Content uploaded by Chakchai So-In
Author content
All content in this area was uploaded by Chakchai So-In on Jun 17, 2021
Content may be subject to copyright.
e-ISSN: 2289-8131 Vol. 9 No. 2-2 23
Video Steganography with LSB Color Detection
Arada Suttichaiya1, Yuwarat Sombatkiripaiboon1, Phet Imtongkhua1, Chatchai Poonriboon1, Chakchai So-In1 and
Paramate Horkaew2
1Applied Network Technology (ANT), Department of Computer Science, Khon Kaen University, Khon Kaen, Thailand.
2School of Computer Engineering, Institute of Engineering Suranaree University of Technology,
Nakhon Rachasima, Thailand.
chakso@kku.ac.th
Abstract—Steganography is the method employed to prevent
unsolicited access and malicious use of sensitive information.
This research proposes an alternative approach to video
steganography by exploiting Least Significant Bit (LSB) in the
binary stream. Its main contribution is incorporating a color
detection technique to optimize steganographic performance.
The proposed encoding and decoding methods were
implemented on MATLAB software to illustrate its applicability
on uncompressed AVI movie. The results showed that it could
conceal text, image, audio, and video data with non-secret
streams. In this study, Peak Signal to Noise Ratio (PSNR) was
used to assess its performance, whereby significant
improvement over generic methods was found
Index Terms—Color Detection; Head Frame; Least
Significant Bit; Video Steganography.
I. INTRODUCTION
Security in data communication and archiving is currently of
utmost interest in the world of digital economy. Among the
most valuable information needed protected are document
and audiovisual contents involving both business transactions
and personal assets. Steganography is therefore needed to
conceal those data to prevent unsolicited access to those files
without consent from communicating parties.
Without the necessary concealment, there are more than
556 million computer users worldwide (equivalent to 1.5
million a day or 18 people every second) currently being
victimized [1]. These attacks can be categorized into
computer virus, malware, worm, Trojan, and espionage both
by the trusted insiders and general public. This type of
computer crimes is considered the immediate threat to our
society which affects both consumers and vendors alike and
hence needs emergent remedy to prevent the unwarranted
interceptions of such sensitive messages.
Nowadays, online media is a crucial means of
broadcasting information to the public such as that managed
by YouTube. According to the recent survey, during 2006 to
2013, the public interests in this kind of communication
channels, e.g., contributors and viewers have increased by 72
percent [2].
Consequently, this research proposes a technique to
conceal sensitive information with video stream. In this
study, the least significant bits (LSB) in the pixel stream were
replaced by those of hidden information. To this end, the
detection of pixel colors and head frame designation was
employed to determine a suitable ratio of concealment in each
color component, so as to increase the amount of information
being encoded.
This article is organized as follows. Section 2 gives
background information of this research, and then related
work will be briefly discussed in Section 3. Section 4 then
explains our methodology for video steganography. Next,
section 5 discusses the results obtained from performance
evaluation. The conclusions and future work are also
presented in the last section.
II. THEORY AND RELATED MATERIALS
This section explains the background theory and relating
material for video steganography.
A. Image Processing
Image processing is a set of techniques which processes
information in an image or a sequence thereof by means of a
computer program. Its objectives are to extract quantitatively
and qualitatively the required information, e.g., extent, shape,
and motion direction of an object in the images. The extracted
information would then by analyzed and used to build a
system in various fields, such as fingerprint recognition,
automatic postcode reader, and face identification.
B. Steganography
Steganography is a method for concealing secret data with
general media, such as image, audio, video, and printed
material. The method differs from cryptography in that the
latter is probable to notice and if appropriate decrypting tool
exists, it can be used to read the encrypted information,
depending on various factors, such as the complexity of the
code and algorithm. Steganography, on the contrary, hides
that information behind sometimes mundane irrelevant data,
ensconcing it from vicious intent. To tighten its security, the
steganogram can be also encrypted with another method of
choice.
