=Paper=
{{Paper
|id=Vol-2698/paper20
|storemode=property
|title=Hiding Information in a Picture File: a System Model
with Experimental Design
|pdfUrl=https://ceur-ws.org/Vol-2698/p20.pdf
|volume=Vol-2698
|authors=Jakub Siłka
|dblpUrl=https://dblp.org/rec/conf/ivus/Silka20
}}
==Hiding Information in a Picture File: a System Model
with Experimental Design==
https://ceur-ws.org/Vol-2698/p20.pdf
Hiding Information in a Picture File: a System Model
with Experimental Design
Jakub Siłkaa
a Faculty of Applied Mathematics, Silesian University of Technology Kaszubska 23, 44-100 Gliwice, Poland
Abstract
The format of information is changing with the developments of computer methods and formats. Graphical information can
be also used to transfer text. In this paper, I am going to describe the new approach via merged text implemented directly
to the image by using alpha channel of the pictures. This kind of encryption can be also used as the opportunity to spread
script guidelines, which subsequently enable assuming control of entire computer. The paper presents a discussion and
experimental set up of the designed solution.
Keywords
Picture, hiding, Data, Invigilation, Instruction
1. Introduction
Nowadays information processing is possible in vari-
ous different levels. We can find many systems which
proceed information from users to get the knowledge
about several aspects. One of the most popular infor-
mation format is an image. We can have photos from
holidays, work, travel or many other occasions. This
type of information is easy to read and possible to use Figure 1: Sample merge scheme when the text code is sim-
in various devices. From images we can have informa- ply inserted into the image by third program to compose an
tion about objects visible there by analyzing pixels of output file.
the image or simply all color channels. In [1] was pre-
sented how pixels can formulate shapes of lung dis-
eases for estimation by artificial intelligence. In [2] images as carriers of information, also text which can
was discussed that pixels from voice spectra can be be visible to computer systems. This possibility gives
recognized as image segments centroids for use verifi- many ways to use it also in bad way, ie. when the text
cation processes, while [3] presented that comparing is a code which can be a virus.
colour aspects of pixels from microscopy images will This paper presents a system model which can use
help on detection of bacteria. graphical file as information transmission, both in good
Other field of working with the information hidden and bad way. The idea presented in this paper is to use
in images is steganography. This field of computer an alpha channel, which is responsible for the resolu-
science is oriented on developments of algorithms for tion level - the difference between the value of 255 and
transferring some hidden information in images. In 254 is imperceptible without special program. The HD
[4] was discussed how permutation on pixels opera- picture contains 921600 pixels (1280x720), so if we use
tions can improve security of transferred information. all of them we get as many as 112,5 kB place to gather
In [5] was given a wide discussion on trends and re- data. Because of that, there is a broad scope of op-
cent advances in steganography. In [6] a discussion portunity to implement variety of scripts there. If the
on these innovative aspects was moved to the domain program is encoded properly, it will host instructions
of coverless images, while in [7] the opposite of cover and send them out with coming pictures. The model
selection was presented. There are many ways to use presented in this paper shows how to use alpha chan-
nel for information buffer. The system is composed
IVUS 2020: Information Society and University Studies, 23 April 2020, to use an input image, merge it with text and forward
KTU Santaka Valley, Kaunas, Lithuania to second part which decomposes it. In the model are
" kubasilka@gmail.com (J. Siłka) discussed different aspects filtering the image to adjust
© 2020 Copyright for this paper by its authors. Use permitted under Creative information in the channel [8, 9].
Commons License Attribution 4.0 International (CC BY 4.0).
CEUR
Workshop
Proceedings
http://ceur-ws.org
ISSN 1613-0073 CEUR Workshop Proceedings (CEUR-WS.org)
Figure 2: On the left side we can see photos before adding coded message (A) - white pixels stand for the value 254 instead
of 255. Below is displayed the message (B), its appearance coded in alpha channel (C) and image after adding the last,
modified layer (D).
2. Proposed data processing
Now we can discuss how the idea of code encryption
into the graphic files works in practice. The system
composed for the research has an experimental set up
which is described below.
