a review on: spread spectrum watermarking techniques in the name of god 1/47
TRANSCRIPT
A Review on:
Spread Spectrum Watermarking Techniques
In the name of GOD
1/47
Table of contents
• Introduction• Spread Spectrum communication
Cox method Adaptive watermarking
• Comparison• Conclusion
2/47
Spread Spectrum Method
3/47
Introduction
• The first papers on data hiding appeared in the early 1990s. (LSB) embedding techniques elementary and non-robust against noise.
• The period 1996–1998: The development of Spread Spectrum method
codes• image watermarking (Cox et al. 1997)• Video watermarking (Hartung and Girod 1998 )
more robust and have been used in several commercial products.
4/47
Spread Spectrum Modulation (SSM)
• The watermarking problem is analogous to a communication problem with a jammer.
• motivated many researchers to apply techniques from SSM (successful against jammers.)
• SSM good for military communication systems (secrecy and robustness due to unauthorized person)
• The jamming problem:• Standard radio or TV communication system:
TX sends a signal in a relatively narrow frequency band. Inappropriate in a communication problem with a
jammer. The jammer would allocate all his power to that
particular band of frequencies.
5/47
Spread Spectrum Modulation (Cont’d)
• SSM system: Allocates secret sequences (with a broad frequency
spectrum) to the TX, which sends data by modulating these sequences.
RX demodulates the data using a filter matched to the secret sequences. (The codes used for spreading have low cross correlation values and are unique to every user)
• Attacker must spread jamming power over all D dimensions.• Owner knows which N dimensions are important.
6/47
the message is spread over a wide frequency BW.
The SNR in every frequency band is small (difficult to detect)
Advantages of Spread Spectrum Communication
• Resist intentional and unintentional interference.• Can share the same frequency band with other
users• Protect the privacy, due to the pseudo random code
sequence.
7/47
Spread Spectrum technique for watermarking
• Adopt ideas from spread spectrum communication
• The WM message is spread over a wide frequency bandwidth (spectrum of the host image) Hide a D-dim. signal (information to embed) in an N-dim. space (part of
original document), N >> D
• The SNR in every frequency band is small (difficult to detect) RX knows the place of WM concentrate weak signals to high SNR
output.
• Generate noise like carrier or hopping sequence with cryptographically secure methods (Security )
• if parts of the message are removed from several bands, enough information is present in other bands to recover the message it is difficult to remove the message completely without entirely destroying
the cover (robustness) 8/47
Embedding a Direct-Sequence Spread-Spectrum Watermark
9/47
Recovering a Direct-Sequence Spread Spectrum Watermark
10/47
Secure Spread Spectrum Watermarks for Multimedia
• The first spread spectrum watermarking method based on DCT Proposed by Cox.
• Cox et al. asserted that in order for a watermark to be robust, it need to be placed in the most significant part of the image.
• the watermark will be composed of random numbers drawn from a Gaussian N(0,1) distribution
11/47
Insertion of the watermark
• Insert X (WM) into V (DCT coeffs.) results in V’• 3 natural formula for computing V’:
• The general form:
• Do not provide a solution for how to compute in order to maximize the robustness of the watermark. (set =0.1)
• How choose N?
i
i i i
xi
i i i
i
v ' v (
v '
v
1 x )
v x
v' ea
= +
=
=
a
+a
ii i iv ' (1 x )v +a=
12/47
Informed Watermarking Method of Cox et al.
13/47
Secure Spread Spectrum Watermarks for Multimedia
• Evaluating the similarity of the watermark The extracted watermark might differ from the
original watermark.
• one should decide on a threshold T, and compare
• Set T to minimize the false positives and false negatives.
**
**
.
.),(
WW
WWWWSim
*Sim(W, W ) T>
14/47
Threshold Comparison
• Reducing False Positive (FP) error Selecting the appropriate threshold
n
iii wwWW
1
** .. ).,0())(,0(~. **
1
2** WWNwNWWn
ii
)1,0(~),( * NWWSim10
6
2/ 108659.9 d 2
1 2
te t
**
**
.