C. Least Significant Bit (LSB) Steganography
LSB steganography is a technique which alters the least N
significant bits of a pixel and replaces them with the message
wish to be hidden [3]. For instance, a letter ‘A’ is to be hidden
behind an 8 pixels image with 1 byte per pixel, whose binary
values are shown below:
10010101
00001101
11001001
10010110
00001111
11001011
10011111
00010000
In this example, the bits with bold face are the least
significant ones. Let the letter ‘A’ be represented by ASCII
code, whose value is 01000001. Each bit in the code
(underlined) will replace those marked above, resulting in
another camouflage stream:
Journal of Telecommunication, Electronic and Computer Engineering
24 e-ISSN: 2289-8131 Vol. 9 No. 2-2
10010100
00001101
11001000
10010110
00001110
11001010
10011110
00010001
The merit of replacing the LSBs with those drawn from the
message is that the resulted stream does not significantly
differ from the original one, effectively hiding the message in
a plain sight. Moreover, the amount of data (number of
encoded bits) remains essentially the same.
D. Video
Video is a multimedia file which represents motion pictures
with embedded sound. There are several types of video, e.g.,
educational video, entertainment video, and commercial
video etc. A video file consists of three types of information
as follows.
Image, whose dimensions characterized by its width and
height in pixels unit. Audio, which can be characterized by its
duration (second or sec), bit rate (kbps) and encoding format
(such as mp3, wma, and wav). Video, which can be
characterized by its frame rate as a speed of image sequence
(fps), data rate as the number of bits that used to represent a
motion picture in a unit of time (kbps), video sample size as
the resolution each image in the sequence (bps), and video
compression as a method used to encode the video into binary
data.
E. RGB Color Representation
RGB is a basic computer color representation. It describes
each picture element or pixel with its color components, i.e.,
red (R), green (G), and blue (B). A combination of these color
components creates different true colors. The number of
colors that can be rendered using this format depends on the
precision of each component. Normally, a personal computer
stores each component in an 8–bits memory, resulting in 256
possibilities for each basis color or 16 million combinations
in total [4].
F. Audio Video Interleave (AVI) Video Format
AVI, standing for Audio Video Interleave, is a movie file
that can be played simultaneously both motion pictures and
audio through a computer. There are a number of computer
programs that support this format, including Windows Media
Player and Quick Time. AVI is considered a standard format
for a computer running Windows operating systems. It
contains very image and sound resolution, and therefore is
normally large. There are variations of this format, i.e., DVD
AVI is suitable for video editing as it best maintains the
original video quality, while Xvid AVI is preferred in
archiving as it is normally six times smaller than the original
but with slight quality drop [4].
III. LITERATURE REVIEWS
This section reviews the preliminary works that addressed
video steganography as follows. In 2013, Yadav et al. [5]
proposed a LSB steganographic method whose advantage
was that the file size did not increase. Its main drawback,
however, was that the locations that messages resided are
systematic, and thus easily predictable. In the same year,
Thakur and Saikia [6] used an 8 x 8 DCT of 128 x 128 which
preprocessed a hidden image by using lossy compression.
Thanks to a unique DCT characteristic, the image
corruption was minimized during the concealment, but it
limits the scope of message types to only that of image. Later,
Dasgupta et al. [7] applied a Genetic Algorithm (GA) and
3:2:2 LSB principles to steganography. In their study, GA
was employed to examine the defects caused by the
concealment in a video file. This resulted in minimal change
between the original and post processed files Moreover, the
noise occurred during the process was eliminated, resulting in
less apparent camouflage.
Similar to others, their technique was still based on LSB
method; and thus it was easily to predict the locations of
hidden data. Bhole and Patel [8] also proposed an LSB
steganography which randomly picked a set of data bytes bits
at which messages were to replace. Compared to previous
techniques, random locations made it harder to guess the
place of hiding. Its limitation, however, is that it was suitable
for text data only. Swathi and Jilani [9] similarly proposed an
LSB based steganography whereby a polynomial expression
was used to determine the alterable bit positions. To avoid the
repetition should it occurred, 2 were added to resulted
position, and the alterable bit were shifted accordingly.
