Robust watermarking scheme for tamper detection and authentication exploiting CA
2019; Institution of Engineering and Technology; Volume: 13; Issue: 12 Linguagem: Inglês
10.1049/iet-ipr.2018.6638
ISSN1751-9667
AutoresPabitra Pal, Biswapati Jana, Jaydeb Bhaumik,
Tópico(s)Digital Media Forensic Detection
ResumoIET Image ProcessingVolume 13, Issue 12 p. 2116-2129 Research ArticleFree Access Robust watermarking scheme for tamper detection and authentication exploiting CA Pabitra Pal, Pabitra Pal orcid.org/0000-0002-2866-7320 Department of Computer Science, Vidyasagar University, West Midnapore, 721 102 IndiaSearch for more papers by this authorBiswapati Jana, Corresponding Author Biswapati Jana biswapatijana@gmail.com orcid.org/0000-0003-4476-3459 Department of Computer Science, Vidyasagar University, West Midnapore, 721 102 IndiaSearch for more papers by this authorJaydeb Bhaumik, Jaydeb Bhaumik Department of ETCE, Jadavpur University, Kolkata, 700 032 IndiaSearch for more papers by this author Pabitra Pal, Pabitra Pal orcid.org/0000-0002-2866-7320 Department of Computer Science, Vidyasagar University, West Midnapore, 721 102 IndiaSearch for more papers by this authorBiswapati Jana, Corresponding Author Biswapati Jana biswapatijana@gmail.com orcid.org/0000-0003-4476-3459 Department of Computer Science, Vidyasagar University, West Midnapore, 721 102 IndiaSearch for more papers by this authorJaydeb Bhaumik, Jaydeb Bhaumik Department of ETCE, Jadavpur University, Kolkata, 700 032 IndiaSearch for more papers by this author First published: 01 August 2019 https://doi.org/10.1049/iet-ipr.2018.6638Citations: 6AboutSectionsPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Share a linkShare onFacebookTwitterLinkedInRedditWechat Abstract In this study, the authors have employed a special type of periodic boundary CA (called CA attractor) for image authentication and tamper detection through the watermarking scheme. Authentication code (AC) has been generated utilising secure hash algorithm-512 on watermark image. The cover image (CI) is subdivided into four sub-sampled interpolated image, where AC and secret watermark bits are embedded to enhance capacity, quality and security. At the receiver end, the secret watermark information and CI are successfully extracted without any distortion from four sub-sampled watermarked images (WIs). Additionally, the proposed scheme can successfully determine any type of distortion within the WI that may be possible to occur by various steganographic attacks. Indeed, better results in terms of capacity and quality are experienced after having compared with similar schemes in vogue. The intended outcome brought into the limelight some remarkable sublime characteristics in the field of image authentication and ownership identification without which the technology life is stunted. Innumerable government and private sector facet including health care, commercial security, defence and intellectual property rights are immensely benefited from this scheme. 1 Introduction In this information age, modern society is changing by virtue of the significant evolution of Internet technology and development of digital communication systems. Nowadays, multimedia documents such as images, audio and video are easy to store, manipulate and disseminate with the help of digital communication through a public communication channel. Therefore, it is hard to preserve the integrity of valuable digital information and raise the problem related to authentication ownership identification and copyright protection. Illegal alteration is becoming an issue in many human-centric applications such as medical images [1], e-governance, military communication, where authentication [2], copyright protection, copy–move forgery [3] and tamper detection [4] are essential. One can take the advantage of digital watermarking scheme to deal with such problems. Digital image watermarking is the technique of implanting an unapparent or apparent watermark (logo) within a digital image [cover image (CI)]. For a long time, watermarking schemes have been used in currency, postage stamps and other government evidence. High capacity