Content uploaded by Khaled Elleithy
Author content
All content in this area was uploaded by Khaled Elleithy on May 04, 2014
Content may be subject to copyright.
Abstract— Due to the high speed of internet and advances in
technology, people are becoming more worried about
information being hacked by attackers. Recently, many
algorithms of steganography and data hiding have been
proposed. Steganography is a process of embedding the secret
information inside the host medium (text, audio, image and
video). Concurrently, many of the powerful steganographic
analysis software programs have been provided to unauthorized
users to retrieve the valuable secret information that was
embedded in the carrier files. Some steganography algorithms
can be easily detected by steganalytical detectors because of the
lack of security and embedding efficiency.
In this paper, we propose a secure video steganography
algorithm based on the principle of linear block code. Nine
uncompressed video sequences are used as cover data and a
binary image logo as a secret message. The pixels’ positions of
both cover videos and a secret message are randomly reordered
by using a private key to improve the system’s security. Then the
secret message is encoded by applying Hamming code (7, 4)
before the embedding process to make the message even more
secure. The result of the encoded message will be added to
random generated values by using XOR function. After these
steps that make the message secure enough, it will be ready to be
embedded into the cover video frames. In addition, the
embedding area in each frame is randomly selected and it will be
different from other frames to improve the steganography
scheme’s robustness. Furthermore, the algorithm has high
embedding efficiency as demonstrated by the experimental
results that we have obtained. Regarding the system’s quality, the
Pick Signal to Noise Ratio (PSNR) of stego videos are above 51
dB, which is close to the original video quality. The embedding
payload is also acceptable, where in each video frame we can
embed 16 Kbits and it can go up to 90 Kbits without noticeable
degrading of the stego video’s quality.
Keywords__ Video Steganography, Hamming Code, Linear
Block Code, Security, Embedding Efficiency, Embedding
Payload.
I. INTRODUCTION
NTERNET makes people’s lives much easier than before;
they can use it to pay their bills, buy their goods, exchange
important messages between parties at far distances, and
many other things. Without protecting that valuable
information, attackers can obtain them in different ways.
Steganography is one of the methods that protects and hides
valuable data from unauthorized people and even without
them having any suspicion of the data’s existence. Human
Visual System (HVS) can’t recognize a slight change that
happens in the media cover such as audio, image and video
[1].
There are two important factors that every successful
steganography system should take into consideration, which
are embedding efficiency and embedding payload. First, the
steganography scheme that has a high embedding efficiency
means good quality of stego data and less amount of host
(carrier) data are going to be changed [2]. Any obvious
distortion to the viewers will increase the probability of the
attacker's suspicion and the secret information can be easily
detected by some of the steganalysis tools [3]. These kinds of
schemes are difficult to be detected by the steganalytical
detectors. The security of the steganography scheme is
depending directly on the embedding efficiency [4].Second,
the high embedding payload means the capacity of secret
information to be hidden inside host data is large. To be more
specific, the two factors embedding efficiency and embedding
payload have a type of contradiction. Increasing efficiency
will cause the capacity of embedding to have a low payload.
Changing the balance between these two factors mainly
depends on the users and the type of steganography scheme
[2].
The rest of the paper is organized as follows. Section 2
presents some of the previous work. Section 3 introduces an
overview of the Linear Block Code and Hamming code, and
then presents our Steganography scheme. Section 4 we discuss
experimental results and analyze them. Section 5 provides the
conclusions.
II. RELATED WORK
In 2009, Eltahir et al presented a video steganography
based on the Least Significant Bit (LSB). Authors tried to
increase the size of the secret message into the video frames.
They analyzed video into frames then each frame was used as
a still image. A 3-3-2 approach has been used which means
taking the LSB of all RGB color components (3-bits of Red,
3-bits of Green, and 2-bits of Blue). The reason for taking 2-
bits of blue color is because the HVS is more sensitive to the
A Highly Secure Video Steganography using
Hamming Code (7, 4)
Ramadhan J. Mstafa and Khaled M. Elleithy, Senior Member, IEEE
Department of Computer Science and Engineering
University of Bridgeport
Bridgeport, CT 06604, USA
rmstafa@bridgeport.edu
I
blue color than the other two colors. The results demonstrated
that the hidden message can take one third of overall video
size. This is considered an improvement of the LSB algorithm
[5].
