Journal of Information Assurance and Security 4 (2009) 470-473
A Watermarking Technique based on Visual Cryptography B Surekha 1, Dr GN Swamy 2, Dr K Srinivasa Rao 3, A Ravi Kumar4 1, 3, 4
Department of ECE, TRR College of Engineering, Patancheru, Hyderabad, Andhra Pradesh, India - 502 319.
[email protected] 2
Department of ECE, Gudlavalleru Engineering College, Gudlavalleru, Krishna District, Andhra Pradesh, India - 521 356
[email protected]
Abstract: With the explosive growth of internet technology robust methods are being developed to protect the proprietary rights of the multimedia. This paper considers the problem of embedding a binary watermark in a gray-level image using the concept of Visual Cryptography (VC). This paper proposes a watermarking scheme which offers better security than Hwang’s method, so that, attackers will not be able to detect ownership information. The proposed scheme embeds the secret image without modifying the host image. In addition, the hidden secret image can be extracted without using the original host image and allows multiple watermarks to be embedded in the same image. The experimental results show that the proposed scheme is robust to withstand several image processing attacks such as JPEG compression, noise adding, sharpening and blurring. Keywords: Copyright Protection, Cryptography, Digital Watermarking, Secret Sharing, Visual Cryptography, Multiple watermarks.
1. Introduction Digital representation of signals brings many advantages such as lossless recording and copying, convenient distribution over networks, easy editing and modification. Unfortunately, these advantages bring in serious problems such as widespread copyright violation and illegal copying, and distribution. Cryptographic techniques do not completely solve the problem of unauthorized copying. Digital watermarking is a technology being developed to provide protection from illegal copying [1]. It involves embedding information into a digital image such that, if the image is copied, then the information is also carried into the copy. Typically, the information is a company logo or a copyright text which identifies the owner of the image. Thus, watermarking can be used to identify the content owner and also guarantees that the embedded mark can be reliably detected after the image has been modified (but not destroyed beyond recognition). Invisible watermarks, when added to the image can’t be perceived as such, and have wider applications than visible watermarks [2]. In general an effective watermarking scheme should satisfy properties such as invisibility, robustness, security, capacity and low computational complexity [3]. Visual Cryptography (VC) is basically a secret sharing
Received June 10, 2009
scheme extended for the images. It has the ability to restore a secret without the use of computations [4]. Visual Cryptography, when used in copyright protection enables resolution of rightful ownership, without using the original image. Also, the host image will not be altered during the embedding process. Various watermarking schemes are available based on Visual Cryptography [5]–[9]. Most of them [5], [7] convert the gray-level host images into two-tone images, using halftone techniques before embedding a binary watermark. Hwang proposed a scheme [6] which hides binary watermarks in gray-level images directly. Hwang’s watermark is based on simple (2, 2) VC scheme. In his method a secret key is used as a seed to select random pixels within the host image. The most significant bits of the selected pixels are used to form the first share, known as Verification Share. Then the binary watermark and the first share are combined to generate the second share, known as Master Share using the principles of (2, 2) VC. The Master Share must be registered with a trusted third party. While resolving the rightful ownership, the same secret key is to be used as a seed, to select the most significant bits of the modified image. The resulted Verification Share and the Master Share are combined to recover the hidden watermark. The Hwang’s scheme doesn’t provide security always. For example, if 90% of the pixels in the host image have graylevels greater than 128, then 90% of the most significant bits will have the value as logic1, regardless of the secret key. The Verification Share is revealed without the knowledge of the secret key, indirectly. This drawback of Hwang’s scheme makes it unsuitable for digital image copyright protection. This paper proposes a watermarking scheme that overcomes the drawbacks of Hwang’s scheme and offers better security. The rest of the paper is organized as follows. Section 2 briefly reviews the basic (2, 2) VC scheme. Section 3 describes the proposed algorithm. Experimental results are illustrated in Section 4. Section 5 concludes the paper.