maintaining good visual quality along with robustness is still a major research issue in unapparent watermarking schemes. In this work, we have used an unapparent digital watermark that is essential in the era of modern digitisation and online social communication. Several watermarking schemes [5-7] were designed to tackle the issue that arises due to the unauthorised access of digital data. These existing watermarking schemes have some pros and cons depending on requirements. Table 1 shows their major achievements and drawbacks. Watermarking is applied to preserve copyrighted documents from being corrupted and allow us to get the information of permissible ownership. The important features of watermarking schemes are imperceptibility, payload, security, tamper detection and authentication with less execution time. Security and robustness [4] are two important requirements for designing any watermarking scheme. The fundamental challenges [10] in watermarking are to resist attacks based on modification, compression, scaling, filtering, cropping, copy and paste, collage etc. So efficient watermarking scheme [11] should make a good trade-off between security, robustness and imperceptibility. Although high embedding capacity is not the main purpose of the watermarking schemes but many of the current researchers [1, 12, 13] are also taken this as one of the important characteristics while possessing the quality of the watermarked image (WI). Table 1. Comparative study on various existing watermarking schemes Author Scheme Watermark Reversible/robust Types Tamper detection/localisation/authentication Su et al. [5] Arnold transform binary yes/yes blind yes/no/no Pal et al. [6] weight matrix colour yes/yes fragile yes/yes/yes Parah et al. [8] interblock pixel difference grey yes/yes blind yes/no/yes Wenyin and Shih [9] LBP binary no/yes semi-fragile yes/yes/yes proposed scheme CA code colour yes/yes fragile yes/yes/yes 2 Literature review Mukherjee et al. [2] proposed a CA-based authentication scheme for digital image and text message, which are better than the scheme based on MD5, secure hash algorithm (SHA)-1 in terms of security and its execution speed. Yoo et al. [14] introduced a novel secured watermarking technique for authentication of printed images, which can easily determine alteration of a printed image since the first printing time. Ye and Li [15] developed a chaotic CA-based novel watermarking scheme to scramble image as a preprocessing for watermarking scheme. Elementary cellular automata (ECA)-based secret sharing scheme has been suggested by Ahmed and Younes [16], where the state transition behaviour of CA with periodic boundary condition has been used. After that, Bhardwaj and Sharma [17] proposed a dual-layer image encryption schemes using two-dimensional CA with non-periodicity behaviour of CA to make the scheme more robust. A reversible image sharing technique has been advised by Shin and Jung [18] to furnish copyright protection in cover work. They achieved 44 dB peak-signal-to-noise ratio (PSNR) with average embedding capacity 524,288 bits. Tralic et al. [3] formulated a new approach to detect copy–move forgery, in which CAs were applied. After that, Tralic et al. [19] designed a novel technique to detect duplicate regions using CA and local binary pattern (LBP). Simplicity and low-computational complexity is the main advantage of their scheme. However, till date, less attention has been paid to accomplish three major research issues, namely tamper detection, authentication and image recovery in a single model when watermarking schemes are designed. Also, most of these research efforts were concentrated on grey-scale images and some of these schemes are susceptible to noise and attacks. The improvement of the visual quality within the recovered colour image has not been investigated, and many of the watermarking schemes concentrated only on the efficiency of tamper detection. So, the development of a watermarking scheme that can detect tamper location and verify authentication is an important research issue, which is essential in health care, military, private and government sectors. A few