In 2010, Feng et al proposed a novel of video
steganography scheme based on motion vectors as carriers to
embed the secret message in H.264 video compression
processing. The algorithm also uses the principle of linear
block codes to reduce motion vectors’ modification rate. The
algorithm has a good quality of stego data, which is proved by
the low modification rate of motion vectors. The PSNRs that
were obtained in both flower and foreman videos are more
than 37 dB [6].
In 2011, Hao et al proposed a novel video steganography
method based on a motion vector by using matrix encoding. A
motion vector component that has high amplitude among both
horizontal and vertical components is chosen to embed the
secret message. The Human Visual System can see the change
that occurs when the object is moving slowly, while if the
same object moves quickly the HVS won’t be able to feel the
change that happens. Motion vectors with large size are
selected for embedding the secret message. The macro blocks
that are moving quickly will generate motion vectors with
large amplitude. The direction of macro blocks depends on the
motion vectors’ components. For example, if the vertical
component is equal to zero that means the macro block
direction is moving vertically. The quality of the tested videos
that was obtained is more than 36 dB [7].
In 2012, Rongyue et al proposed an efficient BCH coding
for steganography which is embedding the secret information
inside a block of cover data by changing some coefficients.
Authors have improved the computational time of the system
and the complexity becomes low because of the system’s
linearity [8].
In 2013, Liu et al proposed a robust data hiding scheme in
H.264 compressed video stream, where they have prevented a
drift of intra-frame distortion. To give the system more
robustness, authors have encoded the message using BCH
code before making the embedding process. The host data is
the DCT coefficients of the luminance I-frame component.
The obtained results have a high quality and robustness [9].
III. THE PROPOSED STEGANOGRAPHY SCHEME
Our algorithm uses an uncompressed video stream which
is based on the frames as still images. First the video stream is
separated into frames and each frame’s color space is
converted to YCbCr. The reason for using YCbCr color space
is that it removes correlation between Red, Green, and Blue
colors. A luminance (Y) part is brightness data, which the
human eyes are more sensitive to than the color parts. As a
result, the color parts (chrominance) can be subsampled in the
video stream and some information will be discarded.
A. Linear block codes
A block code is a linear block code if a summation of
two codewords is also a codeword, and the binary linear
block code is applied to bits of blocks. An (n, k) binary
linear block code has 2k columns and 2n-k rows in a linear
code array. Where k is refers to k-dimensional subspace
and n refers to n-dimensional vector space.
Vn = {(C0, C1, …, Cn-1)|Cj ϵ GF(2)} where n is the
length of the code and k is a number of symbols. In the
standard array, there are no two equal vectors at the same
row. Assume C is a (n, k) code on Galois Field GF (2),
then:
All X vectors of length n belong to a coset of C.
Each coset has 2k vectors.
Two cosets either overlap or intersect completely
or not at all.
If C+Y is a coset of C and X as belong to (C+Y),
then C+X=C+Y.
B. Hamming codes (7, 4)
The Hamming code is one of the most well-known block
code methods that can do both error detection and correction
on a block of data. In the Hamming code technique, the
original information will be coded by adding some extra data
with the minimum amount of redundancy, which is called the
codeword, of length n bits [10]. The added part consists of
parity information of length (n-k) bits where k is the length of
message that is expected to be coded [2]. In this paper, the (7,
4) Hamming code is used that can detect and correct a single
bit error of data or parity. First, the message (m1, m2, m3, m4)
of length k bits (k=4) is encoded by adding three parity bits
(p1, p2, p3) to become the codeword of length n (n=7), which is
ready for transmission. There are different ways to mix both
types of data (message and parity) together and the general
combination is to put the parity bits at position 2i such as (p1,
p2, m1, p3, m2, m3, m4) where i=0, 1, ... ,(n-k-1).
The Hamming codes are linear codes so they have two
matrices: parity-check matrix H and generator matrix G,
which they need for both encoding and decoding. On the
encoding side, each message M, which consists of 4-bits, will
be multiplied by the generator matrix and then have modulo of
2 applied; the result is the codeword X of 7-bits ready to be
sent through a noisy channel.
Where
On the decoding side, for the purpose of checking the
encoded message of 7-bits R (data + parity) will be received,
then will be multiplied by the transpose of the parity-check
matrix, and taking modulo of 2 again.
, where
The result is a syndrome vector Z (z1, z2, z3) of three bits,
which has to be all zeroes (000) if it’s an error-free message.
