Lattice approach to video steganographyjournals.tubitak.gov.tr/elektrik/issues/elk-15-23... · 2....
Transcript of Lattice approach to video steganographyjournals.tubitak.gov.tr/elektrik/issues/elk-15-23... · 2....
Turk J Elec Eng & Comp Sci
(2015) 23: 2074 – 2088
c⃝ TUBITAK
doi:10.3906/elk-1302-21
Turkish Journal of Electrical Engineering & Computer Sciences
http :// journa l s . tub i tak .gov . t r/e lektr ik/
Research Article
Lattice approach to video steganography
Onur UNLU∗, Hacer KARACANDepartment of Computer Engineering, Faculty of Engineering, Gazi University, Ankara, Turkey
Received: 03.02.2013 • Accepted/Published Online: 02.07.2013 • Printed: 31.12.2015
Abstract: As a result of the continuously increasing amount of important data, the demand for information security
increases daily. Steganographic techniques can play a considerable role in solving this issue, since they can hide the
necessary data. Generally, audio and image files have been utilized as hiding media in this area. However, because of
their limited capacity, these media limit the amount of hidden information. On the other hand, the capacity limit can be
exceeded by applying steganography on video files. There are few studies on video steganography in the literature. In this
study, a steganographic library, which is independent of medium and method, was developed for hiding information in
audio, image, and video files by using various methods. Furthermore, distinctly from other studies, the lattice approach
was proposed for this issue. In this approach, security was increased by using the image and audio domain of video
together in the hiding process. In addition, we were able to hide the real-time data in a video taken from a webcam.
Thus, the size of the produced video file could be set, according to the size of the secret file, with this model. Meanwhile,
the possibility of detecting the existence of steganography by comparing to the original video was limited. The success
of the hiding process was also evaluated with various steganalyses and noise measurement methods, and the results are
presented and discussed in the relevant sections.
Key words: Video steganography, steganographic library
1. Introduction
Nowadays, the rapid increase in information-sharing causes various security problems. New security gaps
are emerging daily. Two main ways of providing security in information communication are cryptology and
steganography.
Cryptology is based on the rule of encrypting the information to be sent and decrypting the information
on the receiver side. In order to obtain the information, third parties that are listening to the communication
media must break the password. Here, it is important is to prevent the content of the communication from
being obtained by third parties or software programs that are not participating in the communication itself.
Since cryptologic systems gain their power from the encryption algorithm and secrecy, they can be hacked when
the existence of the algorithm is realized.
Steganography is a technique of hiding a meaningful group of data (implicit/secret object) into another
meaningful group of data (explicit/cover object), which is larger in size. It is based on sending the confiden-
tial/secret information by hiding it in an object that seems innocent to third parties. Thus, it is important not
to make anything appear suspicious to third parties or software. The system is hacked when it is understood
that the file is carrying secret information.
∗Correspondence: [email protected]
2074
UNLU and KARACAN/Turk J Elec Eng & Comp Sci
In the real world, burying treasure at a specific place and then finding and removing it can be a good
analogy for steganography. In a virtual environment, the implicit/secret object can be a group of meaningful
bits, and the explicit/cover object can be any appropriate file type.
Steganography is the science of secret communication. Etymologically the term ‘steganography’ comes
from the Greek word for ‘concealed text’ [1]. The basic idea is to create the stego object by embedding the im-
portant information into an ordinary object (text, image, sound, or video). The stego object is sent/transmitted
and stored in the same way as a cover object. For uninformed observers, stego objects display the content of
the cover object. From the target user’s viewpoint, the stego object carries the important information hidden
by the cover object. Generally, it is a three-phase process. Secret information is buried into the cover object by
using an information-hiding algorithm, and in this way the stego object is created and is sent to the receiver
via communication media. Then the receiver acquires the secret information from the stego object using a data
extraction algorithm. This process is displayed in Figure 1.
Communication
Line
Data Hiding
Algorithm
Secret Information
Cover Object
(Audio, Image, Video)
Stego Object
Stego Object
Data Extraction
Algorithm
Secret Information
Figure 1. Operation logic of steganography.
