Privacy Protection of Digital Images Using Watermarking and QR Code-based Visual Cryptography

Te increase in information sharing in terms of digital images imposes threats to privacy and personal identity. Digital images can be stolen while in transfer and any kind of alteration can be done very easily. Tus, privacy protection of digital images from attackers becomes very important. Encryption, steganography, watermarking, and visual cryptography techniques to protect digital images have been proposed from time to time. Te present paper is focused on the enhancement of privacy protection of digital images utilizing watermarking and a QR code-based expansion-free and meaningful visual cryptography approach which generates visually appealing QR codes for transmitting meaningful shares. Te original secret image is processed with a watermark image (copyright logo, signature, and so on), and then halftoning of the watermarked image has been done to limit pixel expansion. Ten, the halftoned image has been partitioned using VC into two shares. Tese shares are embedded with a QR code to make the shares meaningful. Lossless compression has been performed on the QR code-based shares. Te compression method employed in visual cryptography would save space and time. Te proposed approach keeps the beauty of visual cryptography, i.e., computation-free decryption, and the size of the recovered image the same as the original secret image. Te experimental results confrm the efectiveness of the proposed approach.


Introduction
In most cases, watermarks are translucent, allowing viewers to continue to appreciate the picture even when it contains the watermark. In most cases, one can determine who the photographer is just by looking at the watermark. Te copyright may also be protected by adding a watermark to the photographs. It is protecting the picture from being used by other parties without the consent of the original owner. Capacity, robustness/security, and imperceptibility are the characteristics of a digital watermarking system [1]. Te simultaneous existence of these three needs is what diferentiates digital watermarking from other data-concealing methods.
Visual cryptography is employed in several domains such as data concealment, picture security, color imaging, multimedia, and other related felds. Te topic of data concealment, which is employed in cybercrime, fle formats, and other areas, is where visual cryptography fts in. Te fact that it does not need complicated calculations on the part of the receiver is the primary beneft that the visual cryptography scheme ofers in comparison to traditional encryption methods. Te purpose of this paper is to overcome the drawbacks of previously developed cryptography methods by presenting a new visual cryptography system that maintains the hue of a secret picture.
Te proliferation of information sharing in the form of digital photos presents risks to individual rights to privacy and to their individual identities. While digital photographs are being sent, they are vulnerable to theft, and it is simple to make any type of change to them. As a result, the protection of digital photographs' privacy from malicious actors is becoming more vital. A variety of methods, including encryption, steganography, watermarking, and visual cryptography, have been suggested as potential safeguards for digital photographs on occasion.
Te purpose of this paper is to investigate the use of watermarking and a QR code-based expansion-free and meaningful visual cryptography approach to improve the level of privacy protection aforded to digital images. Te approach produces QR codes that are visually appealing and are capable of transmitting meaning shares. After the processing of the original secret picture with a watermark image (which may be a copyright logo, a signature, or anything else), the watermarked image is then halftoned in order to restrict the growth of the image's pixels. After that, compression of the halftoned image has been done. Ten the compressed halftoned picture is split up into two separate shares using VC. Tese shares have been given signifcance by the incorporation of a QR code into their structure. Te compression used in visual cryptography would result in savings of both time and space.
Te suggested method maintains the attractive qualities of visual cryptography, namely, computation-free decryption, and it ensures that the size of the recovered picture is identical to that of the initial secret image.