Note that the advantage of this technique is that the added
complexity due to polynomial calculation made the message
harder to extract; but, it can conceal only text message.
Involving higher level of image processing technique, Jain et
al. [10] applied zero-crossing operator to determine the
locations where edges passed on or close to. Those locations
were then chosen as message hiding site. Despite higher
computational demand, this method maintains hiding
location regardless of image size.
IV. METHODOLOGY
The abovementioned techniques, based on LSB, DCT, and
zero crossing has laid a common foundation on which our
work was built. In this research, a video steganography based
on LSB color detection is proposed. Specifically, color values
were first examined in order to determine appropriate LSB
format. Head frame where information was to be hidden was
also specified.
With any LSB format (ratio) chosen, a fixed number of
eight LSB bits per each single RGB pixel were allocated for
concealment. These eight locations differed, depending on
relative proportion of the RGB components, so as to ensure
seamless message-image fusion, and so higher PSNR. This
section, thus, discussed the proposed method in greater detail.
A. System Overview
Figure 1 illustrates the overview of the proposed technique.
In this diagram, the processes of encoding and decoding the
video steganography are depicted. During encoding process,
an uncompressed AVI video was used as input and its size,
and the number of frames were calculated. A head frame
where the concealment occurred was also specified.
Subsequently, the message to be hidden was converted into
binary representation, whose size was computed to determine
concealment capacity.
Note that each color component (Red, Green, and Blue)
was averaged per frame to determine the appropriate LSB
format, specifying the amount of LSB bits in each component
replaceable by binary data. During the decoding process, the
head frame was identified, and the steganogram pattern was
then determined. Given the same LSB positions, the
concealed data were extracted and arranged into the original
message. The detailed implementation of encoding and
decoding processes is provided the next sections.
Video Steganography with LSB Color Detection
e-ISSN: 2289-8131 Vol. 9 No. 2-2 25
Figure 1: Overview diagram showing encoding and decoding processes
B. Encoding Process
This process consists of 7 steps as depicted in Figure 2.
Here, each step in the diagram is described as follows. Firstly,
a non-compressed video file in .AVI format was inputted into
the system. Then, video size and number of video frames
were computed; and the frame capacity that can be allocated
for concealment was given by (WIDTH×HEIGHT×3)/8. The
messages that could be concealed by the proposed method are
text in .TXT format, image in .JPG or .JPEG format, audio in
.WAV format, and also video in .AVI format. Before
performing steganography, the message was converted into a
binary (.BIN) stream, without header or structure, etc., by
using dec2bin command, as illustrated in Figure 3.
Figure 2: Diagram of the encoding process
Figure 3: Acceptable message format and message to binary conversion
Note that the head frame identified in the previous step was
used to store relevant meta data, i.e., number of frames,
amount of concealed data, and concealment format/ pattern.
The subsequent frames were then used to conceal the actual
binary stream and the pattern/ format with respect to each
frame. The number of frames required was computed by
(Length of Input Data + Length of head frame)/ Frame Size,
as shown in Figure 4.
Figure 4: The layout of meta and concealed data in video frames
The next step was to compute the averaged values of color
components (R, G, and B) in each frame, as shown in Figure
5. The steganography was then performed on the least 1, 2, 3,
or 4 bits of each R, G, or B component, respectively,
depending to the selected format (ratio). Specifically, 8 bits
from the binary stream were distributed to R, G, and B
components according to the format of that particular frame.
Finally, all the concealing frames were put together into the
original video sequence.
Figure 5: Computing averaged values of all color components
C. Decoding Process
This process, decoding, consists of three steps as follows.