Perceptibility, robustness, and capacity factors can be considered as the success criteria for steganographic
procedures. Generally, these criteria adversely affect each other and are therefore known as the ‘magic triangle’
(Figure 2) [2]. Perceptibility is used as a measure of noticeability of the secret transmission in the medium by
third parties intuitively or by other means. Robustness is used as a measure of acquisition of secret information
after performing operations that cause changes in the object, such as cropping or filtering. Capacity is used
as a measure of the maximum size that can be stored in the media. The relation between these 3 factors waspreviously examined in sound files [3]. Among these factors, perceptibility and capacity were mostly considered
in the studies of the field.
Magic
Triangle
Capacity
Perceptibility Robustness
Figure 2. Magic triangle.
2075
UNLU and KARACAN/Turk J Elec Eng & Comp Sci
2. Steganography in the literature
In the literature, studies on steganography can be grouped into 4 main types as text, sound, image, and video.
2.1. Text steganography
Text steganography is one of the oldest steganography methods. It is the most difficult area for the application
of the steganography [4]. This is due to the lack of repeating unnecessary data compared to media that use
sound and video. This method was successfully applied previously in different languages such as Thai, English,
Japanese, Korean, Chinese, Persian, Arabic, and Turkish [5,6].
In their study, Shirali-Shahreza and Shirali-Shahreza [4] proposed a method where the vertical height of
these letters is used for hiding with the help of letters in Arabic and Persian that have punctuation marks. In
another study, Shahreza [7] performed a study that stored information by changing words that were different in
US and UK English, yet held the same meaning. In a further study, Chinese was mentioned as using simplified
and conventional characters interchangeably, and this property was identified as a possibility for steganographic
purposes. In addition, the usage of conventional characters for ones and simplified characters for zeros was
suggested for users [8].
2.2. Sound steganography
Researchers in Pakistan proposed an advanced type of the least significant bit (LSB) method [9]. With this
method, 8 bits of sound samples are used in the hiding process, and sample and data hiding bits are selected
according to a map (see Table 1). This gives the impression that the information hiding is done randomly,
which increases security against statistical attacks.
Table 1. Secret bit selection map for a sound sample [9]. If the first 2 bits of the sound sample are 1, the secret data
will be hidden in the 2nd LSB of the sound sample.
First 2 bits Secret bit index00 3. LSB01 2. LSB01/11 1. LSB
Spontaneity in bit selection is provided by choosing the first 2 bits of the sound sample. Table 1 provides
the conditions about which index to use. A similar method is used when selecting the sound sample. The first
3 bits of a sound sample indicate a subsequent sound sample that contains the secret. Average capacity is 1 bit
for 4 sound samples. In their study, Yan and Wang performed steganography with the selection of a table in
Huffman coding, which is a property of the MP3 compression standard [10]. Compared to other methods, this
provides high capacity and also ensures security and low deformity requirements. Furthermore, Shirali-Shahreza
proposed a new method that performs real time hiding and has robustness to the MP3 compression by changing
the length of the silent period of speech sounds [11].
2.3. Image steganography
The LSB method is used for image steganography in most studies [12]. With this method, the bit that constitutes
one pixel of the image and has the least effect on the color is changed. Thus, we obtain the area where we can
nest the secret information (see Table 2 for examples).
2076
UNLU and KARACAN/Turk J Elec Eng & Comp Sci
Table 2. Changing of last bits during LSB hiding operations.
Value before hiding Secret bit Value after hiding10111111 (191) 1 10111111 (191)11110011 (231) 1 11110011 (231)10000111 (135) 0 10000110 (134)
Westfeld and Pfitzmaan presented statistical and visual attack methods that are determined when the
standard LSB method is used for hiding [13]. This attack method zeroizes the first 7 bits of the eight-bit color
value, thus turning the color values of the image into zeros and ones. Then the ones are converted to 255, which
is the highest color value, and the zeroes are left as they are. This method can be appropriate for images using
simple LSB; however, when a complex image is chosen, it does not work. Images that have black and white
values depending on the LSBs are given in Figure 3. It can be seen in the figure that some information is hidden
in half of the second one.
Figure 3. Example of a visual attack [13].