Literature Review
Te literature review undertaken for the proposed approach is given in Table 1.
Hsu et al. focused on a multisecret method of visual cryptography based on a circular layout. In order to protect a number of confdential details, a novel visual cryptography technique based on a circle is developed. Te circular stock was subdivided into an outer ring and an inner ring. Both secrets are located on the outer ring, with the second being disclosed by rotating one share at a specifc angle relative to the other. Two additional surprises may be found in the inner ring. Prior to stacking, one of the shares must be turned over to disclose the frst secret. Te second mystery may be solved by rotating one of the shares in the fipped position through an angle. Te shares are encrypted with four secrets that can only be accessed by fipping and rotating them. No one can steal the information in the shares without knowing the angles to turn or which share to fip [2].
Feng et al. provided multiple secrets that may be shared visually. Since traditional visual secret-sharing systems are optimized for a single secret picture, it was inefcient to produce a large number of shared images for a large number of secret photos all at once. As a result, this study proposes a unique visual secret-sharing strategy for a set of secret pictures. To illustrate the encryption functions, the suggested method generates a stacking connection graph of secret pixels and share blocks and then uses this network to build a set of visual patterns that generates two share pictures. Te secret pictures may be derived from the two sharing photos by using the stacking features of these patterns. Te suggested technique allows for an unlimited number of hidden pictures and was also very fexible and adaptable. In this way, the suggested system improves the capacity of visual secret-sharing schemes for keeping a number of secrets at once [3].
Wang presented visual encryption based on a random grid that may be used to identify individual shares. In this research, we ofer a VC technique based on a deterministic RG. By using this technique, we may encode a picture O into two parts, each of which displays the following characteristics: First, each generated share was an RG on the same scale as O; second, no single share reveals any secret information about O; third, the secrets can be revealed by superimposing the two generated shares; fourth, folding up a share will display the identifcation patterns associated with it; and ffth, both the secret information and the identifcation patterns are recognizable to the naked eye without any computation. Visual cryptography methods have never before had the ability to display identifying patterns by just folding them up. It develops a simple and user-friendly interface for users to diferentiate between and manage the multiple shares formed by VC schemes, and it gives important information for identifying the pair of shares belonging to a hidden picture. Tis article was published in the 2011 editions of SPIE and IS&T [4].
Hsu et al. introduced cryptography that can be checked by sight. Tis work introduces a technique for verifying the integrity of VC shares through-out the decoding process. Te concept was to imprint a continuous pattern on the stock certifcates that were part of the same secret picture and then to stack and position the certifcates such that a portion of the pattern is exposed. Evidence for the genuineness of the shares involved in the decoding process comes from the visual consistency between the displayed patterns of all pairs of shares. In contrast to VC methods stated to avoid cheating, the suggested strategy preserves the original pixel expansion in VC without any cheating prevention abilities, and it verifes shares without using an extra verifcation picture. Furthermore, any VC scheme in the literature may have the suggested verifcation method added to it in order to provide legitimate users with the capacity to deter cheating by malevolent players [5].
Lee and Chiu looked insensitive to image size generalpurpose visual cryptography with display quality limits. Traditional VC has issues with pixel expansion and display quality for recovered pictures that cannot be controlled, and it does not provide a framework for building visual secretsharing schemes for universal access structures. In order to solve these problems without resorting to complex code book design, they present a broad and systematic strategy. In the absence of specialized software, this method may be used to de-code binary secret pictures. Instead of utilizing the standard VC-based method, which might lead to pixel enlargement, they create a series of column vectors to encrypt secret pixels. To discover the column vectors for the best VC construction, they frst formulate a mathematical model for the issue and then design an algorithm based on simulated annealing. Te displayed quality of the recovered picture is shown to be higher than that of prior studies, according to the experimental fndings [6].
Askari et al. focused on a new pixel expansion-free visual cryptography algorithm for halftone pictures. To better retrieve hidden pictures in a VC scheme, they provide a novel approach for processing halftone images. In Advances in Multimedia particular, our method reduces the common issues of pixel expansion and contrast loss in VC. Tis work presents and exhibits the outcomes of two applications of VC on halftone pictures based on this processing stage: one application in multiple VC and the other application in extended VC. Both programmers work without any pixel enlargement, and the recovered secret picture looks far better than it did with previous methods from the literature, all while retaining the full security of the pioneering VC method [7].
Petrauskiene et al. reviewed the optical evaluation of chaotic oscillations with the use of dynamic visual cryptography. To evaluate chaotic oscillations optically, they ofer an experimental optical method based on dynamic visual cryptography. One cover picture, with the hidden image embedded within it, was afxed to the surface of the resonating object. Although time-averaged moire´fringes do not develop when the encoded cover picture was oscillated by the chaotic law, it was shown that this visual scheme was useful for the evaluation of chaotic oscillations. Visual examination alone was sufcient for decoding, allowing one to ascertain whether or not the parameters of the chaotic oscillations remain within the acceptable range [8].
Shankar and Eswaran explained visual cryptography's newest k-out-of-n secret image-sharing scheme. Tis study proposes a novel method of (k, n) visual cryptography that may be used to securely share a secret picture. In order to construct the required "n" number of transparency shares, the share generation procedure frst applies a newly stated condition to a set of random matrices before performing XOR operations on them. Te hidden picture may be cracked visually by superimposing a k-subset of see-through layers. However, the superposition of an unauthorized subset cannot provide any confdential information. Te suggested scheme's reliability is verifed by a battery of experiments, statistical evaluations, and security audits on the stocks, including visual testing, encryption quality testing, security analysis, and various assaults. When sending pictures over open networks, the suggested (k, n) VC technique provides reliable security [9].
Punithavathi and Geetha provided biometric template security cancellation using segment-based visual cryptography. Tis article discusses the numerous dangers that a removable biometric system can face. For this reason, it is especially important to protect against attacks that aim to retrieve a person's transformed biometric data from a template database. Researchers also present experimental fndings on a system that combines cancellable biometrics with segment-based visual cryptography, a technique that transforms conventional biocide into novel structures known as shares [10].
Chaturvedi and Bhat proposed improvement schemes for segment-based visual cryptography. Te two primary methods suggested in this category are the seven-segment and sixteen-segment displays, and they were both segmentbased rather than pixel-based and operate on symbols that may be shown as segments. As its proponents put it, "It was trivial to change the hidden pictures and potentially easy to detect for the human eye," making segment-based encryption a powerful tool for hiding sensitive information from prying eyes. In this study, they provide a summary of the several segment-based visual cryptography algorithms available. Tree strategies have been the primary focus of our investigation: 1: confdential information ex-change; 2: symmetric key distribution 3> planted separation by developing regions. Potential areas of investigation for the foreseeable future include automatic seeding algorithms [11].
Chao and Fan introduced an adaptive priority-based secret-sharing technique for visual data generated from a random grid. In this research, they provide a method for progressive visual secret-sharing using a random grid in which the relative importance of each share may be modifed. Using this method, a hidden picture may be retrieved from the company's shares in a roguish manner. So, the more shares you gather the more of the hidden picture you can piece together, and vice versa. Each participant in the secret sharing may also choose their own priority weighting for a share, so that the suggested method can produce shares with varying priority weighting values depending on the desired degree of privacy. It was possible to retrieve several portions of the hidden picture during decryption, all of which are dependent on the relative importance of the stacked shares. In addition, the average light transmission of the reconstructed picture cannot be used to determine the priority level of these shares, ensuring great security [12].
Wu and Yang concealed information visually with a random grid generator. Te vast majority of VSS protocols were used to transmit encrypted binary secret pictures. In these methods, each sharing was disguised as random noise, making it impossible to deduce anything about the hidden picture from any one piece of it. After being rendered with two distinct light transmissions, the recovered secret message may be deciphered by the human eye alone. Gray-level secret pictures are particularly inaccessible to these methods due to the in-ability to retrieve the halftone area. To recover the secret in a multilevel secret picture while maintaining a high degree of visual quality during decryption, they present a VSS technique with multilevel encoding based on a generalized random grid. Te suggested approach does not call for a codebook during encoding and yields a share of the same size as the original secret picture without pixel enlargement [13].