.),(
WW
WWWWSim
SelectThreshold =6
15/47
Experiments
• Watermarked image
16/47
Compression and cropping attacks
17/47
Uniqueness of the Watermark
0 100 200 300 400 500 600 700 800 900 1000-5
0
5
10
15
20
25
30
Random Watermarks
Sim
ilari
ty
Threshold (T=6)
18/47
Disadvantages
• The need to have the original image to be able to detect the watermark.
• Since the DCT transform is based on the whole image , the transform does not allow for any local spatial control of the watermark.
• Does not provide a maximum use of the human visual system
19/47
Image Adaptive Watermarking
• The image adaptive DCT based on 8x8 DCT framework (Wolfgang et al.) Wavelet (Podilchuk and Zheng)
• Use Just Noticeable Difference (JND) Matrix.
• The JND is derived from image independent frequency sensitivity and image dependent luminance sensitivity and contrast masking.
• This assists in determining the maximum amount of watermark signal that can be tolerated. the goal is to place the maximum strength watermark
sequence (Why?). 20/47
Adaptive Image Watermarking
),(JND.),(),( vuwvuIvuI w
1)HVS models as image Adaptive weights of watermarks:
otherwise
JND),( if
),(
),(.),(),(
vuI
vuI
vuwvuIvuI w
2) HVS models as filters of Perceptually important image components:
otherwise
JND),( if
),(
),(.JND),(),(
vuI
vuI
vuwvuIvuI w
3) A combination of Previous two methods.
21
Podilchuk-Zeng (P&Z) Method
• Watermark Embedding:
• : threshold calculated for each level & frequency orientation by
Watson for 9.7 Daubechies Filters.
otherwise
if
.,,,,
,,
,,,,,,,*,,,
,
TX
X
wTXX vu
vu
vuvu
vu
,T
22
Watermark Detection
• Watermark detection:
• Again use similarity and appropriate threshold which is designed to balance false positives and false negatives.
otherwise
),( if
0),(
ˆ,,,,
,
,,,*
,,,*
,,,
vuJXvuJ
XXw vu
vuvu
vu
23/47
Experiment 1 (uniform image)
24/48
Experiment 2 (non-uniform image)
25/48
Podilchuk method (Attacks)
26/48
Quality 100 80 60 40 20 10 5
Cropping(to 25%)+ Compression 1 0.9671 0.9300 0.9114 0.8843 N/A N/A
cropping(to 6.25%) +compression 1 0.9588 0.9200 0.8988 0.8600 N/A N/A
Compression N/A 0.9680 0.9120 0.8262 0.7050 0.4720 0.3020
SSIS Method
27/47
28/47
Results
29/36
Original
watermarked
JPEG Attack
30/36
Noise Attack
31/36
Conclusion
• Spread spectrum communication• watermarking problem is analogous to a
communication problem with a jammer.• the WM message is spread over a wide
frequency bandwidth (spectrum of the host image) Cox method. Adaptive watermarking
32/36
References
• [1] I. J. Cox, J. Kilian, T. Leighton, T. Shamoon, “Secure spread spectrum watermarking”, in IEEE Transaction on Image Processing, vol. 6. 1997.
• [2] C. I. Podilchuk and W. Zeng, “Image adaptive watermarking using visual models”, IEEE journal on selected areas in communication, vol. 16, No. 4, pp. 525-538, May 1998.
• [3] L. M. Marvel, C. G. Boncelet, C. T. Retter, “Spread Spectrum Image Steganography,” IEEE Trans, on Image process., vol 8, no. 8, Aug. 1999.
• [4] W. Lu, H. T. Lu, F. Chang, “Chaos-Based Spread Spectrum Robust Watermarking in DWT Domain,” in Proc. Int. Conf. on Machine Learning and Cybernetics, Guangzhou, Aug, 2005.
33/36
Any Question?
34/36