When changes are performed on the last bits using the LSB method, there will be unwanted effects on
the image. The ratio between the ones and zeros of the last bits is high in the original image; however, it
will be lower in the images carrying hidden information. The attack using this statistical difference is called a
chi-square attack (Figure 4)
Figure 4. Chi-square attack [13].
2077
UNLU and KARACAN/Turk J Elec Eng & Comp Sci
In order to get more robust images against the chi-square attack, a pseudorandom number generator
(PRNG) can be used for the selection of a pixel that is used for hiding. The disadvantage of the PRNG is that
it requires preliminary communication. In addition, if a PRNG is used, the size of the information to be hidden
should be kept small for security purposes.
The least significant bit modulation (LSBM) method, which was first proposed by Sharp, differs from
the LSB method in terms of changing method of the last bit [14]. In this method, the last bit of the cover is
not changed with the hide bit. If the last bit of the cover is different from the hide bit, hiding is achieved by
increasing or decreasing by 1. When this method is used in the frequency area, the stego image becomes more
robust against statistical attacks.
In 2006, Mielikainen improved this method and was able to hide the same amount of information by
changing less pixels. In the proposed method, pixels were used as pairs. One bit of information was hidden into
the last bit of the first pixel, and it obtained the second bit of information from a function of the pixel pair. In
this way, the cover image became more resistant to attacks [15].
Researchers from China developed the method proposed by Mielikainen in their study. They stated
that the plain and smooth areas in the cover image would be visually weak and would provide low security
for information hiding. In order to prevent this, they proposed a method that can choose hiding areas and
perform hiding according to the size of the information and the difference between 2 successive pixels. When
the information to be hidden is small, the method leaves the plain areas of the cover image as is and performs
hiding on the areas where the sharpness is dense, thus increasing security [16].
It can be seen from Figure 5 that the method does not hide plain areas at low hiding ratios. With an
increase in the hiding ratio, the method uses plain areas, and this decreases security.
Figure 5. a) Original image, b) 10% hidden data, c) 20% hidden data, d) 30% hidden data, e) 40% hidden data, and
f) 50% hidden data [16].
Researchers from King Saud University proposed a different method, in which images are split into n
equal blocks (i.e. splitting a 100 × 100 image into four 50 × 50 images) and LSB is applied spirally. They
2078
UNLU and KARACAN/Turk J Elec Eng & Comp Sci
proposed 3 different applications for this method: from the outer corners of the blocks, from the center of the
blocks, or using both by applying them clockwise or counterclockwise. Results showed that this method is
considerably resistant to the chi-square attack [17].
In another study, a method that performs hiding by weighing RGB values was proposed [18]. With this
method, high hiding capacity was achieved with low visual corruption. Since this method depends on changing
the last digit instead of the last bit, it is known as the least significant digit (LSD), which is a derivation from
LSB. Although the last digits of the binary system are used in the LSB method, the last digit is used for hiding
in the decimal system. The number is increased by 10, and the digit to be hidden is subtracted. There are some
cases where exceptional outcomes arise. If the original G and B values are between 250 and 255, the algorithm
produces invalid results. In order to overcome this problem, each G and B value is reduced by 10 [19]. With
this hiding method more information can be hidden, whereas the noticeability of the hiding process increases.
Although the maximum difference between the numbers before and after hiding is 9, when the binary system
is used, it is seen that the upper bits can be changed (Table 3). This situation results in serious corruption of
an image when the application that provides the image performs the calculation based on the bit values.
Table 3. Change of lower bits during LSD hiding operation [18,19].
Value before hiding Secret value Value after hiding
10111111 (191) 111000111(190 + 10 – 1 = 199)
11110011 (231) 411111000(230 + 10 – 4 = 236)
10000111 (135) 710000101(130 + 10 – 7 = 133)
Wang and Chen proposed a two-way block-matching method that is based on block-matching of the cover
image and hidden image [20]. Two images are split into blocks of m × n. The highest similarity block to the
ith block in the cover image is found and matched. This process is repeated continuously. After the process,
the index information of the well-matched blocks and unmatched blocks is compressed and stored in LSB to
be distributed to the cover image. The results of the experiment indicated that it provides low noise and high
capacity [20]. In another study, the present block-matching method was presented as expanding the block search
field, and thus lowered the amount of information to be hidden in the block. In this way, a better quality cover
was claimed to be achieved on the hidden image than in the previous study [21]. Another method, spread-
spectrum image steganography (SSIS), was also presented for hiding the secret information inside the noise [22].