Problem Statement
In the proposed research work, as shown in Figure 1, data security is being considered in relation to QR code-based visual cryptography and watermarking. Previous study work had a problem in that it used traditional methods, but the proposed method overlooked picture quality, security of shares and secret image, and performance. As a result, the suggested study takes into consideration a hybrid method to safeguard multimedia data. Finally, a comparative study of picture quality, performance, and data security ensures that the suggested model provides superior-quality images. Te performance of data transmission is also improved by compressing image fles.

Proposed Methodology
Te objective of the proposed methodology is to keep the size invariant and achieve meaningful shares of the image. To achieve this objective, we use a hybrid approach using watermarking and visual cryptography. In the proposed work, the cover image is processed with the watermark image, and then halftoning of the watermarked image is performed to reduce pixel expansion. Te halftoned image is split using extended VCS into two random shares (share 1 and share 2). A QR code is embedded in both shares in order to make the shares meaningful. Tis step makes the share-handling process easy. After that, the compression of QR-based shares has been done using a lossless compressing mechanism. Finally, the quality of the restored image is evaluated using PSNR, SSIM, and MSE. Te term peak signal-to-noise ratio (PSNR) is an expression for the ratio between the maximum possible value (power) of a signal and the power of distorting noise that afects the quality of its representation.
Mean squared error (MSE) measures the amount of error in statistical models. It assesses the average squared diference between the observed and predicted values. When a model has no error, the MSE equals zero.
Te SSIM index is calculated on various windows of an image. Te measure between two windows x and y of common size N × N is calculated. Te proposed approach is having following stages, as shown in Figure 2: 4.1. Embedding of the Watermark. During the embedding process of the watermark secret image has been considered to perform wavelet decomposition, and it is split afterward. Te watermark image is split as a binary watermark into multiple blocks. Finally, these blocks are embedded into selected wavelet coefcients. Finally, resultant values are obtained after watermark embedding, and the watermarked image is got after wavelet reconstruction. Te algorithm that embeds the watermark takes a secret image (CI) as input and produces a watermarked image (WC) as an output (Algorithm 1).

Extended Visual Cryptography.
Visual cryptography is a cryptography scheme specially designed for images [14]. It is a data security mechanism in which the original secret image is divided into a number of shares. Tese shares are then printed onto the transparencies and given to participants involved in the cryptography process. When a subset or all shares are superimposed, the original secret image is recovered without any need for computation [15]. Te concept of visual cryptography was given by Naor and Shamir in 1991 [16]. But traditional cryptography sufered from two limitations. First, the generated shares were random and meaningless [17]. It becomes difcult to manage random shares. Te second drawback is pixel expansion. Due to this, the size of the shared images is larger than the secret image. In the proposed approach, we propose expansion-free extended visual cryptography. In order to remove pixel expansion, a block-based halftoning operation is done instead of pixel-by-pixel encryption to maintain the size of the secret block and share block (Algorithm 2).

