Anda di halaman 1dari 10

A Duel Layer Trans Domain LSB Stego Method:

On Reducing the Data-Hiding Problem Space in Web


Based Teleradiology Stego Systems

1st Author

2nd Author

3rd Author

1st author's affiliation


2nd author's affiliation
3rd author's affiliation
1st line of address
1st line of address
1st line of address
2nd line of address
2nd line of address
2nd line of address
Telephone number, incl. country code Telephone number, incl. country code Telephone number, incl. country code

1st author's E-mail address

2nd E-mail

ABSTRACT
Security for the general patients medical records is an extremely
vital component of any medical information system infrastructure.
This is however even more sensitive in web based teleradiology
which involves online transmission of medical images e.g. X-rays
and Scans (CAT, MRI etc.) for clinical interpretation and
diagnosis. In this paper, an improved Least Significant Bit (LSB)
trans image domain steganographic method that seeks to mitigate
the data-hiding problem space by increasing the steganographic
capacity and robustness without adversely compromising on the
fidelity of the carrier image (steganographic imperceptibility) is
presented and its effectiveness demonstrated.
A Medical Image (IM) that needs to be transmitted to a qualified
radiologist on line for diagnosis purposes is embedded within a
Random Image Signal (RIS). Since the RIS is a secondary carrier
file, up to eight bits of the image can be used, thereby
significantly increasing the payload capacity. The resultant stego
image is then pseudo randomly dissimulated in a second MI
which acts as the primary carrier file. This is to ensure that the
fidelity of the MI is maintained as it is paramount to maintain it
intact for correct diagnosis. A water mark is robustly embedded in
the transform domain of the primary carrier file in order to
enforce tamper detection and nonrepudiation. The resultant stego
image is then remotely transmitted to radiologist who first tests
the primary carrier for tamper detection before decoding the MI
from the original RIS.
The research adopted an experimental research design where the
statistical characteristics of the carrier files from the proposed
method were compared with those produced the traditional LSB
method..
The experiment results returned a PSNR of above 38dB which is
well above the recommeded 30dB for a fairly imperceptible
carrier file. All steganalysis metrics also recorded noticeable
improvements.

3rd E-mail

Keywords Teleradiology,CapacityRobustnes.Staganography
1. INTRODUCTION
The past one decade has witnessed an increase in automation
of medical records by health care providers with the intent of
making the sharing of such information among the various
stakeholders more efficient and reliable. This has birthed new
possibilities and practices in the medical field among them
telemedicine and teleradiology. Teleradiology is a subset of
telemedicine that allows medical images to be transmitted and
accessed remotely over electronic networks for clinical
interpretation and diagnosis (Nyeem et al., 2013). Teleradiology
and telemedicine have great potential in improving health care
delivery and access particularly in sub Saharan Africa due to
shortage of qualified radiologists. (Prior et al., 2012).The
diagnostic skills and expertise of a reviewing radiologist are
however paramount for high quality and professional medical care
to target patients. These skills are however expensive and scarce
resulting to just a few qualified radiologists in these areas.
Teleradiology therefore comes in handy as a solution to the
challenges of cost effectively providing radiology services to
remote clinics and hospitals outside the major metropolitan areas.
According to Madennis (2007), one of the cardinal goal of a
teleradiology system is to provide timely availability of
radiological images and radiologic image interpretation in
emergent and non-emergent clinical care areas and facilitate
radiological interpretation in on-call situations. Also teleradiology
systems are used in facilitating consultative and interpretative
radiology services in areas of need making services of radiologists
available in medical facilities without onsite radiologists support.
This means that highly confidential image data must be shared
across computer networks among medical professionals, and
researchers (Nyeem et al., 2013). However, this can also expose
such images to possible tampering or theft resulting to serious and
costly ramifications when diagnosing ailments.

In order to ensure security in teleradiology systems,


legislative rules that define the security and privacy requirements
of medical information already exist. However, according to
Kobayashi and Furuie (2009), these measures are not capable of
providing the required security for Radiology Information
Systems. Information hiding in digital files therefore present
interesting possibilities and techniques that can be exploited in
improving security in web based teleradiology (Memon et al.,
2011), (Qershi and Khoo, 2010).