Similar to the encoding, it first took an uncompressed AVI
video that has the message concealed. The next step was to
extract the head frame and corresponding meta data to
determine the number of frames, the amount of concealed
data, and LSB format in the steganogram. The next
subsequent N frames which contained the actual message
would then be processed accordingly. Once the binary data
were successfully drawn in respective pattern, they were then
rearranged into their original stream, producing the hidden
message (in text, image, audio, or video formats).
D. Color Detection
On examining the colors in each frame, their averaged
values of red, green, and blue components were computed to
determine appropriate concealing pattern in each pixel. Table
1 shows PSNR values of red, green, and blue colors, when
concealing 1 – 4 bits data (R1 to R4; G1 to G4; and B1 to B4),
respectively.
From Table 1, from the intensive evaluation, colors which
had PSNR > 37 would be chosen, and their values fell
between 64 and 255. Thus, the LSB formats were specified
according to color intensities as follow.
Color Intensity > 192 = High (H)
64 > Color Intensity > 192 = Medium (M)
Color Intensity < 64 = Low (L)
The specified range would then be used to decide suitable
Uncompressed Video
Input (Text, Image, Video)
Frame Separation
Binary Conversion
Head Frame
Encode
Decode
Color Averaging
Information Hiding
Decoding
Encoded Video
Information
Extraction
Uncompressed Video
(AVI) Input information
(Text, Image, Audio,
Video)
Control Specification
(Frame Header)
Color Value
Computation and
Hiding Frame Format
Binary Conversion
Number of Frame and
Size Calculation
Information Hiding
dec2bin(data)
First Frame N Frame
Info. Hiding Format
. . . . .
Info. Hiding Format,
#Total Frame,
#Hiding Bytes
Red (R) Green (G) Blue (B)
Average(Red) = ? Average(Green) = ? Average(Blue) = ?
Journal of Telecommunication, Electronic and Computer Engineering
26 e-ISSN: 2289-8131 Vol. 9 No. 2-2
LSB format by refereeing to Table 2. In each format, 8–bits
data will be concealed within 1 pixel, consisting of R, G, and
B components. Except that when all the components have
high or low intensity, the medium and low concealing formats
were chosen, respectively.
Table 1
PSNR of R/G/B color when concealing 1-4 bits data
R1/G1/B1
R2/G2/B2
R3/G3/B3
R4/G4/B4
64
128
129
255
94.47,
N/A,
94.47
94.47,
101.25,
94.47
102.48,
109.30,
102.48
N/A,
114.87,
N/A
67.04,
78.98,
67.04
80.91,
86.51,
80/31
94.47,
88.21,
88.82
105.63,
98.78,
105.63
57.32,
60.40,
54.93
70.17,
72.83,
69.97
75.36,
77.93,
75.20
85.89,
82.89,
85.89
43.13,
45.54,
43.13
54.93,
56.08,
54.89
62.28,
62.24,
62.28
68.90,
67.60,
68.95
Table 2
Concealing LSB format from color detection
Red
Green
Blue
Format
High
High
Low
3-3-2
High
Low
High
3-2-3
Low
High
High
2-3-3
High
Low
Low
4-2-2
Low
High
Low
2-4-2
Low
Low
High
2-2-4
High
Medium
Low
4-3-1
High
Low
Medium
4-1-3
Medium
High
Low
3-4-1
Medium
Low
High
3-1-4
Low
Medium
High
1-3-4
Low
High
Medium
1-4-3
High
High
High
3-3-3
Low
Low
Low
1-1-1
For example, suppose the averaged red is 240.5 which is
greater than 192, and thus was put in High range. Following
this criteria, green and blue whose averages are 63.1 and 190
which are lower than 64 and between 64 and 192,
respectively. The blue and green colors are then put in Low
and Medium ranges, respectively. For this particular image,
the suitable format is thus 4-1-3 (High – Low – Medium),
according to Table 2.