Embedding of QR Code.
A QR code, which stands for "quick response code," is a type of matrix barcode (also called a "two-dimensional barcode") that was made by the Japanese company Denso Wave in 1994 [18]. A barcode is an optical label that can be read by a machine and can have information about the thing it is attached to. In real life, QR codes usually have information for a locator, identifer, or tracker that leads to a website or app. QR codes use four standardized encoding modes to store data efciently: numeric, alphanumeric, byte/binary, and kanji. Applications include keeping track of products, identifying items, keeping track of time, and managing documents [19].
A QR code is made up of black squares in a square grid on a white back-ground, along with some fducial markers. It can be read by an imaging device, like a camera, and then processed with Reed-Solomon error correction until the image can be understood. Ten, the needed information is taken from patterns in both the horizontal and vertical parts  Advances in Multimedia of the image. In this stage, the embedding of the QR code into the halftoned compressed shares has been done in order to make the shares meaningful. Te embedding of QR codes in shares does not make much diference in the size of the shares. Moreover, the compression of halftoned shares has also been done in the proposed approach. So, the transmission cost of shares will not be increased. Te embedding of QR codes in the generated shares has been done in order to make random shares meaningful. Tis step makes share handling easy and efcient. Te embedding of the QR codes has been done in the generated shares to make the random shares meaningful and make the share-handling process easy and efcient. In traditional VC, generated shares are random in nature. So in the case of multiple random shares, share handling becomes a tedious task. Tat is why embedding QR codes in the generated shares have been done.

Experiment and Results
Simulation work has been performed on MATLAB environment where the image processing toolbox has been considered for image processing. Several image-processing algorithms are also applied in the present simulation. During the simulation process, the cover image of "Lena" has been considered, and a frog image is considered for watermarking. Figure 3 presents watermarking process where the watermark image has been embedded into Lena image to produce the watermarked image. Tis watermarked image is processed by a halftoned secret image in order to produce share 1 and share 2. Both shares are compressed using a compression mechanism, as shown in Figure 4. Finally, compressed halftone share 1 and share 2 are embedded with randomly selected QR code to produce compressed halftoned QR code-based shares, as shown in Figure 5.

. Various Evaluation Parameters Used
(1) PSNR (peak signal-to-noise ratio): Te term peak signal-to-noise ratio (PSNR) is an expression for the ratio between the maximum possible value (power) of a signal and the power of distorting noise that afects the quality of its representation.
PSNR � 20 log 10 (2) MSE (mean squared error): Mean squared error (MSE) measures the amount of error in statistical models. It assesses the average squared diference between the observed and predicted values. When a model has no error, the MSE equals zero.
Where Leena Image Watermarked Image Watermark image (a) f represents the matrix data of our original image (b) g represents the matrix data of our degraded image in question (c) m represents the number of rows of pixels of the images and i represents the index of that row (d) n represents the number of columns of pixels of the image and j represents the index of that column (e) MAX f is the maximum signal value that exists in our original "known to be good" image.
(3) SSIM Te SSIM index is calculated on various windows of an image. Te measure between two windows x and y of common size N × N is

Comparison Analysis of the Proposed Approach with the Conventional Approach
(1) Te halftoning operation performed on the secret image ensures that the information about the secret image will not reveal (2) Teproposed approach is using QR code-based visual cryptography technique. We will get QR codebased shares after splitting the original image. Te problem of meaningless and random shares has been improved. Moreover, share handling becomes easy (3) While during transmission, QR code-based shares do not attract the invaders. Also, these shares do not reveal anything about the original shares (4) Compression of the shares has been done before transmitting the shares. Tis will save time and space for the transmission of shares.
A comparison analysis of the proposed approach with the conventional approach is shown in Table 2. Te experimental results confrm the efectiveness of the proposed approach.

Conclusion
Te aim of the proposed methodology is to keep the size invariant while achieving meaningful shares of the image [13]. By combining watermarking with visual cryptography greatly improves the privacy protection of secret images.
Simply using the watermarking technique to protect the images has many limitations, as watermarks can be easily removed. Visual cryptography provides a solution to protect the images by splitting the original secret into the number of shares. Te original secret image can be recovered by superimposing a subset of shares or all the shares. But traditional visual cryptography involves the creation of meaningless, share and size-invariant recovered images. Tis paper proposed watermarking using QR code-based visual cryptography techniques for the privacy protection of digital images. Tis approach generates visually appealing QR codes for transmitting meaningful shares that can be stacked to recover the original secret image by the human visual system without the need for any computation. Te experimental results confrm the efectiveness of the proposed approach. In the future, we will explore new techniques that can be combined with visual cryptography in order to improve the quality and contrast of the recovered image.

Data Availability
Te data used to support the fndings of this study are available from the corresponding author upon request.