Evolutionary programming approach for securing medical images using genetic algorithm and standard deviation

ABSTRACT


INTRODUCTION
The modern era is going through a rapid technological volution, and we are witnessing the production of a huge amount of information that can be worthily enhanced with appropriate management and analysis.This advancement in digital technology also raises certain concerns about the illicit access to and dissemination of multimedia information.The genetic algorithm (GA) has been introduced as a solution to these problems [1].Therefore, we work to definitively represent the applications of GA in medical sciences, we note that the genetic algorithm finds its way in various fields of medical sciences such as image segmentation, personalized healthcare, and radiology, and discusses how to successfully apply the principle of the genetic algorithm in these applications.also try to have a comparative discussion of selected apps on different criteria.In general, inserting of hidden text causes degradation of image quality.This degradation is not visible to the human eye.Furthermore, the images are affected with distortions caused by active or passive attacks.
The hidden text procedure must be robust against all the attacks.According to the visibility, domain and permanency, data hiding procedures are classified in to different types.Medical information is the TELKOMNIKA Telecommun Comput El Control  Evolutionary programming approach for securing medical images using … (Shihab A. Shawkat) 1287 prerequisite for any medical treatment nowadays.To obtain the optimal outcome for any diagnosis, several tests have been proposed in a manner that is cost and time effective and that preserves the confidentiality of diagnostic data [2].Therefore, we work to definitively represent the applications of GA in medical sciences, we note that the genetic algorithm finds its way in various fields of medical sciences such as image segmentation, personalized healthcare, and radiology, and discusses how to successfully apply the principle of the genetic algorithm in these applications.also try to have a comparative discussion of selected apps on different criteria.
According to the findings of the literature review, information security is a key factor in determining the success or failure of future progress.There has been a significant growth in the number of information preservation techniques [3], due to the fast advancement of technology.However, security has also become a serious problem as a result of the increasing number of security risks.As a result of widespread digiti-zation, huge amounts of data are transferred from one network to another.An attacker or intruder uses a variety of methods to monitor the network to obtain important information [4].Now with the development of communication and information, information access has become easier, and establishing secure communication has become an essential requirement [5].To ensure secure communication, different methods have been developed.Information steganography is one of the methods exploited in covering data.It is defined as a science or a technique that helps in transmitting information discreetly on a carrier of multimedia (audio, image, text, and video) [6].
Steganography represents a technique used for masking information and data into a cover object to create what is so-called a stego-image.The stego-image is sent by the sender to the recipient by a common object (medium).This prevents the outsiders (interlopers) from noticing the shrouded message in the image.The secret message in the image can be easily extracted with or without a stego-key by the recipient, with the help of an embedding algorithm.The concept of steganography of images is explained in Figure 1.The embedding algorithm requires a cover (image) with a secrete (message) [7].
Data hiding aims at delivering data reliably from sender to receiver without any degradation, alteration, or any third-persons (hackers).Nowadays, steganography has changed the way we look at the development of technologies [8].To prevent hackers or others from seeing or knowing the secret information is not the only purpose of steganography; it also tries to remove any suspicion about the hidden mes-sage or information.The information or the message represents credentials sent and disguised in the medium so that it is hard to detect.Any steganography has two main aspects, the ability to hide information and the inability to recognize hidden content [9].However, these two characteristics confuse each other, as it is hard to raise the capacity while the steganography system is maintaining its imperceptibility.Besides, limited information cloaking methods are still there with protocols of data transmission communication.Such protocols can be unusual, yet, their future is promising [10].
A GA refers to a developmental and computational model influenced by the method of biological evolution.GA is usually used as a function enhancer.GA represents a model employs operators like recombination and selection to create new individuals in the search space, and to improve objective functions.Typically, a genetic algorithm includes two principal problem-based constituents the coding and evaluation functions.Some variables are optimized by either maximizing certain objectives or the error function [11].Optimizing objective function is sup-posed to be a black box of a control dials series characterizing various criterion.The output of the black box refers to the extent to which the evaluation function returns on the optimization problem through a set of parameters.Standard deviation (STD) refers to image contrast change the bigger the value, the clearer the edge contour.To determine both the extreme value and the average value, it is advantageous to figure out the population STD.Calculating STD makes it simple to determine the frequency distribution.As the STD of AC increases, the image becomes more obvious [12].
The integrity and security of medical data have become a big challenge for healthcare services applications.Digital healthcare requires the process of transferring medical data, and has become a daily routine [13].It provides an incorporated communication environment of interdependent platforms and devices by connecting both physical and virtual aspects.Thus, it is important to improve an efficient model to achieve the integrity and security of the transmission and reception of patients' diagnostic data [14].This goal is achieved by employing algorithms of system encryption and techniques of steganography together to mask digital data in an image [15].
The proposed technique has important benefits where it requires no static cover image.Moreover, high quality and resolution are achieved in the pilot study conducted using texts and JPG images.With the basis of the integration of steganography technology and cryptography scheme for a high security system of healthcare, the technique aims at improving the security of transmitting medical data.The study is divided into five sections in addition to the current section.Section 2 explains the previous works related to the subject studied; section 3 illustrates the proposed and suggested model and the algorithms used; section 4 deals with the discussion of the results of the experimental findings.Finally, section 5 is concerned with the main conclusions.