E. Head Frame
Head frame is the first part in a frame that was allocated for
storing relevant meta data and hidden message. In the first
frame, this space was used to store number of steganographic
frames and the amount of data, which will be hidden in the
subsequent frames. More specifically, in the first head frame,
the (1, 1) pixel is reserved for the LSB format. The next 8 bits
are reserved for number of frames involved, and another 14
bits later are used for the amount of data hidden. In the
subsequent frames, their (1, 1) pixels store only the LSB
format used in each particular frame.
V. PERFORMANCE EVALUATION
This section explains the strategies adopted in evaluating
the performance of the proposed technique.
A. Design of Experiments
In our experiments, five types of videos were used to
evaluate the proposed technique. They are stated as follows
[11]:
1. News had a size of 352 x 288 pixels and consisted of
301 frames and its size of this 10 seconds video are 80
MB.
2. Moving Person was 352 x 288 pixels in size, consisting
of 295 frames, and its size of this 9 seconds video are
85.6 MB.
3. Sports had a size of 800 x 600 pixels and consisted of
230 frames, and its size of this 7 seconds video are 315
MB.
4. Landscape was 1280 x 720 pixels in size, consisting of
137 frames, and its size of this 5 seconds video are 316
MB.
5. Animation had a size of 1024 x 575 pixels and
consisted of 129 frames, and its size of this 4 seconds
video are 217 MB.
In this setup, four types of messages were employed as the
hidden data. Each type had 3 different files in the same
format, i.e.,
1. Text consists of Text1.TXT, Text2.TXT and
Text3.TXT whose sizes are 107 KB, 214 KB, and 429
KB, respectively.
2. Picture consists of Pic1.JPG, Pic2.JPG and Pic3.JPG
whose dimensions are 512 x 512, 1160 x 870, and 2048
x 1536 pixels and whose sizes are 89.6 KB, 179 KB,
and 1039 KB, respectively.
3. Voice and Audio consists Audio1.WAV,
Audio2.WAV, and Audio3.WAV whose lengths are 2,
3, and 4 seconds and whose sizes are 55 KB, 534 KB,
and 740 KB, respectively.
4. Moving Picture consists of Video1.AVI, Video2.AVI,
and Video3.AVI whose dimensions and durations are
40 x 22 pixels 6 seconds, 70 x 38 pixels 4 seconds, and
90 x 49 pixels 2 seconds; and whose sizes are 595 KB,
727 KB, and 595 KB, respectively.
In the following evaluations, five different types of videos
were examined, each when concealing four types of
messages, i.e., text, image, audio, and video. The
steganographic streams obtained from the proposed LSB
technique based on color detection were compared against
generic LSB techniques with steganographic ratio of RGB
components of 1:1:1 and 4:4:4, respectively. Specifically,
their performances were assessed by comparing the resulted
streams with their respective originals and by comparing the
number of frames used for concealing.
To this end, four main experimental cases were devised.
The first one was performing steganography on three
different .TXT text files, using five video types (News,
Moving Person, Sports, Landscape, and Animations). The
second, third, and fourth cases were also carried out on the
identical set of videos but doing so on .JPG images, .WAV
audios, and .AVI videos, each of which used three different
files in their respective formats.
B. Measurements
In experiments described here, Peak Signal to Noise Ratio
(PSNR) [12] was employed to assess the performance of the
techniques. Herein, the PSNR was defined as a ratio between
the peak of resulted video stream (Signal) and the difference
between this stream and its original (Noise). The PSNR was
thus expressed as the equation below.
Video Steganography with LSB Color Detection
e-ISSN: 2289-8131 Vol. 9 No. 2-2 27
(1)
In this equation, Max is the maximum pixel intensity of a
frame and MSE (Mean Squared Error) is defined as the mean
of squared difference between resulted and original frames,
i.e.,
(2)
Note that here, m and n is the size of the frames, I and K
are the resulted and original frames, and (i, j) is the pixel
coordinates.
