IEEE-20180
A Novel Copyright Protection Scheme based on Visual Secret Sharing Surekhl1,Dr GN Swamy*2, Dr KRama Linga Redd/3
B
Associate Professor, Department of ECE, TRR College of Engineering, Patancheru, Hyderabad, Andhra Pradesh, INDIA
-
502 3i9
[email protected] *
Professor, Department of ECE, VR Siddhartha Engineering College, Kanuru, Vijayawada, Andhra Pradesh, iNDiA
-
520007
2gavini
[email protected] I
Professor and HOD, Department of ETM, G.Narayanamma Institute of Technology
&
Science
Shai/pet, Hyderabad, INDIA -500008
[email protected]
Abstract-
In this paper the concept of Visual Secret Sharing
Visual Secret Sharing (VSS) is an interesting idea
and Discrete Wavelet Transform (DWT), are used to split the
proposed by Naor and Shamir [I] to share a secret binary
watermark into two random binary images called shares. One share is generated during watermark embedding phase and is kept secret with an arbitrator. The other share can be extracted from the controversial image whenever needed. Both the shares can be combined to decode the original watermark. The proposed scheme aims at re ducing the size of shares, improving the quality of extracted watermark, and decreasing the probability of false positives. Hence it offers
image among be recovered
n
participants such that the secret image can
without
any computations. A codebook
involving pixel expansion is used to split a secret image into
n
random looking images called shares. These shares
are printed onto transparencies and are distributed to the
n
participants. Simple stacking of all these transparencies at a later time can visually recover the secret.
better security when compared to the related schemes. The
Recently, many researchers used the concept of basic
simulation results reveal that the proposed scheme is robust
VSS with two participants to protect the copyrights of
to several image processing attacks.
digital images. The VSS based watermarking techniques defmed in [2]-[4] embeds the secret share in frequency
Keywords-
Copyright Protection, Digital Watermarking,
Discrete Wavelet Transform, Visual Secret Sharing.
domain version of the host image to result high quality watermarked images and to have good resilience to attacks. These techniques are sim ilar to traditional watermarking
I.
techniques in the way they satisfy the performance criteria.
INTRODUCTION
For example high robustness is achieved at the cost of
Due to rapid growth in computer and communication
sacrificing the payload and imperceptibility.
technologies, a variety of multimedia data is available on
To resolve this tradeoff, researchers in [5]-[ 12] have
the internet for open access. Many multimedia agencies are
proposed another category of VSS based watermarking
now depending on the internet for their commercial
schemes which doesn't alter the host image at all. Before
purposes. The ultimate availability of their data on the
publishing the host image these schemes first construct one
internet leads to data piracy. Therefore many people are
random image called an ownership share using the original
interested in protecting their intellectual data by adapting
watermark image, a secret key and some features extracted
copyright
protection
techniques.
Digital
image
from the host image. The ownership share is confidentially
watermarking is one such technique. In this process, before
time stamped and is kept secret with the arbitrator. During
an image is published, the copyright information of the
verification, a second random image called verification
owner
share is extracted from the controversial image using the
is
embedded
into
it
by
using
embedding algorithm
and
a secret
a
watermark
key. Whenever a
same secret key. The verification share is then overlaid on
controversy happens, the watermark can be extracted from
the ownership share to extract the watermark. Due to some
the controversial image using the same secret key and a
kind of pixel expansion used in creating shares the
watermark
extracted watermark have large size and low contrast.
extraction
algorithm.
The
watermark
embedding and extraction algorithms should be designed,
Since shares are needed to be saved by the arbitrator for
such that the hidden watermark can't be tampered after
copyright verification, the actual size of the shares should
attacks on the copyrighted image. The term robustness in
be as small as possible. This problem is more pronounced
watermarking
Other
when multiple watermarks are to be embedded into the
watermarking
same host image. In this case multiple shares (one for each
important
refers
criteria
to
to
this
be
attack
satisfied
resilience. by
algorithms are simplicity, imperceptibility, high capacity
owner) are to be generated and are to be registered with the
and security.
arbitrator. If the shares are large, it becomes heavy burden for the arbitrator to store all the shares. Hence, it is
ICCCNT'12 26 th_28th July 2012, Coimbatore, India
IEEE-20180
important to reduce the size of the shares as much as possible. In this paper, a novel watennark hiding scheme based on modified VSS and Discrete Wavelet Transfonn (DWT) is proposed. This scheme works in the same way as specified in the previous paragraph. When compared to the related
works,
the
proposed scheme has
(a)
three main
(b)
advantages. 1) The quality of extracted watennark is good. 2) The scheme results in no pixel expansion while creating the shares. 3) It reduces the probability of false positives. In addition to these advantages, the proposed scheme achieves
high robustness to several
image processing
attacks. The rest of the paper is organized as follows. In Section 2, we give the preliminaries needed in understanding the proposed watennarking scheme. Section3 describes the proposed
watermark
Experimental
results
hiding and
and
verification
comparisons
are
phases. given
II. VISUAL SECRET SHARING (VSS)
(c) Share2 (d) Decoded image The results of basic VSS with two participants are shown
in
Fig.1.