Transmitter Transmitter is the program that codes
guidelines for the receiver, downloads picture and links
text file from proper folders and checks whether the
pictures have correct format. If not, the program would
turn it into the demanded one. Sample scheme is vis-
ible in Fig. 1. Afterwards, the predefined statement
is encoded, whereby program has already determined Figure 3: Sample decoding scheme from the receiver side,
format and ending of the instruction. Therefore it is where the code is reformatted in the file and verified by the
not necessary for the receiver to analyse entire picture program.
to indicate the guideline.
Receiver The program is receiving instructions, first one, which does not contain any guidelines. Sample
checks if the folder name is “downloaded”. Then the scheme is visible in Fig. 3. Of course, it is possible
history of files is constructed with proper format in to adjust the program in a way that enables it to turn
order to prevent conducting the same instruction. The itself from receiver into the transmitter and to encode
system script is getting overwritten by meaningless new instructions for every file with proper format. This
writing and consecutively deleted, so it would not be type of the structure is also often used in viruses since
possible to regain it. After carrying the instructions it can infect solely one computer in the inner network,
out the program overwrites the picture by the new because it will spread virus to each system that re-
146
Figure 4: Attached is the picture, in which the original photo is displayed on the left,the determined quantity of bites that
were cut off, starting from A) 6 bites, B) 5 next bites, consecutively in C and D are presented on the right.
ceives graphical information from attacked computer,
and thus the whole network can be infected easily, at
the short period of time.
The system result from input to the output is pre-
sented in Fig. 2. On the entire process an information
is merged with the initial input to compose a final out-
put. On the way an alpha channel is used to fit within
the text format information.
2.1. Message Encoding
In regard to restricted transcript place, it benefits to
use certain an approach for data compression. In pro-
gram, for instance, the Shannon-Fano coding is used
through its easy implementation and satisfying level
of compression. The Shannon-Fano coding is a type
of lossless compression - it finds prefix code for each
discreet source.
Example.
a) S={a,b,c,d}
b) p={0.6,0.2,0.15,0.05}
For the sequence a) which corresponds to the proba-
bility, i.e. the number of occurrence in the text is coded
to the number of all digits b). The situation is visible
in Fig. 5. The program is checking itself whether the
implementation if dictionary at the beginning of cod- Figure 5: On the left is mark a coded as 0. On the right is
ing would take less time than the message. Thereby, mark b coded as 1-0.
the program optimally exploits place in picture.
2.2. Alternative Use for regular computer user. The real intent of a mes-
sage can be therefore easily hidden due to innovative
In regard of increasing surveillance and extended value method. Let us now think how to use the above sys-
of high computer performance, casual ways of infor- tem for this. The first pixels of a graphic are changed
mation protection are becoming steadily unassailable into the desirable file-format, which entails that the
147
Figure 6: Preview of polling picture on the left and version where one part of the picture is magnified for better view on
the right.
Figure 7: Above on, the accuracy of projecting the highest compression of picture is presented. Left photo is an original
source (21 bites = pixel), whereas the next ones show higher compression: 7 bites and 1 bite consecutively.
receiver itself can on an going basis decode messages 2.3. Composed Model of Image Data
conveyed by its download. The same approach can be Encoding
applied to text files. To enable it, the program firstly
changes given file into binary system pattern. The The picture itself contains much more information than
scheme requires to use adequate amount of pictures simple text, so one should label which starting format
to code the message. The receiver, after getting an in- is expected. There are three options, hence, it is nec-
formation turns it into the desirable format. essary to decrease the amount of data in the picture.
a) Colours accuracy
148
Figure 8: Picture shows how the lack of colour changes original file, from left is RGB, RG, GB, RB.
b) Less accuracy of colours projection high loss of the details. The maximum polling is used,
the blurs phenomena takes place, and thereby the pic-
c) Amount of colours ture loses its sharp edges - and ultimately, everything
The color accuracy relies on the differences between is mixed together.
particular colors. It means that value assigned to color In opposite, using the minimum polling reflects in
does not belong already to the interval (0,255), but in- the enhancing the image noises, whereby the trans-
stead is one of the lower power of 2, determined by the parency undergoes decrease. Thus, is is highly sig-
user. Thereby, once the graphics is decoded, the reduc- nificant to adjust properly polling to picture. Sample
tion of pixels will go ahead, which is formed from the schemes of both ways is visible in Fig. 6, while the
formula resulting quality of processing is visible in Fig. 7.
