 Open Access
 Total Downloads : 109
 Authors : R. Surya Prakasa Rao, Prof. P. Rajesh Kumar
 Paper ID : IJERTV6IS060208
 Volume & Issue : Volume 06, Issue 06 (June 2017)
 DOI : http://dx.doi.org/10.17577/IJERTV6IS060208
 Published (First Online): 13062017
 ISSN (Online) : 22780181
 Publisher Name : IJERT
 License: This work is licensed under a Creative Commons Attribution 4.0 International License
Digital Signature based Image Watermarking using GA and PSO
R. Surya Prakasa Rao
Research Scholar, Dept., of ECE,
AU College of Engineering, Visakhapatnam, India.
Prof. P. Rajesh Kumar Professor & HOD, Dept., of ECE,
AU College of Engineering, Visakhapatnam, India.
Abstract: This paper proposed a new digital image watermarking scheme based on Integer Wavelet transform and Singular value decomposition. The proposed approach adopts two metaheuristic algorithms such as Genetic Algorithm and Particle Swarm optimization for optimization purpose. For an enhanced authentication, a new signature generation and signature embedding procedure is also developed in this paper. The proposed signature generation and signature embedding procedure is completely based on the singular values. Various simulation tests are performed to reveal the Excellency of proposed approach and the obtained simulation results illustrates the efficiency of proposed approach.
Keywords: Image watermarking, IWT, SVD, GA, PSO, PSNR, NC.

INTRODUCTION
Recently, the vast increase in the pirated digital media due to the World Wide Web availability and the speed of distribution has led to the need to protect the media against attacks. The digital media (such as video, image, audio or text) can be modified easily by attackers who can then claim its ownership. So, owners, authors, publishers and providers of that media, are reluctant to grant the distribution of their documents in a networked environment [1]. The need to develop robust methods to protect the intellectual property rights of data owners against unauthorized copying, and redistributing it on the network became the main objective of researchers in digital watermarking. Traditional methods such as copy protection or encryption could not solve the problem of unauthorized copying entirely.
Digital watermarking presents a viable solution to that problem by marking the digital media, then it can be easy to be spread and later track it [1]. This can be referred to as a digital signature. The technology used to apply the digital signature on the digital media is called copyright protection. Digital watermarking is described as technologies and methods that hide information sometimes called a signal or watermark; for example, a number or text, into media files such as images, videos, audio, and text. A digital watermark can be visible or invisible to the human visual system. Logos that are often seen added to the corners of images or videos as a way to prevent copyright infringement is an example of visible digital watermark. These watermarks can be easily defeated and removed by replacing or cropping it from the digital media. As a result, the recent effort of research intends to develop watermarking systems to protect the media content. These systems should satisfy the imperceptibility, the robustness, the security and the capacity requirements.
According to the domain used for embedding, the current schemes of digital watermark are basically classified into two types [2]: (i) Spatial domain; (ii) Transform domain. The transform domain involves discrete Fourier transform (DFT), discrete cosine transform (DCT), and discrete wavelet transform (DWT), etc. Many related watermarking schemes have been proposed [3]. In recent years, the watermarking scheme, which is based on singular value decomposition (SVD), became a hot area during the watermarking procedures. In 2002, Liu and Tan [4] first proposed the SVD based watermarking algorithm. In their scheme, the SVD transform is applied on the original image, and the watermark is embedded into the singular value matrix. Kamble et al. [5] presented a digital image watermarking method based on Arnold transform, SVD, and DWT. A watermark is encrypted with a secret key obtained by Arnold transform. The original image is decomposed by twolevel wavelet transform, and the low resolution approximation matrix can be acquired. Then, the SVD transform is applied on the low frequency subband, and the encrypted watermark is embedded into the singular value matrix of the low frequency subband. Fazli and Moeini