Otherwise, any change in the message during transmission
will lead to flipping one or more bits of the message; then it
needs an error correction process.
Example: Assume we have a message M1 of 4-bits (1, 1, 1, 1)
and the Hamming code (7, 4) is done by the following steps:
1- Calculate , the result is (3, 3, 3, 1, 1, 1,
1) then taking modulo of 2, the result is the codeword
X=1111111, which is sent through the
communication channel.
2- At the destination, to get the correct message the
syndrome vector Z must be zero. The first
assumption R=1111111 is received without any
errors. Then Z will become (0, 0, 0), where
.
3- In the second assumption, suppose that during
transmission due to the noisy channel one of the bits
has changed. The received data will be R=1111011,
then calculating the syndrome we will get Z=433, and
taking modulo of 2 the syndrome will become
Z=011.
4- Comparing Z value with the parity-check matrix H, it
appears that the Z value (0, 1, 1) is equal to the 5th
row (0, 1, 1) of the H matrix, which means that the 5th
bit of R has changed.
5- Correcting the 5th bit of R by flipping it to 1, R then
is corrected to become (1, 1, 1, 1, 1, 1, 1).
6- The four first bits are the original message M1 (1, 1,
1, 1) and the last three other bits will be ignored.
C. Data embedding phase
Data embedding is a process of hiding a secret message
inside host videos, and it can be done by the following steps:
1- Convert the video stream into frames.
2- Separate each frame into Y, U and V components.
3- Change the position of all pixels in three components
Y, U and V by a special key.
4- Convert the message (which is a binary image) to a
one dimension array, and then change the position of
the whole message by a key.
5- Encode each 4 bits of the message using Hamming
(7, 4) encoder.
6- The result of the encoded data, which consists of 7
bits (4 bits of message + 3 bits of parity) is XORed
with the 7 bits of random value using a key.
7- Embed the result of those 7 bits in one pixel of YUV
components (3-bits in Y, 2-bits in U and 2-bits in V).
8- Reposition all pixels of YUV components to the
original frame pixel position.
9- Rebuild the video stream again from those embedded
frames.
There are three keys that have been used in this work,
which give to our steganography scheme an improvement in
both security and robustness. Those keys are shared between
sender and receiver in both data embedding and extracting
processes. The first key is used to reposition pixels in Y, U, V,
and the secret message into a random position, which makes
the data chaotic. In order to select the locations for embedding
the secret message into the host data, the second and third
keys are used. They are used to pick the random rows and
columns respectively in each chaotic Y, U and V component.
The XOR function that has been used increases the quality of
the system. The block diagrams of the data embedding phase
and the data extracting phase are illustrated in Figure 1 and
Figure 2 respectively.
Figure 1: Block diagram for data embedding phase.
D. Data extracting phase
Data extracting is a process of retrieving the secret
message from the stego videos which can be done by the
following steps:
1- Convert the video stream into frames.
2- Separate each frame into Y, U and V components.
3- Change the position of all pixel values in the three Y,
U, and V components by the special key that was
used in the embedding process.
4- Obtain the encoded data from the YUV components
and XOR with the random number using the same
key that was used in the sender side.
5- Decode 4 bits of the message by the Hamming
decoder.
6- Reposition the whole message again into the original
order.
7- Convert the message array to 2-D.
Figure 2: Block diagram for data extracting phase.
IV. EXPERIMENTAL RESULTS AND ANALYSIS
In this paper, a database of nine standard Common
Interchange Format (CIF) video sequences is used, with the
size (288 X 352) and the format 4:2:0 YUV. All video
sequences are equal in length with 300 frames in each one.
The secret message is a binary image logo for the University
of Bridgeport (UB) with a size of 128 X 128 pixels. The
MATLAB software program is used to implement this work
and test our experiment results.
In Figure 3, an example of one frame (frame no. 111) in
the Foreman video is chosen. The first part of the figure shows
that the three components of the 111th frame are separated.
Then it shows some locations that have been chosen randomly
for the secret message. The embedded locations are different
in each component inside one frame and they differ from one
frame to next, which mainly depends on the private key. The
second part of the figure shows frame no. 111 before and after
the embedding process. The last part of the figure shows the
whole message that has been embedded and extracted 100%
correctly.