𝑟 = log2 𝑀𝑎𝑥 (1) The amount of colours is crucial through each new
one is adding another dimension to the picture, that
The interval size is a power of 2, due to the writing has been already scaled-down. Thereby quantity of
ongoing in binary form, so that is is possible to use photos being required for sending solely one photo is
the whole length of bites greatly enlarged, i.e. it is possible to establish diversi-
fied options of recreating the colour either by using the
𝑓 (𝑥) = 𝑥 ∗ 28−𝑟 (2) RGB’s values or in the Grayscale, which would save
the place. The changes of colour channels for the in-
Whereas x is value of the pixel and r is square root of
put image are visible in Fig. 8.
the interval size. Sample results from pixels encoding
is presented in Fig. 4. In order to not complicate both,
coder and decoder, the transition takes place in a dis- 3. Conclusions
tinct program, which also allows the independent ac-
tion of a program. It is worth stressing, that the coder There are plenty of risks that occur while one is us-
is adjusted to binary data even though the pixel has ing an internet, hence, we do not want our private
spectrum 256 and still operates at maximum values. information to get into others control. That is why,
The difference must be as little as it is possible due to the developed scheme for quiet conveying instructions
security matter - indeed, once somebody knows the for other computer or sending the graphics and text
communication pattern, the entire scheme is not ade- messages is an important to be considered in common
quate anymore. information-protection manners. Moreover, the pa-
The precision of colors projection relies on the sim- per points out the possible, sensible protection of an-
plification of the graphics by using polling method. tivirus systems. It is important to notice that the send-
The picture is converted as scantily as possible due to ing graphics process is problematic action, however,
149
under certain circumstances it seems to be only avail-
able approach for sending sensible information-media
graphics, for instance under high censorship condi-
tions. The idea will be continuously developed toward
optimal encoding. I do believe it is the most optimal
approach in case of protecting the data.
References
[1] G. Capizzi, G. L. Sciuto, C. Napoli, D. Polap,
M. Woźniak, Small lung nodules detection based
on fuzzy-logic and probabilistic neural network
with bio-inspired reinforcement learning, IEEE
Transactions on Fuzzy Systems (2019).
[2] D. Połap, M. Woźniak, R. Damaševičius,
R. Maskeliūnas, Bio-inspired voice evalua-
tion mechanism, Applied Soft Computing 80
(2019) 342–357.
[3] M. Woźniak, D. Połap, L. Kośmider, T. Cłapa, Au-
tomated fluorescence microscopy image analysis
of pseudomonas aeruginosa bacteria in alive and
dead stadium, Engineering Applications of Artifi-
cial Intelligence 67 (2018) 100–110.
[4] Y. Wang, L. Kong, Z. Qian, G. Feng, X. Zhang,
J. Zheng, Breaking permutation-based mesh
steganography and security improvement, IEEE
Access 7 (2019) 183300–183310.
[5] I. J. Kadhim, P. Premaratne, P. J. Vial, B. Halloran,
Comprehensive survey of image steganography:
Techniques, evaluations, and trends in future re-
search, Neurocomputing 335 (2019) 299–326.
[6] J. Qin, Y. Luo, X. Xiang, Y. Tan, H. Huang, Cover-
less image steganography: A survey, IEEE Access
7 (2019) 171372–171394.
[7] Z. Wang, X. Zhang, Secure cover selection for
steganography, IEEE Access 7 (2019) 57857–57867.
[8] F. Bonanno, G. Capizzi, G. L. Sciuto, C. Napoli,
Wavelet recurrent neural network with semi-
parametric input data preprocessing for micro-
wind power forecasting in integrated generation
systems, in: 2015 International Conference on
Clean Electrical Power (ICCEP), IEEE, 2015, pp.
602–609.
[9] G. Capizzi, S. Coco, G. Sciuto, C. Napoli, A new
iterative fir filter design approach using a gaussian
approximation, IEEE Signal Processing Letters 25
(2018) 1615–1619.
150