proposed a robust hybrid image watermarking scheme using DWT, DCT, and SVD. In this method, the original image is partitioned into four areas: upleft, upright, downleft, and downright. Each area is decomposed by DWT. Then the coefficient matrix HL and LH are partitioned into 8Ã—8 nonoverlapping image blocks. DCT is performed on each block, and the coefficients in the location of (1, 2) and (2, 1) are obtained to organize a new matrix. After applying SVD on the new matrix, the singular value matrix of the new matrix is used as the embedded location of encrypted watermark. Ganic et al. [11] inserted the singular values of the gray scale watermark into the singular values of all one level DWT subbands. Rastegar et al. [9] also suggested a hybrid watermarking scheme. They applied radon transform to the host image, then decomposed it into three levels using DWT. The singular values of the thirdlevel DWT subbands are modified by the singular values of the binary watermark. Lagzian et al. [8] replaced DWT by RDWT and followed the same steps used by Ganic et al. [11]. These schemes performed well against some attacks, but their schemes have a weakness in the security because they applied SVD for the watermark in the embedding process [13, 14, 12]. Lai et al.

scheme could overcome this security issue. They decomposed the image using DWT, divided the watermark into two halves and then embedded each half into the singular values of LH and HL subbands of DWT transform of the
host image. Although, Lai et al. scheme fulfilled the requirements of digital image watermarking, their schemes capacity is still insufficient.
In this paper, we met all the requirements of watermarking, especially robustness, embedding capacity and imperceptibility. Moreover we attempted to overcome the lack of security problem. Various approaches are proposed in earlier to investigate this issue. Loukha et.al [15] and guptha
should be inflated, since they may be less sensitive to the variations within the image.
This can be illustrated through the following concept. Lets consider the image I is denoted as = [1, 2, , ] , where
is a 1 Ã— row vector that represents the ith row of matrix I, then
= 2 (1)
Where
et.al [16] proposed two solutions to solve the security issue in
= =
(2)
image watermarking. The first solution [17] is applying a
I.e.,
=1
oneway hash function to the U and V. This gives the two hashing values HU and HV. Then the singular values are modified based on these hash values. [16] Proposed another solution to this problem. The authors proposed a signature based authentication mechanism for both U and V before continuing the extraction procedure. Based on Loukha and
Guptha, [18] proposed a new solution for solving the security
is the eigenvector of the covariance matrix
corresponding to Eigen value 2, = 1,2, , . Similarly, Lets consider the image I is denoted as = [1, 2, , ] , where is a Ã— 1 row vector that represents the ith Colom of matrix I, then
= 2 (3)
=1
Where
problem. In [18] the SHA1 algorithm was applied on U and
= =
(4)
V to obtain two hashing values. Then a new result (R1) is generated by XORing them. In addition a new secret key is generate with the same dimensions of U and V. The final signature is obtained by XORing the result R1 and the binary version of secret key. A new procedure of singular values modification through the signature embedding procedure. Though this approach achieved an enhanced performance, there is no strict process of secret key generation and also didnt given any clarification about the band selection at signature embedding. One more issue with [18], 1level DWT applied to host image which cant provide much information about the resolution levels of host image.
To overcome the problems with conventional approaches, this paper propose a new intelligent digital image watermarking scheme based on the 3level Integer Wavelet transform (IWT), Normalized Singular Value Decomposition (NSVD) and a new signature generation and embedding procedure. The proposed approach also applies Genetic Algorithm (GA) and particle swarm Optimization for the optimization purpose. The simulation results are tested over various attack types to show the effectiveness.
Rest of the paper is organized as follows; section II gives the basic details about the technologies used in the proposed approach. Section III illustrates the complete details of proposed approach. Section IV gives the details of experimental results and finally the conclusions and future scope is provided in section V.


PRELIMINARIES

Normalized SVD (NSVD)
From the SVD point of view it was noticed that every image matrix has the wellknown SVD for any given single matrix A, the larger Singular Values (SVs) are very sensitive to variations in the image such as noise changes in the host image. Upon the occurrence of attack on the watermarked image, there may be effect on the pixel intensities. But the SVD are very sensitive to these variations. To alleviate the variations in image, a normalized SVD approach is proposed with mainly two ideas such as the weights of host image should be deflated since they are every sensitive to the variations in the image itself and weights of base images corresponding to relatively small
I.e., is the eigenvector of the covariance matrix
corresponding to Eigen value 2, = 1,2, , .
Hence, the should be inflated, since they may be less sensitive to the variations within the image, a new SVD formulation can be derived by modifying the standard SVD evaluation as,
= (5)
Where U, S and V are the corresponding matrices, and is the normalizing constant. In order to achieve the requirements, the needs to satisfy the following condition.
0 1 (6)