independent
Here
random
all
the
selection
pixels of
are
coded
by
Thus
no
columns.
infonnation is obtained by observing any group of pixe Is
A. Basic Visual Secret Sharing ( VSS) Scheme 2-o{-2
(d)
Fig. 1 Example of basic VSS scheme (a) Secret image (b) Sharel
in
Section4. Finally, Section 5 concludes the paper.
A
(e)
visual secret sharing is the basic model. It
splits a secret image into two random images called shares. The splitting is done in the following way. For each pixel in the secret image, a code consisting of four sub-pixels is substituted in each of the shares using a codebook given in Table 1. Each code has half white and half black sub pixels, independent of whether the corresponding pixel in the secret image is black or white. A white pixel is shared into two identical codes. A black pixel is shared into two complemental)' codes. While creating the shares, if the given pixel p in the original image is black, then the encoder randomly chooses one of the first two columns of Table.1. If the given pixel p is white, then the encoder randomly chooses one of the last two columns of Table. 1. The security of VSS lies in the random selection of these columns. To achieve high security the probability of selecting each column, for either pixel color should be same.
on each share. To decode the secret image, each of these shares has to be xeroxed onto transparent sheets. Stacking both these sheets will reveal the original secret. When the two shares are overlaid one above the other, as in Fig.I.d, the black pixels in the original image remain black and the white pixels turns gray. This contrast loss is due to the process of pixel expansion during share creation.
B. Modified Visual Secret Sharing ( VSS) Scheme In the basic VSS, each pixel in the original image is replaced with a code consisting of four sub-pixels in each share. This increases the size of each share to four times the original image and results in loss of contrast in the recovered proposed
image.
To
watermarking
overcome scheme
these
problems
exploits
a
codebook which is shown in Table 2. Here the codes are one pixel size and the decoding operation used is logical XOR. The results of using a modified codebook are shown in Fig. 2. Note that the decoded image in Fig. 2d is exactly
TABLE I
same in size and contrast as the original secret image.
CODEBOOK OF BASIC VSS SCHEME
TABLE 2
Pixel
White
CODEBOOK OF MODIFIED VSS SCHEME
Blac k
Pixel Prob.
Share I
Share2
Sharel Share2
+
the
modified
50%
50%
50%
50%
� � � � � � � � � � II II
Black
White
Prob.
50%
50%
50%
Sharel
•
D
•
D
Share2
•
D
D
•
EB
D
D
•
•
Sharel Share2
ICCCNT'12 26 th_28th July 2012, Coimbatore, India
50%
IEEE-20180
Pixel in watermark
(a)
Black
White
1
Element in Y matrix
0
Ownership share
•
D
0
1
D
•
(b)
B. Watermark Extraction Phase The inputs to the watennark extraction algorithm are a gray level controversial image and the ownership share image. The corresponding output is a binary watennark image.
(c)
To extract a secret binary image Y from the features of
(d)
controversial image, the owner uses the same process and secret key which is used in the watennark hiding process.
Fig. 2 Example of modified VSS scheme (a) Secret image (b) Sharel
A verification share is constructed using the rules given in
(c) Share2 (d) Decoded image
Table4.
III. THE PROPOSED WATERMARKING SCHEME Un like
traditional
watennarking
schemes,
the
watennark is not embedded physically into the digital image.
Instead,
the
proposed
method
A
bitwise
Exclusive
OR
constructs
an
verification share and the extracted watennark is exactly
TABLE 4
VERIFICATION SHARE CONSTRUCTION RULES
A. Watermark Hiding Phase
Element in Y matrix
0
Verification share
•
The inputs to the watermark hiding algorithm are a gray and a binary watennark
then
same since there is no pixel expansion.
of the host image to hide a binary watermark.
H
is
the secret. Note that, the size of the ownership share,
ownership share and a verification share from the features
level host image
operation
perfonned on ownership and verification shares to reveal
B.
The
1
D
corresponding output is a binary image called ownership share. The size of the ownership share is same as the size IV. SIMULATION RESULTS
of watermark image. The host image
H is
first decomposed into} DWT levels
such that the size of each sub-band in the lowest DWT level is larger than the size of the watennark. A secret key
K
is used as a seed to generate a random list of
positions from the LL} sub-band, where watermark.
A
feature
image
P
F with
P
pixel
is the size of the
P
elements
is
constructed, by performing the average on a window of size 5x5 pixels centred at each selected pixel position Let
M be
Pi.
the population mean obtained by averaging all
the sample averages in the feature image.