This is illustrated in figure 1 based on Fridrichs diagram for the


data-hiding problem space, which depicts the mutually
competitive nature of these parameters.

These techniques however while already well established in a


wide range of other applications are only just beginning to be
explored for healthcare and medical information systems (D. Que
et al, 2014). The two techniques for information hiding have
therefore recently attracted considerable attention in medical
image applications because of their various desirable attributes.
(Das and Kundu, 2011).

1.1 Steganography Requirements


All steganographic methods and algorithms must comply
with a few key requirements. The most important requirement is
that a steganographic embedding method has to be imperceptible.
This means that in order to effectively conceal the existence of
hidden information, it is important that the embedding process
does not produce perceptible distortions to in cover file. Bender
related this concept to the magicians trick of misdirection, which
allows something to be hidden while it remains in plain sight
(Bender, 1996). Imperceptibility also defends steganography
against computer-based steganalysis by preserving certain
statistical characteristics of the cover-medium. It is concerned
with the stego-mediums consistency with the statistical
characteristics of the original cover-file (Fridrich et al., 2001).
Research in Information hiding techniques has established
two more aspects for rating the performance of a stego system.
These include the payload capacity and robustness against image
manipulation. Payload capacity or the embedding capacity was
defined by Lin and Delp as the size of information that can be
hidden relative to the size of the cover file (Lin and Delp, 2009).
It is the amount of secret information that can be hidden in a cover
image without compromising imperceptibility. A good
steganographic algorithm should therefore have high payload
capacity without compromising imperceptibility. Robustness is
the amount of modification that a stego file can withstand before
an adversary can destroy hidden information (Suhud et al., (2014).
There is always a possibility that image manipulation, such as
rotating or cropping, can be performed on the image before it
reaches its destination. Depending on the manner in which the
message is embedded, these manipulations may destroy the
hidden message. It is preferable for steganographic algorithms to
be robust against either malicious or unintentional changes to the
image.
According to (Marvel et al, 2009), capacity and robustness
are very difficult to maximize at the same time while adhering to
high imperceptibility rates. Bender discusses this trade-off
between capacity and robustness as the data-hiding problem space
(Bender, 2012). He outlines that in order to achieve robustness,
redundant encoding of the embedded data on the cover-medium
must be performed, which in turn definitely compromises
capacity.

Figure 1. The data hiding problem space.

1.2 Stego Methods in Medical


Applications
Coatrienx et al, 2013 states that there are at least three main goals
of stego methods in medical image applications. These are:
1.
2.
3.

The authenticity objective that helps to determine the


source of a document
The integrity objective that helps in ascertaining that the
image has not been tampered with while on transit
The data hiding objective which allows the inserting of
the secret data so that the image is useful as a carrier
file.

In order therefore to effectively achieve all the three goals of stego


methods in medical image applications, there is need to strike a
proper balance between capacity, robustness and imperceptibility
in the carrier file.
This research paper proposes an improved trans domain approach
that uses a RIS to embed the patients MI. The MI is embedded in
the spatial domain of the RIS using an enhanced LSB method.
The RIS serves as the secondary carrier file to ensure that the
intactness of the MI is preserved. This is because medical images
meant for diagonosis purposes are sensitive and even the slightest
alteration can result to wrong diagonosis. The resultant stego file
is then dissimulated in the primary carrier file which is another MI
using an enhanced LSB embedding technique. A water mark is
then robustly embedded in the transform domain of this primary
carrier file in order to enforce tamper detection and
nonrepudiation. The final stego file can then be transmitted to a
remote radiologist who them decodes the MI for diagnosis
purposes.
In order to measure the effectiveness of the proposed method, the
study adopted an experimental research design where the
statistical characteristics of the carrier files from the proposed
method were compared with those produced by method. This
comparative experiment was designed to establish the robustness,
the capacity and the imperceptibility of the proposed method
against steganalysis and in comparison to methods that utilize the

traditional LSB method. The experiment results demonstrated


remarkable improvements in the metrics that constitute the datahiding problem space.