Signature Generation Procedure
A new image authentication mechanism based on the several suggestions proposed in earlier [19, 20, 21] is suggested in this paper. The authors revealed that the flaws of the conventional approaches are due to the utilization of U and V matrices which preserves most significant information. If there is an availability of U and V along with the eigenvectors of S, the information loss in the reconstructed image will be less. When applying inverse SVD, eigenvectors play a significant role in the reconstruction process. Thus, any singular matrix S is used along with these eigenvectors, producing a correlated output instead of an actual output. The correlation will be high if the unmatched singular values are approximately equal to the original singular values, and hence, a security weakness arises, namely, the high probability of false positive watermark detection. This security threat is due to unauthorized embedding by an attacker, in which personal eigenvectors are employed in the extraction process to claim a false ownership. Thus, a signaturebased authentication mechanism for the matrices U and V is proposed in this paper to overcome such drawbacks. The generation of signature considers the orthonormal matrices U and V. Secure Hashing Algorithm (SHA) is used here to generate the digests of U and V. By performing a bitwise XOR operation between the obtained digests of U and V an initial result R1 is formulated. Further a random secret key is generated with the same size of the R1. Here the secret key is completely random in nature. For every test case the nature of random key will vary. Then performing the XOR
operation between the random secret key and result R1 derives a new result R2. For the purpose of authentication, the first eight bits or R2 is chosen as a digital signature, sign.

Signature Embedding Procedure
After the signature generation, it is embedded into the watermarked image in such way that it is robust to various attacks and also the degradation in the quality of image is less. The entire watermarked image is not considered for signature embedding. Only some portion of the watermarked image is considered for signature embedding. This some portion is selected based on the correlation between the image pixels. If the signature is embedded in the pixels with low correlation the attacker can break the security easily and can predict the further information based on the on one pixel. Thus the signature is embedded in the pixels with loss correlation such that the quality of I age also wont affected. The process of signature embedding is explained as follows:

Decompose the watermarked image using 1DWT.

Divide the obtained approximation coefficients into 8*8 blocks.

Evaluate the correlation between the coefficients of every block and also the block correlations.

Find the first eight blocks with high correlation.

Perform SVD over every block.

Round off the obtained singular values to its nearest integers after scaling it with the factor of 10.

Modify the singular value according to the signature bits as follows;
If the signature bit is equal to 1 and the modified singular value is even, or if the signature bit is equals to 0 and the modified singular value is odd, add one to the modified singular value and divide the results by a factor of 10.
Else keep as is its.

Rearrange the all modified blocks in their respective positions such that the coefficient mismatching wont occur.

Perform inverse SVD for all the rearranged blocks.

Perform inverse DWT.



PROPOSED WATERMARKING SCHEME
The complete details of proposed signature based watermarking scheme is illustrated in this section. Total process is carried out in two phases, embedding and extraction. The block diagrams of embedding and extraction phases are shown in the figure.1 (a) and (b) respectively.
(a)

Embedding Procedure
(b)
Figure.1 block diagram (a) Embedding (b) Extraction
3. Apply SVD over all the bands obtained.

Choose one host image and apply normalized block processing.

Apply 3level IWT over the normalized host image such that the obtained subbands approximations and details.

Choose one watermark image and apply IWT over it to get subbands followed by SVD.

Modify the singular values of host image by adding the watermark image with singular values after multiplying
= 10 log(2552 ) (10)
them with a watermarking constant, alpha.

Generate a digital signature and embed it in the
= (,)(,)
((,))2
(11)
watermarked image through the process specified in section IIC.

Apply inverse SVD, Inverse IWT and Inverse Block Processing to obtain the signed watermarked image.



Extraction Procedure

Apply IWT on the signed watermarked image such that the obtained coefficients are approximations and details.

Apply SVD over all the obtained bands.

Perform signature extraction form the modified singular values through the procedure exactly opposite to the process specifed in section IIC.