In order to retrieve the secret information successfully, restoration techniques were used. If the original image
were to be achieved by restoration, the added noise would also be achieved. With these techniques, since the
information cannot be achieved without errors, error control-coding was applied before hiding. In addition, it
was seen that the hidden information could endure the compressed image with well-chosen parameters [22].
Shirali-Shahreza described another steganographic method that he developed for multimedia messaging
services [23]. MMS messages can transmit text up to 30 KB, images up to 100 KB, and videos up to 300 KB.
With the proposed method, the text and image fields in the message were used together. Using the eye’s lesser
sensitivity to blue, the system was utilized with the images and abbreviations in the text field of messages
(for/4, you/u) [23].
Methods using the frequency field generally use discrete cosine transformation (DCT). JPEGs are the
image files where DCT steganography is mainly used. In JPEG files, LSB is utilized after DCT, and quantization
2079
UNLU and KARACAN/Turk J Elec Eng & Comp Sci
and rounding are applied to 8 × 8 image blocks. Steps of this method are shown in Figure 6. Since all blocks will
be affected by a change in a value, choosing the values where the information will be hidden is very important in
this method. J-Steg software that uses LSB on these coefficients was presented in 1997 (http://www.jsteg.org).
Although the method was resistant to visual attacks, it was weak against statistical attacks. In their study,
Westfeld and Pfitzmann detected the hidings performed by J-Steg software by using the chi-square attack
technique [13].
8 x 8 Image Block 8 x 8 Block
A!er DCT
8 x 8 Quantization
Table
8 x 8 Quantized Block
Hiding Bits &
Hu#man Coding Stego
Result
DCT
Division &
Rounding
Figure 6. Hiding data to frequency domain in JPEG files.
Westfeld developed an F5 algorithm that is resistant to visual and statistical attacks [24]. With this
algorithm, it was ensured that the histograms of the DCT coefficients were not corrupted and that the
information was hidden into DCT coefficients in a distributed manner. One year later, Fridrich et al. detected
the information that was hidden by F5 [25]. The key point in this method is the estimation of the histogram
of the image before hiding. This can be achieved by cutting 4 pixels on both sides and compressing them
again to remove quantization in the frequency field of the image. An attack is performed by comparing the
histograms of 2 images [24]. In addition, 2 new methods, namely quantization index modulation (QIM) and
histogram matching (HM), were proposed to allow more information-hiding capability than F5 [26]. These
methods removed the histogram difference that occurred as a result of the attack against F5.
Moreover, there are several methods that can cheat the chi-square attack. One of these methods is based
on sequential binary grouping of the DCT coefficient values (i.e. 2–3, 4–5) [27]. When 2 numbers belonging
to a group are found, one-bit hiding can be performed. This reduces the hiding capacity by half. Hiding is
performed in such a way that the numbers will not be outside the group, and this way a change in any group
will not affect the others.
2.4. Video steganography
The capacity of sound files and motionless images is limited. This capacity constraint also restricts the amount
of information that can be hidden. Anyone that attempts to hide a folder in the file system cannot perform this
for most of the folders. This is because the folder to be hidden exceeds the hiding capacity of the cover image.
In order to overcome this capacity limitation, video files are used.
2080
UNLU and KARACAN/Turk J Elec Eng & Comp Sci
Video files are similar to image files, yet with several important differences. First, there are more pixels
in video files. This also increases the hiding capacity of the videos. In addition, changes in the videos of small
differences cause the repetition of unnecessary information. Compressing methods that are used to prevent this
unnecessary information render the hiding difficult.
In a study, a high rate of information-hiding was implemented using AVI video format [28]. This method
made use of the blue sensitivity of the human eye and achieved information hiding by 1/3 by hiding information
to 3 bits to red and green fields and 2 bits to blue field of one pixel. Additionally, the researchers reported that
crowded videos are more suitable for hiding, since the human eye is more sensitive to the changes in wide and
flat areas, such as sky and sea, than to the changes in mixed areas that include too many objects [28].