2. THE TRADTIONAL LSB METHOD


The traditional LSB method is a common, simple approach
for embedding information in a cover file. It is one of the most
commonly used digital image steganographic methods in image
steganography (Neil and Jajodia, 1998). This method substitutes
the cover images least significant bits with the secret file bits
sequentially until the entire secret file is hidden. It is based on the
idea that since the LSB has a place value of 1, modifying it would
result in a maximum difference of only 1. Because the human eye
is unable to distinguish minimal changes in color, such
modifications would normally be imperceptible in a 24 color bit
depth digital image.
The embedding process consists of choosing a subset {j1,
, j(m)} of cover elements and
0
1
1
performing
the
substitution
0
0
1
operation as follows:
1
1
0
0

Compute index ji where


to store the ith message
bit of m

Sji

LSB(Cj) = Mi (Mi can be either 1


or 0).
(2)

Where:
j

is the cover image bits

is the secret message bits.

The embedding procedure is as


depicted by the pseudo code
below:

Input: Cover C
for i=1 to length (m) do

LSB (Cji) = mi

End for
Output Stego-Object S

In the extraction process, the


LSB of the selected cover-elements are extracted and used to
reconstruct the secret message.

As Bender (1996) explains "To a computer, an image is an array


of numbers that represent light intensities at various points, or
pixels. These pixels make up the images raster data." When
dealing with digital images for use with Steganography, 8-bit and
24-bit per pixel image files are typical. The least significant bit (in
other words, the 8th bit) of some or all of the bytes inside an
image is changed to a bit of the secret message. When using a 24bit image, a bit of each of the Red, Green and Blue color
components can be used, since they are each represented by a
byte. One can therefore store 3 bits in each pixel. An 800 600
pixel image, can therefore store a total amount of 1,440,000 bits
or 180,000 bytes of embedded data (Krenn, R., 2009). For
example a grid for 3 pixels of a 24-bit image could be
represented asshowninfigure8.

Input: Stego-Object S
for i=1 to length (m) do
Compute the jth cover image index where the ith message bit of m is stored
mi = LSB (Cj)
end for
Output Message m

Figure 2. Original Image Bits

When the number 200, whose binary representation is 11001000,


is embedded into the least significant bits of this part of the image,
the resulting grid is as shown in figure 9.
0

Figure 3. Modified Bits


Though the number has been embedded into the first 8 bytes
of the grid, only the three highlighted bits have been changed.
Mostly, only half of the bits in an image will need to be changed
to hide secret data using the maximum cover size. Since there are
256 possible intensities of each primary color, changing the LSB
of a pixel results in small changes in the intensity of the colors.
These changes cannot be perceived by the human eye - thus the
message is successfully hidden.

2.1 The Optimal LSB Insertion Method


This data hiding scheme was developed by Chang and Cheng
and is based an optimal pixel adjustment process to the stego
image obtained by the simple LSB substitution method. The
method improves the stego-image quality by finding an optimal
pixel after performing an adjustment process. Three candidates are
picked out for the pixels value and compared to see which one
has the closest value to the original pixel value with the secret
data embedded in. The best candidate is then called the optimal
pixel and used to conceal the secret data (Jaber et al,. 2014). The
optimal LSB insertion method improves on the quality of the

cover file but the hiding capacity is relatively low. (Suhud et al.,
2014).

2.2 The Pixel Value Differencing


Method
This is a novel steganographic method proposed by (Wu and
Tsai, 2013). The PVD embedding method depends on the images
smooth and edge areas (difference between two consecutive
pixels). A smooth area means a small difference but an edged area
means a large difference. The traditional LSB method is used in
smooth areas while the PVD method is used in edged areas. This
makes the payload of each individual pixel different, and the
resultant stego-image quality is extremely fine with perfect
modification and invisibility. The method can also embed much
more information compared to the optimal LSB method. However
steganalysis is easy as the hidden message is not well spread
across the entire image.

2.3 Blind Hide Method


According to (Bailey and Curran, 2013), this algorithm
blindly hides the secret data in the image starting at the top left
corner of the image and working its way across the image (then
down - in scan lines) pixel by pixel changing the least significant
bits of the pixel colors to match the message. To extract the
hidden information, the least significant bits starting at the top left
are read off. This embedding procedure is not very secure as it's
really easy to read off the least significant bits starting from the
top left corner of the image sequentially.