1554-1010 $ 03.50 Dynamic Publishers, Inc.
471
Surekha et al. 3. Proposed Scheme
2. A (2, 2) Visual Cryptography Scheme Visual Cryptography (VC) was first introduced by Noar and Shamir at Eurocrypt’94 [4]. To encode a secret employing a (2, 2) VC Scheme, the original image is divided into two shares such that each pixel in the original image is replaced with a non-overlapping block of two sub-pixels. Anyone who holds only one share will not be able to reveal any information about the secret. To decode the image, each of these shares is xeroxed onto a transparency. Stacking both these transparencies will permit visual recovery of the secret. Table 1 illustrates the scheme of encoding one pixel in a (2, 2) VC scheme. A white pixel is shared into two identical blocks of sub-pixels. A black pixel is shared into two complementary blocks of sub-pixels. While creating the shares, if the given pixel p in the original image is white, then the encoder randomly chooses one of the first two columns of Table 1. If the given pixel p is black, then the encoder randomly chooses one of the last two columns of Table 1. Each block has half white and half black sub-pixels, independent of whether the corresponding pixel in the secret image is black or white. All the pixels in the original image are encrypted similarly using independent random selection of columns. Thus no information is gained by looking at any group of pixels on a share, either. Table 1. A (2, 2) Visual Cryptography Scheme
Figure 1 shows the results of basic (2, 2) VC Scheme. When the two shares are stacked together, as in Figure 1(d), the black pixels in the original image remain black and the white pixels become gray. Although some contrast loss occurs, the decoded image can be clearly identified. Since each pixel in the original image is replaced by two sub-pixels in each share, the width of the decoded image is twice that of the original image.
This section describes a scheme for embedding binary watermarks in gray-level images, to overcome the security drawbacks of Hwang’s scheme. A watermarking system is usually divided into three distinct phases: embedding, attack and detection. In the embedding phase, a binary matrix is created from the host image and a secret key. This binary matrix, along with the secret binary image (watermark), generates a Master Share according to the predefined encryption rules of the VC Scheme. The Master Share has to be registered with a trusted third party for further verification. The watermarked image is usually transmitted or stored. If a person makes a modification to the marked image, it is called an attack. While resolving the rightful ownership, the same secret key is used to generate the Verification Share. Then the Verification Share and the Master Share (kept by the trusted third party) are combined to recover the hidden watermark. 3.1 Embedding Procedure This scheme assumes that the binary secret image (watermark) S of size wxh is to be embedded into the host image H of size rxc. Let K be a random integer selected by the owner as a secret key. The output of the embedding phase is a watermarked image O of size rxc (same as the original host image) and a Master Share M of size wx2h. Inputs: A Host Image H, a Binary Watermark S, and a Secret Key K Outputs: Marked Image O and a Master Share M The watermark embedding procedure is as follows: Step1: The secret key K is used as a seed to generate wxh random numbers over the interval [1 to rxc]. Let Ri be the ith random number. Step2: Creation of a binary matrix X of size wxh such that the entries in the array are the most significant bits of Rith pixel of the host image. Step3: Creation of a binary matrix Z of size wxh such that the entries in the array are the most significant bits of the Rith random number. Step4: Creation of a binary matrix Y of size wxh such that Yi= XOR (Xi , Zi) Step5: Creation of a Master Share M by assigning a pair of bits for each element in the binary matrix Y according to the predefined encryption rules of VC as shown in Table 2. Finally, the Master Share is registered with a trusted third party. Table 2. Encryption Rules to Create Master Share
Figure 1. Example of (2, 2) Visual Cryptography Scheme