In another study, Cetin and Ozcerit proposed a new method that uses the color histogram of the
environment when choosing pixels to be hidden [29]. According to this method, the histograms of the color
components of the frames are calculated, and a threshold value is determined by comparing the histogram
values of the successive frames. This threshold value is used in pixel selection, and the method is applied as
both frame-based and block-based. The block-based method was successful in terms of perception. In another
study, only type “I” frame of MPEG files was used. During the generation of the “I” frame, it is recommended
to perform hiding into coefficients above certain threshold values, following DCT [30]. In addition, in their
proposed algorithm, researchers from China hid the hiding capacity of GOP and the data holding the starting
point of the hidden information, referred to as control data in the “I” frame. Secret information was hidden in
type “P” and “B” frames. Provided that the change in the motion vector was above a certain threshold value,
the last bit of the vector was used for hiding [31]. In another study, type “I” frames were not used because of
their vulnerability to steganalysis, and the researcher used type “P” and “B” frames for hiding instead [32].
Unlike other studies, when selecting the vectors for hiding, associated macroblock prediction error was used
instead of vector magnitude and the angular properties of the vector [32].
3. Developed application
The aim of the developed software is to create a flexible library and an interface for the direct use of this software
for steganographic applications. The library supports the LSB, LSBM, and LSD methods that were described
in Section 2.3. It also supports raw media environments (WAV, BMP, and AVI) that do not use compressions.
Figure 7 shows the data-hiding interface of the Stegofolderlock application.
Figure 7. Stego folder lock data-hiding interface.
2081
UNLU and KARACAN/Turk J Elec Eng & Comp Sci
For platform independency, the application has been developed by Java programming language. Since the
application will be platform- and method-independent, flexible design was an important requirement. Therefore,
various design patterns, such as bridge, facade, composite, delegation, and factory, were applied. Figure 8 shows
a general class diagram of the developed steganographic library.
Figure 8. System general class diagram.
4. Experiments
Input and output images of information hiding on a BMP image, using standard LSB, LSBM, and LSD, are
given in Figure 9. As seen in the figures, there is no noticeable difference between the first 3 images. On the
contrary, there is a noticeable difference in the fourth image. This is because the LSD method changes the
last 3 bits of one byte densely, and the aircraft image is an image that does not show gray color variations.
The second row of the experimental image shows the result of the visual attack against the last bit plane, the
third row shows the visual attack against the second to last bit plane, and the fourth row shows the attack
against the antepenultimate bit plane. Hidden data can be detected as a result of applied steganalysis on the
second, third, and fourth images. The second row shows that there is a big difference between the last bit plane
of the original image and the last bit planes of stego images, where data were hidden with the LSB, LSBM,
and LSD methods. This difference shows that those bits were changed; in other words, there are hidden data.
At this point, we should mention that the hidden data for all these methods were the same. Whereas the
LSB and LSBM methods hide a single bit into one byte, the LSD method can hide a single byte into 3 bytes.
Because of this capacity difference, the change in the scan area of LSD method is one-third that of the other 2
methods. The LSB method changes only the related bit of a byte. The LSBM method has the same effect as
the LSB method on the last bit plane. The LSBM method is different from the LSB method by incrementing or
decrementing the number by one. Even though this process changes the value of the number by one, there is the
possibility of change on the upper bits of the byte. It is seen that the LSBM method changes the penultimate
and antepenultimate bits of the byte in the third row of Figure 9. Similarly, the LSD method has the possibility
of change on the upper bits of the byte even if it changes the value of the number up to 9.
As shown in Figure 10, when statistical attacks are performed against the aircraft images, it becomes
clearer that steganography was applied in all 3 methods. It is observed that the “0” and “1” balance is equal
in the gray marked area of the last part. As a result of this attack, it can be said that there are hidden data in
the cover image.
2082
UNLU and KARACAN/Turk J Elec Eng & Comp Sci
Last bitplane
Last butone bitplane
Antepenultimatebit plane
a) Original image b) LSB image c) LSBM image d) LSD image
Figure 9. Visual attack against aircraft images.
0-1rate
a) OriginalImage
b) LSBImage
c) LSBMImage
d) LSDImage
Image Row Index
Figure 10. Statistical attack against aircraft images.