2.4 Filter First Method


As discussed by Umamaheswari et al. (2014), this algorithm
filters the image using one of the inbuilt filters. It filters the most
significant bits, and leaves the least significant bits to be changed
for the purpose of hiding the secret data. Because the pixels are
changed, there is needed to be careful about filtering the picture
because one may use information for filtering that might change.
If this happens, then it may be difficult (if not impossible) to
retrieve the message again.
To extract the hidden message, the least significant bits
starting at the top left are read off. This is not very secure since
the message is not completely spread across the image resulting to
only a portion of the image being degraded and hence making
steganalysis easy.

2.5 Algorithm Pixel Swap


This method is proposed by Lee et al. (2013). This works as
follows

Randomly select 2 pixels x1 and x2 from the cover


image using a pseudorandom sequence.

If the two pixels lie within a specified distance (=2


or 3 generally), they are suitable for embedding,
otherwise generate another set of pixels.

Take the specific message bit to hide. If the message bit


is zero, check if x1 > x2 otherwise swap x1 and x2 and
hide the bit in the LSB of the pixel. Do the reverse
operation if the message bit is one.

For extracting the hidden message, select the pixels


using the same pseudo-random sequence. Check if the 2
pixels are within the pre-specified range . If x1>x2, the

message bit is zero (one) otherwise the message bit is


one3.(zero).
Figure
Modified Bits
This method does not add visible distortions to the cover
image since only one bit is changed per pixel but its hiding
capacity is highly limited. An implementation of this is Hermatic
stego version 9.3

Data File

None of the methods in the literature studied that proposes


the use of the separate domains of the carrier image or the use of
the two techniques of information hiding in addressing the hiding
problem space.

Table 1. Comparative LSB Results


IMAGE
FILE

SIZE
(kb)

DIMENSION

PSNR
(dB)

SNR
(dB)

Brain.gif

48

286 X 233

59

52

X-ray
lumber.jp
g
Fluorosc
opy.jpg
VesselKnee.jpg

106

537 X 867

56

49

112

472 X 633

56

49

46

244 X 417

59

53

Encoding Module Stego Image

Stego key

Decoding Module

Data File

Transmission Channel

Stego Key

2.6 Adaptive Image Segmentation


Proposed by EL-Emma (2013), this technique explores the
block similarity between a cover-image and the secret data,
specifically an image. Based on the block difference of the two
images, the best match cover image block for the secret image is
selected. The difference-degree and the quantized-error-matrix
between the cover-image block and secret-image block are then
computed. This error is then used to modify the cover-image
block. This scheme provides improved levels of imperceptibility
but its hiding capacity is fairly low.

Stego Image

Cover Image

Figure 4. Conceptual Model.


Dissimulation of the IM utilized the spatial domain of the image.
The message digest (digital signature) of a user supplied password
was used as a seed to help in pseudo-randomly picking the target
image pixels. To further enhance imperceptibility, each pixel was
decomposed into red, green and blue value octets. Each LSB of
each color octet was replaced by one bit of the current MI octet.
The MI octet therefore requires three pixels to be completely
dissimulated. The first two pixels were used to embed the first 6
bits of the EPHI information octet. The two remaining bits were
dissimulated in the third pixels red and blue components.
According to Cachin (2008), human vision is more sensitive to
green color changes and therefore for this pixel the green value
will be left unchanged.
The technique utilizes a 128-bit Message Digest 5 (MD5) hashvalue of the user-given stego key (password) to seed the marsene
pseudo random number generator which is used to pick the
random pixels and color channels in the cover file i.e. to
determine the specific bits in the pixel bytes of the cover image
where data bits of the data file are to be embedded.
Embedding

1.0

3. PROPOSED SOLUTION
This research paper proposes an improved Trans domain approach
that uses a RIS to embed the patients MI. The MI is embedded in
the spatial domain of the RIS using an enhanced LSB method. A
water mark is then robustly embedded in the transform domain of
the RIS in order to enhance the robustness of the carrier file. The
resultant stego image is finally pseudo randomly dissimulated in a
second MI which acts as final cover file for online transmission to
a remote radiologist who them decodes the MI for diagnosis.
Therefore in order to effectively achieve all the three goals of
stego methods in medical image applications, there is need to
strike a proper balance between capacity, robustness and
imperceptibility in the carrier file.