Perform watermark extraction over the obtained watermarked image.

Evaluate peak Signal to Noise Ratio (PSNR) and Normalized Correlation (NC) between the original ad extracted watermark images.

Apply Genetic Algorithm and Particle Swarm Optimization in an iterative fashion until achieving the optimal fitness function. The fitness function is derived from NC as follows
= 1 () (7) Where
The test images considered for evaluation are shown in figure.2, below

(b)


(c) (d)
Figure.2 test imagery (a) Lena (b) Baboon (c) Cameramann (d) Logo
Various types of attacks are accomplished over the watermarked image to reveal the robustness of proposed scheme. To investigate the robustness of proposed approach, the watermarked image was subjected to eight attacks such as: (1) Gaussian noise Attack (GNA)
(2) salt & pepper noise attack (SPA) (3) Median Filtering attack (MFA), (4) Histogram Equalization attack (HEA),
= 1 (, ,)
(8)
(5) Rotation attack (RA) (6) Contrast Enhancemnet attack
=1
(CEA) (7) cropping attack (CA) and (8) Scalling Attack
Where is original watermark and , represents the extracted watermark through the proposed approach
characterized by the position of the jth particle. The smaller fitness value means the better robustness. Here,
signifies the number of attacks,


SIMULATION RESULTS
The proposed waterarking apporach was implemented using MATLAB. To test the proposed apporach various test images are considreed and the size of host image is kept as 512*512 and the watermark image 64*64. The evaluation of the performance of proposed apporach under various circumstances was conducted in terms of imperceptibility and robustness against various attacks. The most widely used criteria are the peak signal tonoise ratio (PSNR) and the normalized correlation (NC), which are employed consecutively. The PSNR is utilized to estimate the imperceptibility, a term used to evaluate the similarity between a host image and a watermarked image, and can be defined as follows:
(SA). The obtained results for both no attack and attack scenarios is represneted below.
(a) (b)
Figure.3 obtained results under no attack secnario (a) Signed Watermarked image (b) Extracted Watermark
In this test case, the signed watermarked image is not subjeced to any attack. The signed watermarked image is passed to extrcation unit as it is without any modification and for the extracted watermark, the performance metrics are evaluated. In the case of attack scenario, the signed watermaekd image is subjeted to attacks an dthe it is passed to extrcation unit to reveal te performance of
= 1
Where
=1
=1
((, ) (, ))2
(9)
proposed approach. Further the obtained rsults of watermarked image and the extracted image are shown in the following figures.
=original watermark image
= extracted watermark image

Gaussian Noise

(b)
Figure.4 (a) Watermarked Image (b) Extracted Watermark


Salt & Pepper Noise

(b)
Figure.5 (a) Watermarked Image (b) Extracted Watermark


Median Filter

(b)
Figure.6 (a) Watermarked Image (b) Extracted Watermark


Histogram Equalization

(b)
Figure.7 (a) Watermarked Image (b) Extracted Watermark


Rotation

(b)
Figure.8 (a) Watermarked Image (b) Extracted Watermark


Contrast Enhancemnet

(b)