A watermarking Technique based on Visual Cryptography 3.2 Detection Procedure Detection (also called extraction) is an algorithm which is applied to the attacked image to extract the watermark from it. In robust (secure) watermarking applications, the extraction algorithm should be able to reproduce the watermark, even if the modifications were strong. Inputs: Modified image O’, Master Share M, and a Secret Key K Output: Extracted Watermark S’ The watermark detection procedure is as follows: Step1: The secret key K is used as a seed to generate wxh random numbers over the interval [1 to rxc]. Let Ri be the ith random number. Step2: Creation of a binary matrix X of size wxh such that the entries in the array are the most significant bits of Rith pixel of the host image. Step3: Creation of a binary matrix Z of size wxh such that the entries in the array are the most significant bits of the Rith random number. Step4: Creation of a binary matrix Y of size wxh such that Yi= XOR (Xi ,Zi) Step5: Creation of a Verification Share such that, if the element in the binary matrix Y is ‘0’ then Vi = (0, 1) has to be assigned, else Vi = (1, 0) has to be assigned. Step6: The secret image can be extracted by performing logical OR operation as follows: Si’ = OR (Mi , Vi). 3.3 Security Analysis The security in the proposed method is due to the XOR operation performed in Step4 during the embedding phase. In this scheme, the entries of the binary matrix are obtained by taking XOR of the most significant bits of the selected pixels, and the most significant bits of the corresponding random numbers. This indicates that the same large binary matrix resulted from the secret key is to be used in the subsequent extraction process. Otherwise there is no way an attacker can estimate the Verification Share irrespective of the contrast of the host image. Thus this method offers better security than Hwang’s method.
4. Experimental Results The proposed method is tested on a host image of size 512 x 512 gray-level bitmap Lena image [10] which is shown in Figure 2(a). The binary watermark to be embedded is of size 125x125 pixels as shown in Figure 2(b). The Master Share generated in the watermark embedding phase, and the Verification Share generated in the detection phase looks like random images and is similar to the images shown in Figure 1(c).
Figure 2. Original Host Image and Extracted Watermark
472
When there is no attack, the Verification Share generated in the detection phase is as shown in Figure 1(b). Figure 2(c) shows the result of stacking these two shares. In practice, there is a very good chance for the watermarked image to be altered (intentionally and unintentionally) while being transmitted through the channel. These alterations may be a result of intentional attacks such as filtering, blurring, etc. or unintentional distortions such as JPEG compression, channel noise addition etc. To test the robustness of the proposed algorithm, the watermarked images were subjected to various image manipulating operations and compression attacks. All attacks are implemented using the Matlab Image Processing Tool box. Finally, the performance of the algorithm with respect to attack resilience has been established by the results shown in Table 3. The second column of Table 3 shows the modified Lena image after different attacks. The third column shows the extracted binary watermarks. Experimental results reveal that the watermark pattern is well recognizable, even after making severe modifications to the marked image. As long as the extracted watermark is recognizable, the purpose is served. Table 3. Resulting Watermarks from Different Attacks
473
5. Conclusions This paper extends Hwang’s work on watermarking and proposes a new method that can provide better security. The proposed method embeds a binary watermark in a gray-level image using the concept of Visual Cryptography (VC). Experimental results reveal that the proposed watermarking method is robust to most common attacks such as blurring, sharpening, lossy compression and noise adding. The advantage of this embedding scheme is that the host image will not be altered. Also the secret image can be of any size. This scheme allows multiple watermarks to be hidden in a single host image. The security of this scheme is ensured by the properties of Visual Cryptography and XOR operation. The advantage of the proposed extraction scheme is that resolving the ownership can be done without the aid of the original image. However, the proposed scheme is not resistant to some attacks like rotations, cropping, scaling, contrast adjustments and translations. Further work would improve on these lines.
Acknowledgment The authors express their gratitude to Dr MC Chandra Mouly, Professor of ECE & Dean of Engineering, TRR College of Engineering, Hyderabad, for his valuable suggestions in the organization and development of this paper. The authors also acknowledge the management of TRR institutions, Hyderabad, for their constant encouragement in completing this work.