Input
1.1

Data File Cover Image


1.1.1

1.1.2

Process
1.2

Stego Key

1.1.3

Output
1.3

Stego Image

1.3.1

Perform Embedding
1.2.1

The general model framework adopted for the stego system


design is based on the Birgit Pfitzmann generic model

Randomize Embedding
1.2.1.1

Capacity Evaluation
1.2.1.2

Encode Metadata
1.2.1.3

3.2 Extraction Procedures

Figure 5. Hierachical I/O Diagram

3.1 Embedding Procedure


Input

: Cover Image, Secret file (Payload)

Output

: Stego image (image containing hidden file)

Input

: Stego Image, Password message digest

Output

: Secret file

Use Use the Mersenne Twister to


Select a random pixel
Select a random pixel color channel

1. Use the Mersenne Twister to:

Select a random color channel bit

Select a random pixel


Select a random pixel color channel
Select a random color channel bit
2. Let bitToWrite [x][y][channel][bit] denote the selected bit in a
specific color channel for writing
3. Let mi denote the message bit embedded in a color channel bit,
bitToWrite[x][y][channel][bit]
4. For all image color channels:

2. Let bitToRead([x][y][channel][bit])denote the selected bit in a


specific color channel for reading
3. Let mi denote the message bit read in a color channel bit ,
bitToRead([x][y][channel][bit])
4. For all image color channels;
5. If LSB(bitToRead([x][y][channel][bit]) not equal to
mi then
6. Continue

5. If LSB (bitToWrite[x][y][channel][bit]) = mi then

7. If LSB (bitToRead([x][y][channel][bit]) = mi then

6. Continue

8. bitToRead ([x][y][channel][bit])= mi

7. If LSB(bitToWrite[x][y][channel][bit]) not equal to

9. Pack bit in bitSet

mi i then
8. bitToWrite[x][y][channel][bit] = mi
9. while secret file length; Repeat step 1 to 8 to embed
the entire message

10. While secret file length; Repeat step 1 to 9 to read


the entire file
11. Output secret file (M).

10. Output stego image(S)

Stego Key
MI (RIS)
to RIS (Increased Capacity up to 8 bits per color channel usable)
RandomEmbedding
Image Signal

Primary Stego Image

Secondary Stego Image

Extract Primary Stego Image

Original Water Mark

Carrier MI

Primary Stego Image


Stego Key
Embedding Primary stego Image to carrier MI (Improved LSB for better Perceptual Transparency)

Secondary Stego Image

Figure 6. Embedding Diagram

Stego Key

Extract MI

Figure 7. Extraction Diagram

MI

M and N are the height and the width of the image

4. METHODOLOGY
The aim of this study was to measure the effect of using a trans
domain encoding algorithm for information hiding on the data
hiding problem space i.e the significant effect of such an
embedding technique on robustness, imperceptibility and the
payload capacity of the carrier file. Accordingly, an experimental
design represented the best choice for this aim and objective.
According to (Hinkelmann and Kempthorne, 2008), experimental
designs should clearly outline the nature of the problem under
investigation, the type of the experimental design, the
implementation of the experiment, the analysis of the data, and the
interpretation of the results.
Being the most accurate and obvious standard for testing a
hypothesis (Shuttleworth, 2008), the experimental design
methodology was used to achieve the main research objective.
The effect of employing a trans domain technique during the
embedding process represents the variable that was to be studied.
Thus an experiment was carried out to test the relationship
between the specific embedding process ( i.e Proposed method)
and the outcome in the required metric measures ( i.e. in relation
to data hiding problem space).

I is the dynamic range of pixel values, or the maximum value that


a pixel can take, for 8-bit images: I=255.
MSE is the mean square error.

4.3 Signal to Noise Ratio (SNR)


This is an objective quality evaluation metric whose measures are
estimates of the quality of the stego image compared with the
original image. The larger the value of SNR the higher the
imperceptibility level of the embedding algorithm (Mei et al.,
2012).
Equation (1) below is used to calculate the SNR between images:

Where:
Aij Represents one pixel in the original image (before embedding
the data)
Bij Represents one pixel in the stego image (after embedding the
hidden data)

4.1 Evaluation and Discussion of


Results
Objective tests i.e. automated statistical analysis that examines the
statistical properties of an image file were carried out on the stego
images as discussed below. Statistical attacks are crucial in
steganography as they are able to reveal the tiniest modifications
in the statistical properties of an image (Artz, 2011).The following
image quality metrics were employed in these tests

4.2 Peak Signal to Noise Ratio (PSNR)


. PSNR measures the degree of similarity between two images
(similarity of pixels). In the literature, PSNR has shown the best
advantage almost over all other objective image quality metrics
under different image distortion environments and strict testing
conditions (Wang et al., 2002a).It is defined as shown in equations
2.