There seems to be no difference between LSB and LSBM without PRNG against visual and statistical
attacks. In the LSBM method, the change possibility of the penultimate bit, in addition to the last bit, is 50%.
This kind of change may lead to a change in the upper bits, where it does not affect the image files, but causes
sensible distortion to audio files.
2083
UNLU and KARACAN/Turk J Elec Eng & Comp Sci
PSNR, mean squared error (MSE), structural similarity (SSIM) [33], and universal image quality index
(UQI) [34] values are compared in Table 4. It is seen that both LSB, in the first column, and LSBM, in the
second column, have the same destruction effect on the image. This is because the amount of change that the
LSB method caused by changing the last bit is equal to the amount of change that the LSBM method caused
by randomly increasing or decreasing the number by one on the color value of one pixel. In the third column, it
is seen that the noise added with the LSD method is more than in the other two. The fourth column shows nodifference between the images where the LSB and LSBM methods were used. Even if these two methods achieve
the same results on the last bits, the number values are different, and this is the reason for the difference.
Table 4. Aircraft PSNR, MSE, SSIM, and UQI values.
Original image- Original image Original image-LSB-LSBM
LSB LSBM LSBPSNR 51.79 dB 51.79 dB 40.68 dB 48.81 dBMSE 0.43 0.43 5.59 0.86SSIM 0.995 0.994 0.326 0.990UQI 0.998 0.998 0.996 0.997
In Figure 11, which is filled with colors, the application of the LSB method can be detected, whereas the
detectability of the LSD method’s application is very low when visual attack is used. Applied statistical attack
becomes ineffective on color intensive complex images (Figure 12).
Last bitplane
a) Original image b) LSB image c) LSBM image d) LSD image
Last butone bitplane
Antepenultimatebit plane
Figure 11. Colorful image that is more complex than aircraft image.
2084
UNLU and KARACAN/Turk J Elec Eng & Comp Sci
0-1rate
a) OriginalImage
b) LSBImage
c) LSBMImage
d) LSDImage
Image Row Index
Figure 12. Statistical attack against colorful image.
Applied methods were examined by adding noise and using compression in terms of robustness. The salt
and pepper, Gaussian, and speckle noise methods were used. In addition, the images were compressed in JPEG
format and then decompressed. It was seen that these methods do not provide robustness. For the purpose of
redounding robustness, distributed and seldom hiding with PRNG and CRC can be used. It is obvious that
they would reduce capacity.
Stego folder lock application can hide data by using both image and audio domains together in AVI 1.0
video files. For example, if we take a video file that has a 352 × 240 frame size, 29 fps frame rate, and 33 s
length, the image domain of this video will be the size of 352 × 240 × 3 × 29 × 33 ≈ 230 MB. If we assume
that the same video has a stereo audio that has a 16-bit sample size and 44,100 kHz sample rate, the audio
domain of this video will be 44,100 × 16/8 × 2 × 33 ≈ 5 MB. If the entire capacity of both the image and
the audio domain of this video is used, 1/47 percent of hidden information will be placed in the audio domain
and 46/47 percent of hidden information will be placed in the image domain. Because the application hides
secret data as packages, some in the image domain and then some in the audio domain, one after another, this
method can be named as the lattice approach. This way security is enhanced.
Experiments performed on the video steganography-applied files by using the zipper method with stego
folder lock application are given below.
The video’s audio domain wave form and spectrogram graphics, as a result of hiding with the LSBM and
LSD methods, are seen in Figures 13 and 14. The effect of the visual attack on the image domain of the same
video shows that the effects are the same when plain image and single video frames are compared.
5. Conclusion
In this study, a flexible library was designed for data hiding. In addition to being independent from mediums
and methods, the library was also platform-independent, since the Java programming language was used.
In order to increase complexity and security, the lattice approach, which uses image and sound fields of
video files, was proposed (Figure 15). Real-time hiding of the webcam-captured video into the image field was
also provided in the library. Thus, the detection of the difference from the actual video file was prevented, and
the ideal video required for the data to be hidden was taken.
As a result of the tests that were applied on raw image files, it was seen that in flat images, the LSB and
2085
UNLU and KARACAN/Turk J Elec Eng & Comp Sci
Figure 13. Wave forms of audios in the original and stego videos.