For each
element in the feature image, if the sample average value is greater than or equal to the population mean
M,
then assign
logic I as the corresponding element in a binary matrix Y, otherwise assign logic
O.
To
verify
the
effectiveness
of
the
proposed
watennarking scheme a sequence of simulations were perfonned using MA TLAB Image Processing Toolbox. Fig.3 (a)-(c) shows three gray level test images (512x512) which are selected as host images. The binary watennark image to be hidden into the host image is of size 150x200 and is shown in Fig.3 (d). For the host image shown in Fig. 3(a) the resultant ownership share, verification share and the
extracted
watermark
for the proposed watermark
hiding scheme are shown in Fig.4.a-c. Note that the size of all these images is same as the original watermark. Also the extracted watennark quality is exactly same as the original watermark.
Finally an ownership share is constructed using the rules given in Table 3. This ownership share is time-stamped and is confidentially kept secret at arbitrator. During verification of copyright, the owner should provide the same secret key to the arbitrator, to retrieve another share called verification share. When this share is overlaid on the ownership share, the watermark can be revealed.
(a)
(b)
(c)
Fig. 3 Test images used in simulations
TABLE 3
OWNERSHIP SHARE CONSTRUCTION RULES
ICCCNT'12 26 th_28th July 2012, Coimbatore, India
(d)
IEEE-20180
The effectiveness comparison of the proposed scheme with some similar VSS-based watermarking schemes in the literature is given in Table 6. The comparison is mainly focused on the following properties: pixel expansion, extracted watermark quality and the probability of false
(a)
positives or indirectly security. Note that the scheme is
(c)
(b)
ideal if there is no pixel expansion, no contrast loss and if the probability of false positives is less. A false positive is
Fig. 4 Results of simulations (a) Ownership share (b) Verification share (c) Extracted watermark Peak Signal to Noise Ratio Correlation
a result of extraction of a watermark from an unauthorized image, which doesn't actually belong to the owner. Since, false positives encourage the malicious owners in claiming
(PNSR)
and Normalized
other unauthorized images, this problem should be avoided.
are used to evaluate performance of this
The proposed method of watermark hiding utilizes sample
watermarking scheme. PSNR is used to evaluate the
averages and central limit theorem to reduce false alarm
similarity of original and attacked gray level images. It is
probability.
(NC)
defined in terms of Mean Square Error
PSNR =10xlog
2 255
(MSE)
as follows:
From the results in Table 6, it is clear that only the proposed scheme satisfies all the criteria mentioned above.
--
MSE
V. CONCLUSIONS
(1) r
MSE =
and
H'iJ
c
1 -2: 2: (Hi,) _H\})2 rxc
(2)
Where
The concept of Visual Secret Sharing (VSS) is used to
HiJ
hide a digital watermark into Discrete Wavelet Transform of a host image. The features of the image are used to split
i�l }�l
the watermark into two random binary images called
denotes pixel color of original host image
denotes a pixel color of attacked watermarked
image, and
r x c
denotes the image size.
Normalized Correlation
(NC)
is used to measure the
similarity between the original and extracted watermark. It is defined as follows: w
=
One
share
during
watermark
other share is extracted from the controversial image combined to extract the original watermark. Unlike, the existing schemes, the proposed scheme don't expand the In the proposed watermarking method, the shares hide the
watermark without altering the host image. The simulation
_i_�l---,.l�_l
results reveal that the proposed scheme has satisfactory
______
robustness to several image processing attacks.
wxh
Bi,i
generated
during watermark extraction phase. Both the shares are
(3) Where
is
embedding phase and is kept secret with an arbitrator. The
shares. It also results in high quality extracted watermarks.
h
2:2: (Bi,) ffi B'i,)
NC
shares.