researchers [2, 3, 15, 16] have developed some watermarking scheme for tamper detection and authentication through CA. CA-based reversible watermarking scheme having high embedding capacity, better security with high perceptibility is an important area of investigation. So, it is a great challenge for the researcher to make a good trade-off among the robustness, visual quality, payload and security. In light of this discussion, we have introduced a colour image-based watermarking method using ECA for tamper detection and image authentication. The scheme has been designed in such a way that the watermark bits are not embedded within the cover media directly to keep the unvarying security level. The watermark bits are encrypted using ECA attractor before embedding within the cover work. This provides secrecy of the watermark (logo) against public extraction and detection. In embedding phase, the encrypted watermark bits are generated for four sub-sample images by employing CA attractor in coloured CI and authentication bits are computed by employing cryptographic hash function SHA-512 on secret data. We then embedded the information (authentication bits) within the sub-sample of colour images. In watermark extraction phase, it is possible to retrieve secret data and authentication code (called Extracted Authentication Code (EAC)) from sub-sample images. Finally, in authentication phase, the AC code and reconstructed (called Reconstructed Authentication Code (RAC)) from recover watermark image using SHA-512, and compared with EAC for verifying the authentication of CI. The tamper location can also be detected. In this investigation, our aim is to formulate a reversible watermarking algorithm using CA for practical applications. The proposed scheme provides good visual characteristics of WI. 2.1 Motivations and objectives The motivation and objective of this scheme are as follows: Reversibility: Watermarking [6] is one of the best solutions for image authentication and tamper detection. However, after embedding, the watermark can damage the important information that is present in the original CI. So, it is difficult to get back an original CI at the receiver end. Recently, recovery of original CI is essential in many application areas such as military, medical etc. In such applications, reversible watermarking scheme is essential instead of conventional (apparent) watermarking scheme. Robustness: Robustness of a scheme is measured by evaluating the strength of the scheme against existing steganographic attacks. From the existing literature, it has been observed that there are few watermarking techniques that are developed which are robust against attacks in various practical applications. Tamper detection: Tampering is an intentional modification of multimedia documents in such a way that would make them harmful to end users. So, it is essential to reveal the watermark as well as the original CI from tampered WI during the extraction process. Imperceptibility: Imperceptibility is the basic requirement for any invisible watermarking approach. In other words, maintaining a good visual quality after embedding the watermark is a challenging but necessary requirement. Authentication: Authentication is the process of recognising the owner of the object and its identity. It is determining claimed user identity by checking user provided evidence that was embedded previously in the WI. The evidence, which is provided by the user during the process of authentication, is called a credential. 2.2 Contribution of this paper The major contributions on the proposed watermarking scheme are: Sub-sample images with interpolation are used to increase data hiding capacity, security and achieve reversibility. It is hard to extract secret message without simultaneous sub-sampled images, which are a special case of secret sharing. AC is generated using SHA-512 from watermark image and embed within sub-sample images. Shared secret key has been exclusive ORed (XORed) with watermark bits and generate encrypted message, which is used in CA for distribution in sub-sampled image to increase security. It is hard to retrieve secret message