Figure 14. Spectrograms of audios in the original and stego videos.
LSBM methods did not cause a perceptible difference, although the LSD method did. All 3 methods applied
to the flat images can be detected by visual attacks against the last bits and by chi-square attacks. On the
other hand, all 3 methods seem to be imperceptible, and chi-square attacks show no result when the hiding is
performed on complex images. All 3 methods can be detected by visual attacks; however, the LSD method was
found to be more resistant than the other two.
2086
UNLU and KARACAN/Turk J Elec Eng & Comp Sci
Video File
Hea
der
Audio
Image
Secret
Object
Image and Audio
Steganography
Let Ayşe Go on Holiday
Figure 15. Applied lattice approach to enhance security.
The number of applications that can perform data-hiding in videos is very limited. In this study, unlike
previous studies, the sound field as well as the image field within the video file were used for hiding. Consequently,
complexity and security levels were increased.
References
[1] Cheddad A, Condell J, Curran K, McKevitt P. Digital image steganography: survey and analysis of current methods.
Signal Process 2010; 90: 727–752.
[2] Cvejic N, Seppanen T. A wavelet domain LSB insertion algorithm for high capacity audio steganography. In: IEEE
2002 Digital Signal Processing Workshop; 13–16 October 2002; Oulu, Finland. New York, NY, USA: IEEE. pp.
53–55.
[3] Gopalan K, Shi Q. Audio steganography using bit modification—a tradeoff on perceptibility and data robustness
for large payload audio embedding. In: ICCCN 2010 International Conference on Computer Communications and
Networks; 1–6 August 2010; Zurich, Switzerland. New York, NY, USA: IEEE.
[4] Shirali-Shahreza MH, Shirali-Shahreza M. A new approach to Persian/Arabic text steganography. In: IEEE/ACIS
International Conference on Computer and Information Science; 10–12 July 2006; Honolulu, HI, USA. New York,
NY, USA: IEEE. pp. 310–315.
[5] Samphaiboon N, Dailey MN. Steganography in Thai text. In: Proceedings of ECTI-CON 2008. pp. 133–136.
[6] Meral HM, Sankur B, Ozsoy AS. Watermarking tools for Turkish texts. In: IEEE 2006 Signal Processing and
Communications Applications; 17–19 April 2006; Antalya, Turkey. New York, NY, USA: IEEE. pp. 1–4.
[7] Shirali-Shahreza MH, Shirali-Shahreza M. A new synonym text steganography. In: IIHMSP 2008 International
Conference on Intelligent Information Hiding and Multimedia Signal Processing; 15–17 August 2008; Harbin, China.
New York, NY, USA: IEEE. pp. 1524–1526.
[8] Sun X, Meng P, Ye Y, Hang L. Steganography in Chinese text. In: ICCASM 2010 International Conference on
Computer Application and System Modeling; 22–24 October 2010; Taiyuan, China. New York, NY, USA: IEEE.
pp. 651–654.
[9] Asad M, Gilani J, Khalid A. An enhanced least significant bit modification technique for audio steganography.
In: ICCNIT 2011 International Conference on Computer Networks and Information Technology; 11–13 July 2011;
Abbottabad, Pakistan. New York, NY, USA: IEEE. pp. 143–147.
[10] Yan D, Wang R. Huffman table swapping-based steganography for MP3 audio. Multimed Tools Appl 2009; 52:
291–305.
[11] Shirali-Shahreza MH, Shirali-Shahreza S. Real-time and MPEG-1 layer III compression resistant steganography in
speech. IET Inform Secur 2010; 4: 1–7.
2087
UNLU and KARACAN/Turk J Elec Eng & Comp Sci
[12] Akar F, Yalman Y, Varol HS. Data hiding in digital images using a partial optimization technique based on classical
LSB method. Turk J Electr Eng Co 2013; 21: 2037–2047.
[13] Westfeld A, Pfitzmann A. Attacks on steganographic systems. Lect Notes Comput Sc 2000; 1768: 61–76.
[14] Sharp T. An implementation of key-based digital signal steganography. Lect Notes Comput Sc 2001; 2137: 13–26.