denotes pixel color of extracted w atermark
image from the original host when it is not altered and
B'iJ
denotes a pixel color of extracted watermark image when it is altered. To check attack resilience of the proposed scheme, some common image processing attacks were performed on the three host images. All the attacks were performed using MA TLAB software. The corresponding PSNR and NC values are listed in Table 5. The JPEG compression attack is performed by compressing the image with a quality factor 10. The blurring, sharpening and median filtering of the images were done with a window of size 3x3. A noise added image is obtained by adding 20% salt and pepper noise to the host image. The scaling of an image is done by first reducing the original host image size from 512x512 pixels to 256x256 pixels, and then zoomed to its original size by means of pixel replication. The cropped image is obtained by cropping 25% of the original host image. To test the robustness against rotation attack the original host image is rotated 10° in counter clock wise direction. From Table 5 it is clear that the proposed scheme results in most NC values closer to 1, indicating satisfactory
REFERENCES [1] M. Naor and A. Shamir, "Visual Cryptography," in Proceedings of Advances in Cryptology : Eurpocrypt '94, vol. 950, Springer,
Berlin, pp. 1-12, 1995. [2] C.S. Hsu and S.F. Tu, "Digital Watermarking Scheme with Visual Cryptography", in Proceedings of The International Multi Conference of Engineers and Computer Scientists, Hong Kong, China, 2008., pp. 659-66, [3] S-F Tu, and C-S Hsu, "Digital Watermarking Method Based on Image Size Invariant Visual Cryptographic Scheme", in IEEE Symposia and Workshops on Ubiquitous, Autonomic and Trusted Computing, 2009, pp.362-366. [4] Nagaraj V. Dharwadkar and B.B.Amberker, "Watermarking Scheme for Color Images using Wavelet Transform based Texture Properties and Secret Sharing", International Journal of Signal Processing, vol.6, no.2, pp. 93-100, 2010. [5] Hwang, R., "A Digital Image Copyright Protection Scheme based on Visual Cryptography," Tamkang Journal of science and Engineering, vol.3, no.2, pp. 97 - 106, 2002. [6] C-c. Chang and J-c. Chuang" "An image intellectual property protection scheme for gray-level images using visual secret sharing Strategy," Pattern Recognition Letters, vol. 23, no. 8. pp.931-941,2002. [7] M Hassan A, and M Khalili A, "Self Watermarking based on Visual Cryptography," in proc. of World Academy of SCience, Engineering and Technology, Oct 2005, vol. 8, pp. 159-162. [8] B Surekha , GN Swamy , K Srinivasa Rao , and A Ravi Kumar , "A Watermarking Technique based on Visual Cryptography," Int.
robustness for the given attacks.
ICCCNT'12 26 th_28th July 2012, Coimbatore, India
IEEE-20180
Journal of Information Assurance and Security, voL 4, no,6, pp, 470-473, 2009, [9] D,C Lou, H,I(, Tso, and J,L Liu, "A copyright protection scheme for digital images using visual cryptography technique," Computer Standards and Interfaces, voL29, no,I, pp,125-131, 2007, [10] G'o'Park, EL Yoon, and KY Yoo, "A new copyright protection scheme with visual cryptography," inProc' of the Second Inl, Con! on Future
Generation Communication and Networking Symposia, 2008, pp, 60-63, [II] Y Xing, and JH He, "A new robust copyright protection scheme for digital images based on visual cryptography," in Proc,of the 2010 International Conference on Wavelet Analysis and Pattern Recognition,Qingdao, 11-14 July 2010, pp,6-1 \, [12] Chen TH, Chang CC, Wu CS, and Lou nc, 'On the security of a copyright protection scheme based on visual cryptography', Computer Standards & Inteifaces, voL 31, no,l, pp, 1-5, 2009,
TABLES
SIMULATION RESULTS OF PROPOSED WATERMARKING SCHEME Mandril
Boat
Lena
PSNR
NC
PSNR
NC
PSNR
NC
(dB)
(%)
(dB)
(%)
(dB)
(%)
Attacks JPEG (QF- I O%)
32.93
0.9842
30.42
0.9812
35.36
0.9752
Blurring (3x3)
30.52
0.9532
29.12
0.9427
32.64
0.9654
Sharpening (3x3)
29.61
0.9876
28.29
0.9749
32.54
0.9898
Histogram Equalization
26.97
0.8704
27.57
0.9842
4l.94
0.9630
Median Filter (3x3)
36.42
0.9943
31.69
0.9882
40.71
0.9972
Salt & Pepper Noise
34.05
0.9514
34.12
0.9383
34.20
0.9386
Scaling (1/2)
33.60
0.9843
30.99
0.9752
36.81
0.9876
Rotation (3° left)
29.59
0.8433
28.69
0.8483
30.68
0.8678
Cropping (10%)
36.11
0.8774
36.11
0.8546
36.21
0.9255
TABLE 6
COMPARISON OF VARIOUS SCHEMES
Domain
Pixel Expansion
Contrast loss
Frob, of false positives
Security
Hwang [5]
Spatial
2
Yes
Low
No
Chang and Chuang [6]
Spatial
9
Yes
High
High
Hassan and Khalili [7]
Spatial
2
Yes
Low
High
Surekha and Swamy[8]
Spatial
2
Yes
High
Low
Lou, Tso and Liu[9]
Frequency
2
Yes
Very high
No
Park, Yoon and Yoo[lO]
Frequency
No
No
Low
No
Xing and He[ll]
Frequency
2
Yes
High
Yes
Proposed Scheme
Frequency
No
No
Very low
High
� Schem e
ICCCNT'12 26 th_28th July 2012, Coimbatore, India