without knowing shared secret key and proper CA rule. It provides two-fold security. Tamper detection and authentication of both cover and secret image are achieved by employing CA attractor with shared secret key. We have designed an innovative scheme using colour image, which is less susceptible to noise and attacks due to minimum modification in pixel level. The computational cost of the suggested scheme is less than the previous scheme, which may be used for real-time application. Rest of this paper is organised as follows: the proposed watermarking scheme is discussed in Section 3. Experimental results and comparison with existing related works are provided in Section 4. Finally, conclusions are drawn in Section 5. 3 Proposed watermarking scheme 3.1 Watermark embedding phase The block diagram of the embedding, extraction and authentication phases of the proposed scheme has been depicted in Fig. 1a–c, respectively. The watermark embedding phase mainly consists of two sections as: (i) key generation phase and (ii) watermark embedding procedure. The first part, i.e. the key generation phase is the main phase for any security-related scheme. Fig. 1Open in figure viewerPowerPoint Block diagram of watermarking process (a) Watermark embedding process, (b) Watermark extraction process, (c) Authentication process 3.1.1 Key generation A shared secret key of length 128 bit is considered for watermark embedding. Now, a watermark bit stream is taken from the watermark image (W) and 512 bit AC is generated using SHA-512 algorithm (shown in Fig. 1a). After that, a sequence of numbers between 0 and 8 have been generated using pseudorandom number generator with as the seed value and are stored in the array NS[]. Four secret vectors {, , , } are chosen sequentially from NS[], such that the sum of all four secret vectors is (i.e. ). Hence, any -block CA attractor ( for i = 1 to ) has to be considered for watermark embedding (Fig. 1a). Here, CA rule-42 is applied to choose attractor block with initial state 53. For example, has been collected from NS[] as four secret vectors ( for ) and are considered as -block CA attractor ( for to 8). For the next block, is updated using (1) to get a new sequence of four secret vectors ( for ) from updated NS[] (1)where is the average pixel value of the corresponding block of CI. 3.1.2 Embedding procedure In embedding phase, a colour CI is separated into three red, green and blue (RGB) planes , and . After that, first colour block is considered and divided into image blocks. Then, four Sub-sample Image (SI) are produced from this image blocks using (2). A sample generation and interpolation method is depicted in Fig. 2a. Four interpolated image blocks Interpolated Sub-sample Image (ISI) are constructed from each sub-sample using (3) (2) (3)where and m and n are considered as row and column of the corresponding image block. The interpolated image ISI is generated from each sub-sampled image SI, which corresponds to CI individually presented in (3). Here, a new interpolation scheme is proposed with an Additional Displacement (AD) variable that provides better quality interpolated image. Now, the watermark bits are embedded into each ISI. First -bit data is taken from the watermark bits (M). Then, an XOR () operation is performed between and to get the value . After that, are transformed into four parts (for to ) to embed in four different interpolated image samples (shown in Fig. 1a). Also, the randomness of the steps can be achieved by choosing the number of embedded in the interpolated sample images. The and are calculated using (4) and (5), respectively (4) (5) Fig. 2Open in figure viewerPowerPoint Numerical illustration of watermark embedding phase (a) Sub-sample generation and interpolation, (b) Embedding procedure Now, (for to 4) are converted to 8-bit binary number. Each two bits from this number are embedded into two bits starting from Least Significant Bit (LSB) of four different pixels of the ISI shown in Fig. 1. In this way, all the bits of are embedded into the first block of the corresponding shares ISI. The two bits AC are embedded into the middle pixel () of the first