[15] Mielikainen J. LSB matching revisited. IEEE Signal Proc Lett 2006; 13: 285–287.
[16] Luo W, Huang F, Huang J. Edge adaptive image steganography based on LSB matching revisited. IEEE T Inf
Foren Sec 2010; 5: 201–214.
[17] Mathkour H, Assassa GMR, Al Muharib A, Kiady I. A novel approach for hiding messages in images. In: 2009
International Conference on Signal Acquisition and Processing; 3–5 April 2009; Riyadh, Saudi Arabia. New York,
NY, USA: IEEE. pp. 89–93.
[18] Akar F, Varol HS. A new RGB weighted encoding technique for efficient information hiding in images. J Nav Sci
Eng 2004; 2: 21–36.
[19] Yalman Y, Akar F, Erturk I. An image interpolation based reversible data hiding method using R-weighted coding.
In: IEEE 2010 International Conference on Computational Science and Engineering; 11–13 December 2010; Hong
Kong. New York, NY, USA: IEEE. pp. 346–350.
[20] Wang R, Chen Y. High-payload image steganography using two-way block matching. IEEE Signal Proc Lett 2006;
13: 161–164.
[21] Chen S, Wang R. High-payload image hiding scheme using k -way block matching. In: Proceedings of the 2010
International Conference on Intelligent Information Hiding and Multimedia Signal Processing; 15–17 October 2010;
Darmstadt, Germany. New York, NY, USA: IEEE. pp. 70–73.
[22] Marvel LM, Boncelet CG, Retter C. Spread spectrum image steganography. IEEE T Image Process 1999; 8: 1075–
1083.
[23] Shirali-Shahreza M. Steganography in MMS. In: IEEE 2007 International Multitopic Conference; 28–30 December
2007; Lahore, Pakistan. New York, NY, USA: IEEE. pp. 1–4.
[24] Westfeld A. F5—a steganographic algorithm. Lect Notes Comput Sc 2001; 2137: 289–302.
[25] Fridrich J, Goljan M, Hogea D. Steganalysis of JPEG images: breaking the F5 algorithm. Lect Notes Comput Sci
2003; 2578: 310–323.
[26] Noda H, Niimi M, Kawaguchi E. Application of QIM with dead zone for histogram preserving JPEG steganography.
In: IEEE 2005 International Conference on Image Processing; 11–14 September 2005; Genova, Italy. New York,
NY, USA; IEEE. pp. 1082–1087.
[27] Famili Z, Faez K, Fadavi A. A new steganography based on χ2 technic. Lect Notes Comput Sc 2009; 5856: 1062–
1069.
[28] Eltahir ME, Kiah MLM, Zaidan BB, Zaidan AA. High rate video streaming steganography. In: Proceedings of 2009
International Conference on Future Computer and Communication; 3–5 April 2009; Kuala Lumpur, Malaysia. New
York, NY, USA: IEEE. pp. 550–553.
[29] Cetin O, Ozcerit AT. A new steganography algorithm based on color histograms for data embedding into raw video
streams. Comput Sec 2009; 28: 670–682.
[30] Stanescu D, Stratulat M, Ciubotaru B, Chiciudean D, Cioarga R, Micea M. Embedding data in video stream using
steganography. In: SACI 2007 International Symposium on Applied Computational Intelligence and Informatics;
17–18 May 2007; Timi?oara, Romania. New York, NY, USA: IEEE. pp. 241–244.
[31] Xu C, Ping X, Zhang T. Steganography in compressed video stream. In: Proceedings of the 2006 International
Conference on Innovative Computing, Information and Control; 30 August–1 September 2006; Beijing, China. New
York, NY, USA: IEEE. pp. 269–272.
[32] Aly HA. Data hiding in motion vectors of compressed video based on their associated prediction error. IEEE T Inf
Foren Sec 2011; 6: 14–18.
[33] Wang Z, Bovik AC, Sheikh HD, Simoncelli EP. Image quality assessment: from error visibility to structural
similarity. IEEE T Image Process 2004; 13: 600–612.
[34] Wang Z, Bovik AC. A universal image quality index. IEEE Signal Proc Lett 2002; 9: 81–84.
2088