Figure 3: A sample result of frame number 111 for the Foreman video. a)
Shows the selected areas for embedding in YUV components for frame
number 111. b) Shows the 111th frame both the original and the stego frames.
c) Shows the embedded and extracted message.
Figure 4 shows an example of frame no. 222 in the Akiyo
video. The first part of the figure shows the separating of the
YUV video components and also shows the areas that have
been selected for embedding the secret message. The selected
areas in this frame are different from the selected areas in
other frames in the same video, which are chosen randomly by
the private key. This gives the system more security and
robustness against attackers. The second part of the figure
shows the original and the stego frames. The third part of the
figure shows the hidden message before and after embedding.
Figure 4: A sample result of frame number 222 for Akiyo video. a) Shows the
selected areas for embedding in YUV components for frame number 222. b)
Shows the 222nd frame both the original and the stego frames. c) Shows the
embedded and extracted message.
In Table I, the average PSNR for all video sequences is
shown for each Y, U, and V component and all are greater
than 51dB. The quality of the stego videos are mostly the
same as the original videos.
Figure 5 shows the PSNRs of 300 stego frames in the
Mother-daughter video. The quality of the results that have
been obtained from our proposal are very close to the quality
of the original videos before embedding. In general, PSNRs
are greater than 51 dBs, and the V component has a better
quality among the three components.
Figure 6 shows the comparison of visual quality between
nine stego videos. The PSNR of each component, Y, U, and
V, is calculated, of which the average is 300 frames per video.
All the results of PSNRs are between 51 and 52.5 dBs, which
are considered very good results with regard to the purpose of
quality.
V. CONCLUSIONS
In this paper, a secure video steganography has been
proposed based on the Hamming code concepts. The
steganography scheme used frames as still images. It divides
the video stream into frames and then converts the frames to
the YUV format. This algorithm is considered a high
embedding efficiency algorithm due to the low modification
on the host data that makes the stego videos have a very good
TABLE I
THE AVERAGE PSNR OF Y, U, AND V FOR ALL VIDEO SEQUENC ES
Sequences
Frame No.
PSNRY
PSNRU
PSNRV
Foreman
1-100
51.901
51.940
51.920
101-200
51.857
51.924
52.049
201-300
51.817
52.059
52.038
Akiyo
1-100
51.881
51.988
52.431
101-200
51.859
51.978
52.458
201-300
51.859
51.943
52.428
Coastguard
1-100
51.835
51.664
51.854
101-200
51.824
51.682
51.806
201-300
51.823
51.655
51.795
Container
1-100
51.821
52.146
52.067
101-200
51.806
52.117
52.008
201-300
51.785
52.056
51.970
Hall
1-100
51.787
52.084
52.021
101-200
51.797
52.079
52.016
201-300
51.785
52.063
52.005
Mobile
1-100
51.862
52.127
52.065
101-200
51.829
52.076
52.074
201-300
51.834
52.064
52.072
Mother-daughter
1-100
51.686
51.857
51.995
101-200
51.702
51.868
51.992
201-300
51.687
51.876
51.946
News
1-100
52.027
52.167
51.781
101-200
52.012
52.139
51.769
201-300
51.998
52.135
51.764
Stefan
1-100
51.885
52.082
51.961
101-200
51.810
52.111
51.964
201-300
51.848
52.081
51.904
quality. The visual quality is measured by the PSNR and all
the obtained experimental results have a PSNR above 51 dBs.
By having a good visual quality for stego videos, attackers are
not likely to be suspicious. Regarding security purposes our
algorithm is a secure enough to thwart any endeavor by
unauthorized users to retrieve the secret message, even if they
are suspicious of a message’s existence. Security has been
satisfied by having more than one key to embed and extract
the secret message. In addition to the three keys that we have
used, we also encode and decode the message before and after
embedding, which improves the security of our scheme to be
even better.
Figure 5: PSNR of 300 stego frames for the Mother-daughter video.
Figure 6: Comparison between the averages of the PSNR Y, U, and V
components for nine video sequences
With regard to the embedding payload, in each frame the
University of Bridgeport’s logo, which has 16 Kbits, has been
embedded. We have been used all 300 frames in each video
the total number of secret information that we can hide in one
video 300 times 16Kbits (300 X 16 Kbits). Our steganography
scheme can increase the capacity up to 90 Kbits in each frame
with the slight degradation of visual quality. Finally, our
proposal have added new parameters in term of security
purpose to the encoded message of Hamming encoder (7, 4) to
make it more secure before and after embedding process.