block of each share. Before going to the next block, the secret key is updated using (1). Similarly, other watermark bits and authentication bits are embedded within the rest of pixel blocks of ISI. Finally, interpolated sub-sampled WIs () are generated by combining the RGB colour components. A numerical illustration is shown in Fig. 2. The corresponding algorithm is shown in Algorithm 1 (Fig. 3). The embedding algorithm of the proposed scheme contains two procedures: Embedding() and EmbedBitsInImageBlock(). The detailed procedures are described as follows: Embedding(): Here, EmbedBitsInImageBlock() function is called for four interpolated image block separately to embed secret message through the XOR operation. Then, three updated colour components are combined by CreateSampleWatermarkedImage() function to form the sub-sampled colour WI (). Fig. 3Open in figure viewerPowerPoint Algorithm 1: watermark embedding algorithm EmbedBitsInImageBlock(): In this function, the encrypted watermark bits are embedded within the specified places. 3.2 Watermark extraction phase The watermark extraction phase has been clearly described using a numerical illustration depicted in Fig. 4 and an algorithmic illustration shown in Algorithm 2 (Fig. 5). Fig. 4Open in figure viewerPowerPoint Numerical illustration of watermark extraction phase of the proposed scheme Fig. 5Open in figure viewerPowerPoint Algorithm 2: watermark extraction algorithm 3.2.1 Extraction procedure A pixel block of original CI is constructed from the unaffected pixels and from each pixel blocks of all the shares (). In this way, the whole CI is constructed from four shares. Now, two bits starting from LSB are collected from the pixels and of the first block of , to form an -bit binary number. This number is converted to its decimal equivalent and stored into the corresponding . Again, the LSB of is collected from each share and appended to form a string (EAC) shown in Fig. 2. Now, a shared secret key is considered and XOR operation is performed with to get . This is converted to -bit binary string and appended with . Then, the average of px is calculated from the first block of the reconstructed CI. After that, this is used to modify using (1). Then, the above process is continued for the rest of the pixel blocks of all the shares and recovered from each iteration is appended with . So, contains all the extracted watermarked bits in binary format. The is then reconstructed from shown in Fig. 4. The AC (RAC) is regenerated from this watermark using SHA-512 algorithm. Finally, RAC is compared with EAC to check the authenticity of recovered CI. The watermark extraction phase has been clearly described using a numerical example illustrated in Fig. 4 and algorithmic presentation is shown in Algorithm 2 (Fig. 5). The extraction algorithm of the proposed scheme contains two procedures: Extraction() and EmbedBitsFromImageBlock(). The detailed procedures are described as follows. 4 Experimental results and comparison A set of benchmark [20-23] colour images of size are considered to assess the effectiveness of the proposed scheme. Here, three different sizes of logo images are considered as a watermark as shown in Fig. 6 to measure the quality and corresponding capacity. Performances of the related schemes are compared with the proposed schemes. Mean square error (MSE) [6], PSNR [6], structural similarity index measurement (SSIM) [5] and quality index (Q-index) are computed to test the perceptible characteristics after embedding. Furthermore, normalised correlation coefficient (NCC) [9], bit error rate (BER) [8], standard deviation () [24] and correlation coefficient () [24] are computed for tamper detection in a watermark image. Performance of the proposed scheme is assessed on the basis of computation time and it has been compared with other existing schemes. Fig. 6Open in figure viewerPowerPoint Watermark images (logo) with different sizes are tested in the proposed scheme 4.1 Quality measurement and payload analyses The fundamental necessities of any watermarking scheme are robustness and imperceptibility. Usually, the qualities of WIs are evaluated from