References [1] R. J. Anderson, Ed., “Information Hiding”, In 1st International Workshop, Lecture Notes in Computer Science, Springer-Verlag, Berlin, Vol. 1174, pp.1-7, 1996. [2] G. W. Braudaway, K. A. Magerlein, and F. Mintzer, “Protecting Publicly-available Images with a Visible Image Watermark”, In Proceedings of SPIE, Vol. 2659, pp.126–133, 1996. [3] Cox, I. J., Kilian, J., Leighton, T., and Shamoon, T., “Secure Spread Spectrum Watermarking for Multimedia”, In IEEE Transactions on Image Processing, Vol. 6, No. 12, pp. 1673−1687, 1997. [4] M. Noar and A. Shamir, “Visual Cryptography”, Advances in Cryptography Eurocrypt’94, Lecture Notes in Computer Science, Springer-Verlag, Berlin, Vol. 950, pp. 1-12, 1995. [5] Y-C Hou, P-M Chen, “An Asymmetric Watermarking Scheme based on Visual Cryptography”, In Proceedings of ICSP, Vol. 2, pp. 992 -995, 2000. [6] R. Hwang, “A Digital Image Copyright Protection Scheme based on Visual Cryptography”, Tamkang Journal of science and Engineering, Vol.3, No.2, pp. 97-106, 2002. [7] M. S. Fu and 0. C. Au, “Joint Visual Cryptography and Watermarking”, In Proceedings of IEEE International Conference on Multimedia and Expo, pp. 975-978, 2004.
Surekha et al. [8] Chang, C.C., Chaung ,J. C., “An Image Intellectual Property Protection Scheme for Gray-level Images using Visual Secret Sharing Strategy”, Pattern Recognition Letters, Vol. 23, pp. 931−941, 2002. [9] Chang, C. C., Hsiao, J. Y., and Yeh, J. C., “A Color Image Copyright Protection Scheme based on Visual Cryptography and Discrete Cosine Transform”, The Imaging Science Journal, Vol. 50, pp. 133−140, 2002. [10] www.ece.rice.edu/~wakin/images
Author Biographies
Ms B Surekha is currently working as an Associate Professor in the Dept of ECE, TRR College of Engineering, Hyderabad, AP, India. She has received the B. Tech degree from the Nagarjuna University, India and the M. Tech degree from the JNT University, India, both in Electronics and Communication Engineering. She is currently pursuing PhD degree at JNT University, India. She has several publications in international conferences. Her research interests include Cryptography and Copyright Protection.
Dr GN Swamy is currently working as a Professor and Head of the Department, Dept of ECE, Gudlavalleru Engineering College, Gudlavalleru, India. He has received his Ph D degree in Signal Processing from the Andhra University, India He has 16 years of teaching experience and is actively associated with national professional bodies like IETE and ISTE, India. He has several publications to his credit at national and international level. His research interests include Electronic Devices, Microwaves, Signal Processing and Cryptography.
Dr K Srinivasa Rao is currently working as Professor of ECE and Principal, TRR College of Engineering, Hyderabad, AP, India. He received his Ph D degree in Antennas from the Andhra University, India. He has 25 years of experience in the field of Education. He is an active member of IEEE, ISTE, SEMCEI, BMESI. He is a Fellow of IE (I) and IETE. He has several publications in various journals and conferences of national and international repute. His research interests include Communications, Antennas and Cryptography.
Prof. A.Ravi Kumar is currently working as a Professor and Head of the Department, Dept of ECE, TRR College of Engineering, Hyderabad, AP, India. He has received the B. Tech degree from the Nagarjuna University, India and the M. Tech degree from the JNT University, India, both in Electronics and Communication Engineering. He has 25 years of teaching and industrial experience, and is associated with IETE and IE (I).He has several publications to his credit. He is presently the Co-opted Member, Academic & Distance Education Committee, IETE Head Quarters, New Delhi, India, and Special Invitee, IETE, Hyderabad centre,India.