REFERENCES
[1] H. Yuh-Ming and J. Pei-Wun, "Two improved data hiding schemes," in
Image and Signal Processing (CISP), 2011 4th International Congress
on, 2011, pp. 1784-1787.
[2] C. Chin-Chen, T. D. Kieu, and C. Yung-Chen, "A High Payload
Steganographic Scheme Based on (7, 4) Hamming Code for Digital
Images," in Electronic Commerce and Security, 2008 International
Symposium on, 2008, pp. 16-21.
[3] L. Guangjie, L. Weiwei, D. Yuewei, and L. Shiguo, "An Adaptive
Matrix Embedding for Image Steganography," in Multimedia
Information Networking and Security (MINES), 2011 Third
International Conference on, 2011, pp. 642-646.
[4] W. Jyun-Jie, C. Houshou, L. Chi-Yuan, and Y. Ting-Ya, "An embedding
strategy for large payload using convolutional embedding codes," in ITS
Telecommunications (ITST), 2012 12th International Conference on,
2012, pp. 365-369.
[5] M. E. Eltahir, L. M. Kiah, and B. B. Zaidan, "High Rate Video
Streaming Steganography," in Information Management and
Engineering, 2009. ICIME '09. International Conference on, 2009, pp.
550-553.
[6] P. Feng, X. Li, Y. Xiao-Yuan, and G. Yao, "Video steganography using
motion vector and linear block codes," in Software Engineering and
Service Sciences (ICSESS), 2010 IEEE International Conference on,
2010, pp. 592-595.
[7] B. Hao, L.-Y. Zhao, and W.-D. Zhong, "A novel steganography
algorithm based on motion vector and matrix encoding," in
Communication Software and Networks (ICCSN), 2011 IEEE 3rd
International Conference on, 2011, pp. 406-409.
[8] Z. Rongyue, V. Sachnev, M. B. Botnan, K. Hyoung Joong, and H. Jun,
"An Efficient Embedder for BCH Coding for Steganography,"
Information Theory, IEEE Transactions on, vol. 58, pp. 7272-7279,
2012.
[9] Y. Liu, Z. Li, X. Ma, and J. Liu, "A Robust Data Hiding Algorithm for
H. 264/AVC Video Streams," Journal of Systems and Software, 2013.
[10] A. Sarkar, U. Madhow, and B. S. Manjunath, "Matrix Embedding With
Pseudorandom Coefficient Selection and Error Correction for Robust
and Secure Steganography," Information Forensics and Security, IEEE
Transactions on, vol. 5, pp. 225-239, 2010.
Ramadhan J. Mstafa
Ramadhan Mstafa is originally from Dohuk, Kurdistan Region,
Iraq. He is pursuing his Doctorate in Computer Science and
Engineering at the University of Bridgeport, Bridgeport, Connecticut,
USA. He received his Bachelor’s degree in Computer Science from
University of Salahaddin, Erbil, Iraq. Mr. Mstafa received his
Master’s degree in Computer Science from University of Duhok,
Duhok, Iraq. His research interests include image processing, mobile
communication, security and steganography.
Prof. Khaled M. Elleithy
Dr. Elleithy is the Associate Dean for Graduate Studies in the
School of Engineering at the University of Bridgeport. He is a
professor of Computer Science and Engineering. He has research
interests are in the areas of wireless sensor networks, mobile
communications, network security, quantum computing, and formal
approaches for design and verification. He has published more than
two hundred fifty research papers in international journals and
conferences in his areas of expertise.
Dr. Elleithy is the editor or co-editor for 12 books by Springer. He
is a member of technical program committees of many international
conferences as recognition of his research qualifications. He served
as a guest editor for several International Journals. Also, he is the
General Chair of the International Joint Conferences on Computer,
Information, and Systems Sciences, and Engineering virtual
conferences.
51
51.2
51.4
51.6
51.8
52
52.2
52.4
1
20
40
60
80
100
120
140
160
180
200
220
240
260
280
300
PSNR
Frame Number
PSNR_Y
PSNR_U
PSNR_V
51.2
51.4
51.6
51.8
52.0
52.2
52.4
52.6
Average PSNR for 300 Frames
Video Type
PSNR_Y
PSNR_U
PSNR_V
