8 × 8 pixels of the digitized image and calculate the average of brightness for each of them. To implement the bit with a value of “0” it is necessary to have less average brightness value in the first block than in second. It is necessary to redistribute brightness. Duplication of the embedded hidden information, based on a pseudo-random distribution of blocks in a graphic document enhances the security of a digital watermark from external influences on geometric stegocontainer. At the same time, block-check of the information sequence by finding the same key elements allows to identify possible areas for modification of the image. As an illustration of the proposed algorithm and its properties, we will build in to the image-container DW with size of 25 × 10 pixels, and then we will subject image with an embedded DWM a series of geometric attacks and compression of the JPEG format (Figure 3).
Maksim Bukharmetov et al. / Procedia Engineering 178 (2017) 507 – 514 а)
b)
Fig. 2. Averages of brightness of two neighboring blocks: (a) - Y = 162.978; (b) - Y = 161.468.
Because of cropping block boundaries integration has shifted, it results in a shift of DWM. A digital sign has also lost 10% of the information, though the original image-container has lost about 40%. At such enormous modifications of the image, the digital sign remained well readable, and it also showed possible options of the modified areas. These results show that the DWM, built-in with the use of the developed algorithm, is resistant to a set of geometric transformations of image-container and to compression by JPEG format.
Fig. 3. The result of the extraction of DWM from the modified image: (a) - the image-container and built-in DWM; (b) - modified image and the recovered DWM; (c) - image-container modification fields.
4. Conclusion In this work, the new algorithm of block embedding of DWM having the following advantages compared to the existing methods is presented. The algorithm does not require the use of extra procedures of pre-processing (such as encryption) of the digital watermark image. The algorithm does not require the use of fixed digital watermark to detect modifications, which provides digital watermark resistance to so-called “attack with a fixed DWM” (watermark template attack, deliberate attack of an experienced intruder making possible to modify images, even in the presence of a built DWM). The algorithm performs faster by utilizing brightness modulation rather than calculating matrixes of discrete cosine transformation.
513
514
Maksim Bukharmetov et al. / Procedia Engineering 178 (2017) 507 – 514
Using the redistribution of brightness as a basis for withholding the information gives unlimited possibilities to improve the algorithm. One can use different algorithms for brightness distribution within the block for tighter implementation bits of the digital watermark and for correction of boundaries of blocks with a pronounced redistribution of brightness. The most common steganography algorithms to protect data by embedding an arbitrary information sequence of a given volume into the image containers were studied. Based on the evaluation of hiding and capacity of the embedding algorithms the four most stable steganographic algorithms were identified. Reduction in hiding of the embedded data was demonstrated. The influence of the level of embedding the information into a graphic document was shown. When using steganography algorithms operating at a higher level of the data compression, hiding of the embedded information sequence decreases.
References Abuadbba, A. and Khalil, I. (2015) Wavelet based steganographic technique to protect household confidential information and seal the transmitted smart grid readings. Information Systems, 53, 224–236. Chernyi, S. (2016) Use of Information Intelligent Components for the Analysis of Complex Processes of Marine Energy Systems.Transport and Telecommunication Journal, 17(3), 202–211. DOI:10.1515/ttj-2016-0018. Glumov, N.I. (2011) Algorithm for embedding of the semi-fragile digital water marks for the authentication tasks and hidden data transmission. Computer optics, 35(2), 262–264. Glumov N. I.and Mitekin V.A. (2011) Algorithm of embedding of digital water marks for tasks of authentication of images and the hidden information transfer. Computer optics, 2, 262–264. Jafar, I., Darabkh, K., Al-Zubi, R. and Saifan, R. (2016) An efficient reversible data hiding algorithm using two steganographic images. Signal Processing, 128, pp. 98–109. Konakhovich, G.F. (2014) Evaluating the effectiveness of methods of embedding steganographic information into the spectral area of the image. Automated control systems and automation devices, 168, 59–63. Kwok, T.andThao, N. (2005) An automatic electronic contract document signing system in a secure environment. Seventh IEEE International Conference on E-Commerce Technology (CEC'05), 497–502. Miano, G. (2005) Formats and algorithms of the image compression in action.TRIUMF, 330. Na, S. and Lee, S. (2008). Design of Security Mechanism for Electronic Document Repository System. 2008 International Conference on Convergence and Hybrid Information Technology. Nyrkov, A., Sokolov, S.and Belousov, A. (2015) Algorithmic Support of Optimization of Multicast Data Transmission in Networks with Dynamic Routing. Modern Applied Science, 9(5), 162–176. DOI: 10.5539/mas.v9n5p162. Nyrkov, A.P.andVikulin, P.V. (2011) Traffic Control Algorithm for vessels on crossing courses. Journal of the Water Communications University, 1, 100–105. Qiang, J., Chen, P., Ding, W., Xie, F. and Wu, X. (2016) Multi-document summarization using closed patterns. Knowledge-Based Systems, 99, 28–38. Song, W.W., Cheung, D.andTan, C.J. (2000) A semantic similarity approach to electronic document modeling and integration. Web Information Systems Engineering, 2000. Proceedings of the First International Conference, 1, 116-124. DOI:10.1109/WISE.2000.882382. Sutherland, C., Luxton-Reilly, A. and Plimmer, B. (2016) Freeform digital ink annotations in electronic documents: A systematic mapping study. Computers & Graphics, 55, 1–20. Zhou, J-J. (2013) Study on several confidentiality protection technologies for electronic document. Mechatronic Sciences, Electric Engineering and Computer, 2282–2285.