their subjective and objective quality indices. The subjective characteristics of the WIs are evaluated in the proposed schemes and these are depicted in Fig. 7. The evaluation results of the proposed scheme in terms PSNR and Q-index after embedding different number of bits of watermark are presented in Table 2. It is observed from Fig. 7 that no visual distortions are detected after embedding maximum payload of 393,216 bits watermark. The Q-index values are close to unity, which establishes the acceptability of the proposed scheme. From Table 2, it is also found that the PSNR value is 53.03 dB after hiding 98,304 bits watermark into the CI and this value has been changed to 50.21 dB after hiding 393,216 bits watermark. So, it has been noted that after hiding watermark, the visual quality has been degraded only 5%. Thus, without degrading the quality of WI, a huge amount of data can be embedded. This higher visual quality has been achieved due to the insertion of watermark bits in LSB position of each pixel. Table 2. Capacity, PSNR, Q-index and bpp values for standard benchmark images Dataset Image Capacity, bits PSNR Q-index bpp USC-SIPI [22] Lena 98,304 53.03 0.99999 0.375 196,608 51.35 0.99998 0.75 393,216 50.21 0.99997 1.5 UCID [21] Jerusalem 98,304 53.17 0.99999 0.375 196,608 52.67 0.99997 0.75 393,216 50.91 0.99996 1.5 STARE [20] Im0001 98,304 53.15 0.99998 0.375 196,608 50.93 0.99996 0.75 393,216 49.62 0.99995 1.5 High Dynamic Range (HDR) [23] Medical1 98,304 53.59 0.99998 0.375 196,608 51.35 0.99997 0.75 393,216 49.27 0.99995 1.5 Fig. 7Open in figure viewerPowerPoint Standard colour images University of Southern California - Signal and Image ProcessingInstitute (USC-SIPI) [22], Structured Analysis of the Retina (STARE) [20], HRD [23] and Uncompressed Color Image Database (UCID) [21] of size pixels are used as CI The proposed scheme has been tested taking more than 100 sample images from four different standard benchmark image databases and experimental outcomes are exhibited in Table 3. The experimental results are categorised into three different cases. At first, we consider 25, 50 and 100 images and the average PSNR (dB) is calculated. It has been noted that, ∼50 dB average PSNR can be achieved after embedding a maximum amount (393,216 bits) watermark. Table 3. Average PSNR (dB) of various yardstick image datasets considering 25–100 images Database Size of image Number of image Average PSNR USC-SIPI [22] 25 50.09 50 49.65 100 50.21 UCID [21] 25 49.93 50 50.34 100 49.91 STARE [20] 25 50.73 50 49.83 100 49.62 HDR [23] 25 50.79 50 49.94 100 49.27 Fig. 8 depicts the variation of PSNR for different schemes considering 1338 images from UCID image database [21] with different levels of embedding capacities. Fig. 8 represents that after embedding 98,304, 196,608 and 393,216 bits watermark, we achieved ∼53, 51 and 50 dB average PSNRs, respectively. Fig. 8Open in figure viewerPowerPoint Graphical representation of PSNR on UCID image database Table 4 exhibits the test results in terms of MSE, PSNR, NCC, SSIM, Q-index and BER for colour CIs of four different benchmark image databases. From Table 4, it is found that the visual qualities for the aforesaid image databases are >50 dB PSNR. The NCC, SSIM and Q-index values of the proposed scheme are close to one, and BER values closer to zero, which establish the effectiveness of the proposed algorithm. Table 4. Results of different evaluation metrics (MSE, PSNR, NCC, SSIM, Q-index and BER) for different images Dataset Images PSNR NCC SSIM, % Q-index BER SIPI [22] 1Lena 50.22 0.99998 99.46 0.9999 0.019 2Baboon 49.97 0.99998 99.57 0.9999 0.019 3Tiffany 49.57 0.99999 99.13 0.9996 0.019 4Barbara 49.91 0.99997 99.39 0.9999 0.019 average 49.96 0.99998 99.34 0.99986 0.0185 HDR [23] Bird 47.88 0.99996 99.37 0.9999 0.019 Jeruslem 49.70 0.99994 99.32 0.9998 0.019 Redrock2 50.11 0.99996 99.43 0.9999 0.019 Safari04 50.03 0.99997 99.26 0.9999 0.019 average 50.01 0.99995 99.30 0.99985 0.0187 UCID [21] ucid00085 47.91 0.99997 99.78 0.9999 0.019 ucid00091 49.95 0.99997 99.56 0.9999 0.019 ucid00104 49.97 0.99998 99.42 0.9999 0.019 ucid00341 49.90 0.99996 99.78 0.9999 0.019 average 