Table 2. Test Images


FILE NAME

DIMENSIONS

FILE SIZE

Brain.gif

286 x 233

48KB

x-ray lumbar .jpg

537 X 867

106KB

fluoroscopy.jpg

472 X 633

112KB

vessels-knee.jpg

244 X 417

46 KB

2
Table 2. Secondary carrier Images
Where:
FILE NAME
Pulmonary
th

th

Xij is the i row and the j column pixel in the original (cover)
image,
Xij is the ith row and the jth column pixel in the reconstructed
(stego) image,

DIMENSIONS
820 x 720 Pixels

Systen.jpg
Table 2. PSNR and SNR Results

FILE SIZE
113KB

IMAGE

SIZE
(kb)

DIMENSION

Brain.gif

48

X-ray

106

FILE

PSNR
(dB)

SNR(
dB)

286 x 233

61

54

537 X 867

57

50

lumbar.jp
g
fluorosco

112

472 X 633

57

50

py.jpg
vessels-

46

244 X 417

61

55

knee.jpg

5. SECTIONS
The heading of a section should be in Times New Roman 12-point
bold in all-capitals flush left with an additional 6-points of white
space above the section head. Sections and subsequent subsections should be numbered and flush left. For a section head and
a subsection head together (such as Section 3 and subsection 3.1),
use no additional space above the subsection head.

5.1 Subsections
The heading of subsections should be in Times New Roman 12point bold with only the initial letters capitalized. (Note: For
subsections and subsubsections, a word like the or a is not
capitalized unless it is the first word of the header.)

5.1.1

Subsubsections

The heading for subsubsections should be in Times New Roman


11-point italic with initial letters capitalized and 6-points of white
space above the subsubsection head.

5.1.1.1 Subsubsections
The heading for subsubsections should be in Times New Roman
11-point italic with initial letters capitalized.

5.1.1.2 Subsubsections
61
60
59
58
57
56
55
54
53

PSNR (dB)
Traditional
LSB
PSNR (dB)
Proposed
Solution

Figure 8. Peak Signal to Noise Ratio

56
SNR(dB)
Proposed
Solution

54
52

SNR(dB)
Traditional
LSB

50
48
46
1

The heading for subsubsections should be in Times New Roman


11-point italic with initial letters capitalized.

6. ACKNOWLEDGMENTS
Our thanks to ACM SIGCHI for allowing us to modify templates
they had developed.

7. REFERENCES
[1] Bowman, M., Debray, S. K., and Peterson, L.
L. 1993. Reasoning about naming systems.
ACM Trans. Program. Lang. Syst. 15, 5 (Nov.
1993), 795-825. DOI=
http://doi.acm.org/10.1145/161468.16147.
[2] D. Que, X. Wen, and B. Chen(2011) "PACS
model based on digital watermarking and its
core algorithms," in MIPPR 2009 - Medical
Imaging, Parallel Processing of Images, and
Optimization Techniques: 6th International
Symposium on Multispectral Image Processing
and Pattern Recognition, Huazhong University
of Science and Technology; National Natural
Science Foundation of China; China Three
Gorges University.Ding, W. and Marchionini,
G. 1997. A Study on Video Browsing Strategies.
Technical Report. University of Maryland at
College Park.
[3] Kobayashi. L and S. S. Furuie, (2009)
"Proposal for DICOM multiframe medical
image integrity and authenticity," Journal of
Digital Imaging, vol. 22, pp. 71-83.
[4] K. Stine and Q. Dang, "Encryption Basics,"
Journal of AHIMA, vol. 82, pp. 44-44-47,
2011.