Figure.9 (a) Watermarked Image (b) Extracted Watermark

Cropping

(b)
Figure.10 (a) Watermarked Image (b) Extracted Watermark


Scaling

(b)
Figure.11 (a) Watermarked Image (b) Extracted Watermark
Here initially the Cameramann image is embedded in the Lena image and the obtained MSE, PSNR, NC and the SSIM for both attack and noattack cases are represneted in table.1. Similarly the logo image is embedded in the baboon image and its respective results are represented in table.2.
Table.1 Performance metrics for the test case of Lena and Cameramann Image
Attack
Nasrin et.al [18]
Proposed Approach
MSE
PSNR
NC
SSIM
MSE
PSNR
NC
SSIM
NA
2.2772
44.5568
0.9967
0.9822
0.1905
55.3326
0.9974
0.9908
GNA
3.8015
42.3312
0.9852
0.9645
0.4355
51.7412
0.9895
0.9796
SPA
3.6539
42.5032
0.9955
0.9447
0.6319
50.1244
0.9964
0.9586
MFA
5.1345
41.0258
0.9338
0.9412
0.7077
49.6321
0.9383
0.9552
HEA
57.2282
30.5547
0.9258
0.9333
1.7373
45.7823
0.9299
0.9444
RA
185.556
25.4458
0.9752
0.8724
7.5828
39.3325
0.9817
0.8899
CEA
19.9664
35.1278
0.9338
0.9323
2.4614
44.2189
0.9452
0.9598
CA
389.502
22.2257
0.9645
0.8552
60.1817
30.3358
0.9723
0.8645
SA
33.5483
32.8741
0.9947
0.9631
9.5706
38.3214
0.9966
0.9828
Table.2 Performance metrics for the test case of Lena and Logo Image
Attack
Nasrin et.al [18]
Proposed Approach
MSE
PSNR
NC
SSIM
MSE
PSNR
NC
SSIM
NA
1.2018
47.3325
0.9920
0.9813
0.2009
55.1010
0.9962
0.9941
GNA
1.8721
45.5132
0.9824
0.9645
0.4875
51.2513
0.9869
0.9805
SPA
2.5228
44.1120
0.9902
0.9447
0.6338
50.1111
0.9913
0.9586
MFA
3.6079
42.5582
0.9377
0.9423
0.9989
48.9842
0.9593
0.9555
HEA
66.5933
29.8965
0.9238
0.9339
2.1821
44.7421
0.9279
0.9445
RA
84.6702
28.8535
0.9795
0.8878
3.8898
42.2315
0.9804
0.9366
CEA
15.5453
36.2148
0.9373
0.9489
1.7164
45.7845
0.9422
0.9709
CA
172.119
25.7725
0.9712
0.8557
31.6673
33.1247
0.9747
0.8757
SA
26.5601
33.8885
0.9902
0.9689
5.3036
40.8851
0.9923
0.9833
60
Nasrin.et.al[18] Proposed
50
1
0.99
0.98
Nasrin.et.al[18] Proposed
40 0.97
PSNR(dB)
0.96
NC
30 0.95
0.94
20
0.93
10 0.92
0.91
0
NA GNA SPA MFA HEA RA CEA CA SA
Scenario
0.9
NA GNA SPA MFA HEA RA CEA CA SA
Scenario

(b)


Figure.12 Comparitive analysis for the test of Lena with Cameramann image with reference to (a) PSNR (b) NC
60
Nasrin.et.al[18] Proposed
50
1
0.99
0.98
Nasrin.et.al[18] Proposed
0.97
40
PSNR(dB)
0.96
NC
30 0.95
0.94
20
0.93
0.92
10
0.91
0
NA GNA SPA MFA HEA RA CEA CA SA
Scenario
0.9
NA GNA SPA MFA HEA RA CEA CA SA
Scenario
(a) (b)
Figure.13 Comparitive analysis for the test of Lena with Logo image with reference to (a) PSNR (b) NC
Figure.12 illustrates the perfomrance of proposed approach under various test cases for a given Lena host and Cameramann watermark image. The PSNR is observed to be high for the proposed apprach compared to the conventional one for the both attack an dnoattack scenarios. Since there is optimization at both fetaure extrcation leve and also atembedding phase, the quality of image is preserved in an efficient manner. Along with the PSNR, the NC also observed to be high for the proposed approach compared to the conventional apporach.Since the proposed apporach adopts a new signature based authentication mechanism the watermarked image is more robust to all types of atacks. Similarly figure.13 reveals the performance of proposed apporach for the test case of Lena with logo image. For this case also the proposed approach is said to be achiveed better results compare to conventional approach.


CONCLUSIONS
In this paper a new secure imag ewatermarking approach is proposed to enhance the security ofmultimedia images during their transmission. Here the proposed apporach achieved an excellet performance and outperforms the conventionl apporaches. Various test scnearios conducted over various images revealed the enhanced performance of proposed. Approximately the PSNR of proposed apparach is increased by 8dB. Under simulation the proposed apporahc was analyzed by adopting varius attack senarios. From the results the aproposed apporahc is revealing an excellent performance.