49.88 0.99997 99.45 0.9999 0.019 STARE [20] im0001 47.62 0.99995 98.17 0.9998 0.018 im0370 49.91 0.99997 98.68 0.9999 0.018 im0371 49.89 0.99996 98.61 0.9999 0.019 im0373 49.91 0.99997 98.60 0.9999 0.018 average 49.83 0.99996 98.46 0.99987 0.0184 The comparison to respect to PSNR (dB) and payload (bpp) for Lena, Aeroplane, Baboon, Tiffany, Boat and Pepper images are presented in Table 5 and graphically depicted in Fig. 9. From these results, it is seen that the proposed scheme provides better results in terms of capacity compared with other existing schemes. Table 5. Comparison of different RWTs in sub-sample image with respect to PSNR (dB) and embedding capacity (bpp) Image Lin and Tsai [25] Chang et al. [26] Parah et al. [8] Shin and Jung [18] Lin and Chang [27] Proposed PSNR bpp PSNR bpp PSNR bpp PSNR bpp PSNR bpp PSNR bpp Lena 39.16 1/4 40.92 (t−1)/3 40.58 3/64 45.13 (t−2)/4 46.95 1/2 50.22 3/2 Baboon 39.15 1/4 40.92 (t−1)/3 39.60 3/64 43.9 (t−2)/4 46.92 1/2 49.97 3/2 Aeroplane 39.21 1/4 40.87 (t−1)/3 41.18 3/64 45.48 (t−2)/4 46.90 1/2 49.57 3/2 Peppers 39.20 1/4 40.96 (t−1)/3 40.43 3/64 44.41 (t−2)/4 46.85 1/2 50.18 3/2 Boat 39.18 1/4 40.93 (t−1)/3 41.32 3/64 44.11 (t−2)/4 46.96 1/2 49.96 3/2 Tiffany 39.13 1/4 40.89 (t−1)/3 41.35 3/64 45.1 (t−2)/4 46.84 1/2 50.13 3/2 average 39.18 1/4 40. 92 (t−1)/3 40.74 3/64 44.68 (t−2)/4 46.91 1/2 50.01 3/2 Fig. 9Open in figure viewerPowerPoint Comparison in terms of PSNR with existing schemes based on sub-sample image for Lena, Baboon, Aeroplane, Pepper, Boat and Zelda image 4.2 Robustness analysis Robustness of the proposed scheme is analysed by evaluating the quality metrics such as NCC [9], BER [8], and [24]. Furthermore, the proposed scheme has been assessed against salt and pepper noise, cropping and copy–move forgery attacks. 4.2.1 Analysis of standard deviation () and correlation coefficient () The proposed technique is evaluated on the basis of statistical distortion using parameters such as standard deviation and correlation coefficient . The experimental results are illustrated in Table 6. From Table 6, it is observed that there is no substantial difference of between the CI and the WI. Average values of original CI and the WI are 127.8765 and 127.7015, respectively, and their difference is 0.175 for Lena image after embedding 393,216 bits watermark. The average value between the original image and the WI is 0.9999 for Lena image. So, finding the watermark from the WI becomes quite difficult. These results represent that the proposed scheme provides a good camouflage of watermark with a decrease in the probability of watermark detection and ensures the robustness of the scheme. Table 6. Standard deviation and correlation coefficients of CI and WI Database Image –CI–avg –WI–avg –avg SIPI [22] Lena 127.8765 127.7015 0.9999 Baboon 123.1512 123.1908 0.9999 Tiffany 77.2014 76.9679 0.9997 Barbara 141.2811 141.3621 0.9999 Zelda 137.5968 137.6723 0.9999 Pepper 135.3631 135.3238 0.9999 BoatsColour 161.0483 161.0385 0.9999 UCID [21] ucid00006 218.9172 218.8121 0.9999 ucid00085 173.9327 173.9272 0.9999 ucid00091 176.5012 176.6390 0.9999 ucid00104 171.1419 171.0576 0.9999 ucid00341 138.9078 138.9214 0.9999 ucid00786 143.0516 143.0617 0.9999 ucid00797 245.1584 244.6556 0.9999 HDR [23] Bird 140.2114 139.6596 0.9999 Jeruslem 118.9881 119.0321 0.9998 Redrock2 199.4528 199.6689 0.9999 Sedona01 150.8240 150.9086 0.9999 Stonehouse 185.1848 184.7796 0.9999 STARE [20] im0001 104.4130 104.6566 0.9998 im0370 188.0491 188.0544 0.9999 im0371 146.7582 146.7793 0.9999 im0373 172.8893 172.8928 0.9999 im0374 108.2031 108.2079 0.9998 4.3 Brute force attacks Robustness of the proposed scheme is analysed by evaluating the quality metrics such as PSNR, SSIM, Q-index, NCC and BER in the presence of salt and pepper noise, cropping and copy–move forgery attacks. The attacking results on the proposed scheme after applying salt and pepper noise, cropping and copy–move forgery attack with different noise levels are depicted in Figs. 10-12, respectively. The salt and pepper noise levels 0.01, 0.1 and 0.5 are applied in watermarked Lena image (shown in Fig. 10). After that, our
Referência(s)