It is worth noting that observing that the acceptable results
are defined as the steganogram with the PSNR not lower than
37.
C. Experimental Results
The evaluations reported herein followed these steps. The
encoding process started by first (1) selecting the type of
message to be hidden (Text, Image, Audio, or Video) and
then browsing the required file (2). The user was then asked
to decide which technique to be applied between auto and
manual selection (3).
When choosing auto mode, the LSB format determined by
the proposed color detection would be used, while manual
mode allowed user to set the LSB format (Red: Green: Blue)
(4). Next, the source videos were chosen from 5 available
uncompressed AVI files (5). Encoding command finally
started the steganography using the provided parameters and
data (6). Figure 6 shows the GUI of this process.
Figure 6: Graphic user interface (GUI) of encoding and decoding
On decoding a message from a video, first its type needed
to be specified (1). Then, the encoded stream was chosen and
Decode command was executed, i.e., (5) and (7),
respectively.
Note that the encoded stream in each message-video
combination as described in Section 4 Methodology was
compared with their original. Table 3 shows the PSNR and
number of frames used in the concealing text message in the
videos.
It can be concluded from the Table 3 comparing the LSB
video steganography on text messages with 1:1:1, 4:4:4, and
automatically selected format by color detection that, the
1:1:1 method gave the highest PSNR, followed by color
detection and 4:4:4 LSB methods, respectively. Comparing
the number of frames used revealed that the 4:4:4 LSB
method required the least number of frames to conceal such
messages, followed by color detection and 1:1:1 LSB
methods.
Table 3
PSNR/Number of frames used for text steganography
Video
Size (Bytes)
LSB (1:1:1)
LSB (4:4:4)
LSB (CD)
Akiyo
110,000
69.87/4
56.28/1
60.47/2
Akiyo
220,000
67.45/7
53.26/2
58.70/3
Akiyo
440,000
67.77/13
50.19/4
56.44/5
Foreman
110,000
70.58/3
56.18/1
60.46/2
Foreman
220,000
67.82/6
53.18/2
58.73/3
Foreman
440,000
64.93/12
51.73/3
64.93/5
Aspen
110,000
68.31/1
48.26/1
55.98/1
Aspen
220,000
68.32/1
48.37/1
55.97/1
Aspen
440,000
65.43/2
48.50/1
56.00/1
Minion
110,000
72.29/1
52.13/1
59.82/1
Minion
220,000
72.22/1
52.24/1
59.78/1
Minion
440,000
69.20/2
52.31/1
59.74/1
Sport
110,000
74.81/1
54.46/1
61.52/1
Sport
220,000
71.64/2
54.44/1
61.53/1
Sport
440,000
69.00/3
54.60/1
61.67/1
The same conclusions can be drawn from Table 3 in the
image case, as shown in Table 4, which are that the 1:1:1 LSB
method produced the highest PSNR, and that the 4:4:4 LSB
method needed the least number of frames. To further
emphasize these findings, the steganography was also
performed on audio and video messages. Their PSNR and
frame counts were shown in Tables 5 and 6, respectively.