REFERENCES

T. Hai, C. M. Li, J. M. Zain, and A. N. Abdalla, Robust image watermarking theories and techniques: A Review, Journal of Applied Research and Technology, vol. 12, no. 1, pp. 122138, Feb. 2014.

X. Y. Ye, X. T. Chen, M. Deng, S. Y. Hui, and Y. L. Wang, A multiplelevel DCT based robust DWTSVD watermark method, in: Proceedings of the 10th International Conference on Computational Intelligence and Security, Kunming, China, Nov. 15 16, 2014, pp. 479483.

S. M. Mousavi, A. Naghsh, and S. A. R. AbuBakar, Watermarking techniques used in medical images: A survey, Jounal of Digital Imaging, vol. 27, no. 6, pp. 714729, Dec. 2014.

R. Z. Liu and T. N. Tan, An SVDbased watermarking scheme for protecting rightful ownership, IEEE Transactions on Multimedia, vol. 4, no. 1, pp. 121128, Mar. 2002.

S. Kamble, V. Maheshkar, S. Agarwal, and V. K. Srivastava, DWTSVD based robust image watermarking using Arnold map, International Journal of Information Technology and Knowledge Management, vol. 5, no. 1, pp. 101105, Jan.Jun. 2012.

S. Fazli and M. Moeini, A robust image watermarking method based on DWT, DCT, and SVD using a new technique for correction of main geometric attacks, Optik, vol. 127, no. 2, pp. 964972, Jan. 2016.

Lai CC, Tsai CC. Digital image watermarking using discrete wavelet transform and singular value decomposition. IEEE Trans Instrum Meas 2010;59(11):30603.

Lagzian S, Soryani M, Fathy M. A new robust watermarking scheme based on RDWTSVD. IJIIP Int J Intell Inform Process 2011;2(1):229.

Rastegar S, Namazi F, Yaghmaie K, Aliabadian A. Hybrid watermarking algorithm based on singular value decomposition and radon transform. Int J Electron Commun (AEU) 2011;65(7):65863.

Bhatnagar G. A new facet in robust digital watermarking framework. International Journal of Electronics and Communications (AEU) 2012;66(4):27585.

Ganic E, Eskicioglu AM. Robust embedding of visual watermarks using discrete wavelet transform and singular value decomposition. J Electron Imaging 2005; 14(4):0430049.

Lai CC. An improved SVDbased watermarking scheme using human visual characteristics. Optical Communications 2011;284(4):93844.

Ling HC, Phan RCW, Heng SH. On the security of a hybrid watermarking algorithm based on singular value decomposition and radon transform. Int J Electron Commun (AEU) 2011;65(11):958 60.

Zhang XP, Li K. Comments on an SVDbased watermarking scheme for protecting rightful ownership. IEEE Trans Multimedia 2005;7(3): 5934.

Khaled Loukhaoukha, JeanYves Chouinard, M. Haj Taieb, Optimal image watermarking algorithm based on LWTSVD via multi objective ant colony optimization, J.Inf. Hiding Multimed. Signal Process. 2 (2011) 303319.

A. Gupta, M. Raval, Arobust and secure watermarking scheme based on singular values replacement, Sadhana 37 (2012) 425440.

K. Loukhaoukha, J.Y. Chouinard, Hybrid watermarking algorithm based on SVD and lifting wavelet transform for ownership verification, in: Canadian Workshop on Information Theory, 2009, pp.177182.

Nasrin M.Makbol, A new robust and secure digital image watermarking scheme based on the integer wavelet transform and singular value decomposition, Digital Signal Processing 33 (2014) 134147.

X.P. Zhang, K. Li, Comments on An SVDbased watermarking scheme for protecting rightful ownership, IEEE Trans. Multimed. 7(3) (2005) 593594.

H.C. Ling, R.C.W. Phan, S.H. Heng, On the security of a hybrid watermarking algorithm based on singular value decomposition and
radon transform, AEÃœ, Int. J. Electron. Commun. 65(11) (2011) 958960.

R. Rykaczewski, Comments on An SVDbased watermarking scheme for protecting rightful ownership, IEEE Trans. Multimed. 9(2) (2007) 421423.