Figure 9. Signal to Noise Ratio

[5] Artz, D. (2001) Digital steganography: hiding


data within data, Internet Computing, IEEE,
vol. 5, Issue: 3, pp. 75-80
[6]
[7] Das. S and Kundu (2010)."Lossless ROI
Medical Image Watermarking Technique with
Enhanced Security and High Payload
Embedding," in Pattern Recognition (ICPR),
20thInternational Conference on, 2010, pp.
1457-1460.
[8]
[9] Bender, W. (1996). Techniques for Data
Hiding, IBM Systems Journal, 35(3&4), pp
313-336.
[10] Frhlich, B. and Plate, J. 2000. The cubic
mouse: a new device for three-dimensional
input. In Proceedings of the SIGCHI
Conference on Human Factors in Computing
Systems (The Hague, The Netherlands, April 01
- 06, 2000). CHI '00. ACM, New York, NY,
526-531. DOI=
http://doi.acm.org/10.1145/332040.332491.
[11] Coatrieux. G, L. Lecornu, B. Sankur, and C.
Roux (2013). "A Review of Image
Watermarking Applications in Healthcare," in
Engineering in Medicine and Biology Society,
2011. EMBS
[12] '06. 28th Annual International Conference of
the IEEE, 2006, pp. 4691-4694.
[13] Tavel, P. 2007. Modeling and Simulation
Design. AK Peters Ltd., Natick, MA.
[14] Fridrich, J., Goljan, M. and Du, R. (2001).
Reliable Detection of LSB Steganography in
Color and Grayscale Images. IEEE
Multimedia. 8, 22-28.
[15] Madennis Michael (2009). Security in
Teleradiology Systems . University of Arizona.
[16] Nyeem, Hussain, Wageeh, Boles, & Boyd,
Colin, 2013. A review of medical image
watermarking requirements for teleradiology.
Journal of Digital Imaging, 26(2), pp. 326-343.
[17] Prior. F, M. L. Ingeholm, B. A. Levine, and L.
Tarbox,(2012). "Potential impact of HITECH
security regulations on medical imaging,"
Annual International Conference of the IEEE
Engineering in Medicine and Biology Society.
Piscataway, NJ, USA, 2009, pp. 2157-60.

[18] Forman, G. 2003. An extensive empirical study


of feature selection metrics for text
classification. J. Mach. Learn. Res. 3 (Mar.
2003), 1289-1305.
[19] Marvel, L.M., Boncelet, C.G. and Retter, C.T
(1999). Spread Spectrum Image
Steganography, IEEE Transactions on Image
Processing, 8(8), pp 1075-1083.
[20] Lin E. and Delp E. (2209). A Review of Data
Hiding in Digital Images, Video and Image
Processing Laboratory, School of Electrical and
Computer Engineering, Purdue University.
[21] Brown, L. D., Hua, H., and Gao, C. 2003. A
widget framework for augmented interaction in
SCAPE. In Proceedings of the 16th Annual
ACM Symposium on User Interface Software
and Technology (Vancouver, Canada,
November 02 - 05, 2003). UIST '03. ACM,
New York, NY, 1-10. DOI=
http://doi.acm.org/10.1145/964696.964697.
[22] Suhad Shakir Jaber, Hilal Adnan Fadhil,
Zahereel, Abdul Khalib and Rasim Azeez
Kadhim, (2014). Survey on Recent Digital
Image Steganography Techniques. Journal of
Theoretical and Applied Information
Technology Vol:66 No:3 pp 714-728
[23] Yu, Y. T. and Lau, M. F. 2006. A comparison of
MC/DC, MUMCUT and several other coverage
criteria for logical decisions. J. Syst. Softw. 79,
5 (May. 2006), 577-590. DOI=
http://dx.doi.org/10.1016/j.jss.2005.05.030.
[24] Neil, F. J. and Jajodia, S. (1998). Exploring
Steganography: Seeing the Unseen, IEEE
Computer, 31(2), pp 26-34.
[25] Wang, Z., Bovik, A. C. and Lu, L. (2002a).
Why is image quality assessment so difficult?
IEEE International Conference on Acoustics,
Speech, and Signal Processing, (ICASSP '02),
4, 3313-3316.
[26] Spector, A. Z. 1989. Achieving application
requirements. In Distributed Systems, S.
Mullender, Ed. ACM Press Frontier Series.
ACM, New York, NY, 19-33. DOI=
http://doi.acm.org/10.1145/90417.90738.

Columns on Last Page Should Be Made As Close As


Possible to Equal Length