Table 4
PSNR/Number of frames used for image steganography
Video
Size (Bytes)
LSB (1:1:1)
LSB (4:4:4)
LSB (CD)
Akiyo
110,000
71.14/3
56.13/1
63.86/1
Akiyo
220,000
68.13/6
53.12/2
60.89/2
Akiyo
440,000
61.01/31
47.45/8
53.06/12
Foreman
110,000
70.58/3
56.02/1
63.87/1
Foreman
220,000
68.55/5
52.92/2
60.87/2
Foreman
440,000
61.32/28
48.13/7
53.48/11
Aspen
110,000
68.30/1
48.24/1
56.04/1
Aspen
220,000
68.31/1
48.34/1
56.11/1
Aspen
440,000
62.43/4
49.00/1
53.30/1
Minion
110,000
72.33/1
52.19/1
59.89/1
Minion
220,000
72.31/1
52.26/1
59.90/1
Minion
440,000
65.28/5
49.44/2
57.16/3
Sport
110,000
75.04/1
54.91/1
62.00/1
Sport
220,000
72.17/2
54.80/1
62.13/1
Sport
440,000
67.00/6
51.84/2
57.35/3
Table 5
PSNR/Number of frames for audio steganography
Video
Size (Bytes)
LSB (1:1:1)
LSB (4:4:4)
LSB (CD)
Akiyo
110,000
72.89/2
55.76/1
63.65/1
Akiyo
220,000
63.88/16
50.00/4
55.86/6
Akiyo
440,000
62.50/22
48.39/6
54.77/9
Foreman
110,000
72.12/2
55.52/1
63.53/1
Foreman
220,000
63.98/15
49.96/4
55.89/6
Foreman
440,000
62.75/20
49.28/5
54.77/8
Aspen
110,000
68.28/1
48.19/1
55.99/1
Aspen
220,000
65.41/2
48.52/1
56.23/1
Aspen
440,000
63.67/3
48.68/1
56.34/1
Minion
110,000
72.35/1
52.16/1
59.84/1
Minion
220,000
67.52/3
52.46/1
60.10/2
Minion
440,000
66.25/4
52.51/1
57.00/1
Sport
110,000
75.09/1
55.26/1
62.20/1
Sport
220,000
69.06/4
55.49/1
59.42/2
Sport
440,000
67.82/5
51.96/2
59.11/2
Journal of Telecommunication, Electronic and Computer Engineering
28 e-ISSN: 2289-8131 Vol. 9 No. 2-2
Table 6
PSNR/Number of frames for video steganography
Video
Size (Bytes)
LSB (1:1:1)
LSB (4:4:4)
LSB (CD)
Akiyo
110,000
63.37/18
48.54/5
54.86/7
Akiyo
220,000
62.50/22
48.04/6
53.95/9
Akiyo
440,000
63.37/18
48.72/5
55.01/7
Foreman
110,000
63.45/17
48.42/5
54.90/7
Foreman
220,000
62.75/20
48.85/5
54.51/8
Foreman
440,000
63.70/16
49.72/4
55.73/6
Aspen
110,000
65.40/2
48.30/1
55.99/1
Aspen
220,000
63.66/3
48.49/1
56.14/1
Aspen
440,000
65.40/2
48.39/1
56.10/1
Minion
110,000
67.54/3
52.26/1
56.80/2
Minion
220,000
66.27/4
52.49/1
56.89/2
Minion
440,000
67.53/3
52.39/1
56.89/2
Sport
110,000
69.02/4
56.26/1
59.46/2
Sport
220,000
67.99/5
52.78/2
59.60/2
Sport
440,000
68.98/4
56.51/1
59.54/2
It is evident that in all cases the PSNR is acceptably high.
Although the 1:1:1 LSB offered highest PSNR, it required the
most number of frames to conceal the messages, compared to
4:4:4 LSB and color detection methods. On the contrary, the
4:4:4 LSB required the least number of frames, at the cost of
scarifying PSNR in the resulted streams. It is therefore safe
to note here that, LSB format automatically determined by
color detection offered the optimal compromise between
PSNR and number of frames required to conceal the
messages.
To elucidate the above observation, the graphs comparing
PSNR of resulted steganography using the three LSB
methods on different types of message in each video are
shown in Figure 7 (a to e). It can be clearly seen in these
graphs that 1:1:1 LSB gave consistently high PSNR values,
regardless of message types. The LSB by using color
detection offered slightly lower PSNR but still higher than the
4:4:4 LSB. It is worth noted that the differences are small
when concealing video messages.
VI. CONCLUSION AND FUTURE WORK
This research proposed the use of color detection in the
LSB bits (LSB Color Detection) to adjust ratios of concealing
data according to respective pixel RGB values.
With this technique the number of bits replaced by
steganogram increased from that was possible in the
conventional 1:1:1 method, effectively improving the
steganographic performance. In addition, when compared
with the 4:4:4 LSB method, the proposed LSB color detection
generated video steganography with higher PSNR.
When considering the number of frames needed to encode
these messages, it was found that color detection based LSB
can effectively conceal a similar amount of data compared to
4:4:4 LSB, given a specific number of frames. However,
unlike 4:4:4 LSB, it could do so without having to scarify the
PSNR.
(a)
(b)
(c)
(d)
(e)
Figure 7: Encoding performance (PSNR) with different file formats: (a)
Akiyo, (b) Foreman, (c) Aspen, (d) Minion, and (e) Sport
REFERENCES
[1] it24hrs. 2014. The shift in hacker attack behavior so as to gain higher
reward. www.it24hrs.com/2014/data-breach-impact.
[2] Gesengues, A. 2013. Study: Number Of People Who Share Or Post
Videos Online Has Doubled Since 2009, Pew Research Center.
[3] Pavani, M. Naganjaneyulu, S. and Nagaraju, C. 2013. A Survey on LSB
Based Steganography Method. Int. J. Of Engr. and Comput. Sci. 2(8):
2464–2468.
[4] Hunt, R.W.G. 2004. The Reproduction of Colour. Wiley–IS&T Series
in Imaging Sci. and Technol., 724 .
[5] Yadav, P. Mishra, N. and Sharma, S. 2013. A Secure Video
Steganography with Encryption Based on LSB Technique,” Proc. IEEE
Int. Conf. on Comput. Intell. and Comput. Research. 1–5.
[6] Thakur, V. and Saikia, M. 2013. Hiding Secret Image in Video. Proc.
Int. Conf. on Intell. Syst. and Signal Process. 150–153.
[7] Dasgupta, K. Mondal, J. K. and Dutta, P. 2012. Hash based Least
Significant Bit Technique for Video Steganography (HLSB), Int. J. of
Sec., Privacy and Trust Management. 1(2): 1–11.
[8] Bhole A. T. and Patel, R. 2012. Steganography over Video File using
Random Byte Hiding and LSB Technique, Proc. IEEE Int. Conf. on
Comput. Intell. and Comput. Research. 1–6.
[9] Swathi A. and Jilani, S.A.K. 2012. Video Steganography by LSB
Substitution Using Different Polynomial Equations. Int. J. of Comput.
Engr. Research. 2(5): 1620–1623.
[10] Jain, N. Meshram, S. and Dubey, S. 2012. Image Steganography Using
LSB and Edge–Detection Technique. Int. J. of Soft Comput. and Engr.
2(3): 2231–2307.
[11] Video Dataset (Producer, Marin Kocs, Nbadleague). 2015. Akiyo,
Foreman, and Aspen [media.xiph.org/video/derf]; Funny Video 3D and
Nbadleague [www.youtube.com]
[12] Huynh-Thu, Q. and Ghanbari, M. 2008. Scope of validity of PSNR in
image/video quality assessment. Electro. Letters. 44(13): 800–801.
Text
Image
Audio
Video
0
20
40
60
80
LSB (1:1:1)
LSB (4:4:4)
LSB (CD)
PSNR
Text
Image
Audio
Video
0
20
40
60
80
LSB (1:1:1)
LSB (4:4:4)
LSB (CD)
PSNR
Text
Image
Audio
Video
0
20
40
60
80
LSB (1:1:1)
LSB (4:4:4)
LSB (CD)
PSNR
Text
Image
Audio
Video
0
20
40
60
80
LSB (1:1:1)
LSB (4:4:4)
LSB (CD)
PSNR
Text
Image
Audio
Video
0
20
40
60
80
LSB (1:1:1)
LSB (4:4:4)
LSB (CD)
PSNR