RELATED WORKS
There are many uses for digital images, and sometimes they are used in some applications and important areas.Data should be safeguarded against unauthorized access, where the attackers should not be able attack throughout transmission.The outcomes of such works have, nonetheless, slightly improved.This section discusses a few of previous studies conducted by some authors.A very important research direction was introduced by Arsalan et al. [16].It presented genetic programming (GP) depending on a reversible watermarking technique.To inset the optimum level of a watermark in the chosen pixels and to choose a domain of coefficients in integer wavelet transform, the researchers use GP in their technique.In its coefficients, the least significant bit (LSB) embeds watermarks.A trade off between imperceptibility and payload is gained by using the sub band category of wavelet coefficients and amplitude information.In this case, GP introduces GA, depending on the reversible integer wavelet transformed commanding technique.To obtain payload and a higher level of imper acceptability, GA is employed for the coefficient selection.
Anwar et al. [17] proposes an algorithm which depends on an artificial immune system and image segmentation.A block has been chosen following the algorithm proposed and after the cover medium (image) is segmented.Then, the artificial system of immune is used to cover the bits of the message in the best available place.In a domain of frequency, a data hiding approach is presented by employing a genetic algorithm.First, the cover medium (image) is mapped to a suitable domain with the use of a genetic algorithm and adaptive wavelet transform.The data and information to be covered are hidden and encrypted in frequency coefficients characterizing the edges of the image in the spatial domain.A hybrid model for securing data of the diagnosis text in medical images is suggested [18].The model suggested is improved via integrating one of the following techniques: 2-D discrete wavelet transform 1 level (2D-DWT-1L) or 2-D discrete wavelet transform 2 level (2D-DWT-2L).However, either one is done with a proposed hybrid scheme of encryption.To start with, it encrypts the hidden data; then it conceals the output in a cover image.To conceal various sizes of text, images of both color and grayscale are employed as cover images.
Ahmadi and Sajedi [19] an approach which uses picture segmentation and a system of artificial immunity is suggested.According to the suggested approach, the system of artificial immunity is used to cover the bits of the message in the most suitable location.This is done after the cover picture has been split into blocks.The cover picture is firstly transferred to the proper frequency domain using a genetic algorithm and adaptive wavelet transform.Then, a method for concealing data in the frequency domain is described.The information is encoded and concealed in frequency coefficients that represent the spatial edges of the picture.Elhoseny et al. [20] presents a model of hybrid security as a suggestion for protecting the textual diagnostic information in medical pictures.The suggested model is made by combining a suggested strategy of hybrid encryption with one of the following steganography approaches: the 2D-DWT-1L or 2D-DWT-2L.First, the secret data is encrypted; after that, a cover picture is used to conceal the outcome.To hide various text sizes, cover graphics in both color and grayscale are employed.
Karakus and Avci [21] states that using stego-images and increasing the amount of data to be concealed are to guarantee high image quality by making use of pixels similarities.As a cover medium, medical images of different sizes are used.The doctor's comments on the situation of the patient are covered in medical images of 256×256 size without utilizing any technique of data compression.A robust and reversible framework of watermarking and a secure multiparty computation are used to achieve encrypted images [22].To embed a watermark, the protected multi-party computation method is used where the scheme is not assessed for hybrid attacks and the system computational complexity is greater.
In the open environment, in Shawkat et al. [23] argues, the cloud infrastructure resources are shared via cloud computing on the internet.This maximizes the problems of security such as data integrity, confidentiality, and availability.Therefore, such problems can be solved by adopting data encryption, which is important to secure users' data.A comparative study is conducted between the two algorithms of security in TELKOMNIKA Telecommun Comput El Control  Evolutionary programming approach for securing medical images using … (Shihab A. Shawkat) 1289 a cloud platform known as eyeOS.The study concludes that the algorithm of Rivest Shamir Adlemen (3kRSA) outperforms the algorithm of the triple data encryption standard (3DES) in terms of output bytes and complexity.
A fusion method for multimodal medical images is proposed in Arif and Wang [24] based on the curvelet transform.The GA method depends on the curvelet transform to solve the wavelet transform drawback in processing curvy forms and contours of objects that often appear in medical images.Any diffuses and suspicions existing in the input image can be solved by the method, which can further improve the characteristics of image fusion.The method proposed has been verified and tested in various places of medical pictures.Also, it is compared with new techniques of medical image fusion.

THE PROPOSED MODEL
This method ensures that the ciphertext and the stego-image (image after embedding) maintain the highest level of image authenticity and quality.By utilizing the potent qualities of crossover and mutation GA operations, the suggested approach employs a GA twice, for producing an approach for encoding the text and picking the optimum region to mask.STD for each section is then finding value of a standard deviation, (i.e., be adopted as locations to hide) and that by including all bit of message.The algorithm was applied to more than one image, where the standard deviation value is measured, as the lower the dispersion value of the image data, which indicates better results, as the peak signal to noise ratio (PSNR) value was high, which indicates that there is no visible change in the image after masking compared to the cover image.On the basis of the GA and STD, a novel scheme of text encryption and embedding is suggested to maintain the security of medical image files.On a few random texts and well-known images, the encoding technique suggested in this paper is tested, and promising results are noted.

Encryption and embedding text
The GA is used, in the process of optimization, to optimize (1).To find the value of (1), let's suppose the count of 0s is shown by  and the count of 1s is by .Besides, we have  which shows a threshold value.Next,  1 ,  2 and  3 are calculated and are added as in (1) to conclude the objective function of the key of the input.Thus, the value of 1 s and 0 s in the key is calculated and that of  1 is returned to either one in the condition of equality and zero in inequality.Whereas, the second function begins when the significance of  is entered.It refers to the required value of block units, where  = 8.The value of n in the key string is searched for and the value is returned to either one in the case of a size block  or zero in the absence of a block of size .Besides, it is returned to either one in the case of a measure gap  − 1 or zero in the absence of the presence of a gap of measure  − 1.
The GA main section is the fitness function, and it is used to evaluate whether a member is useful or not.A new fitness function, in this study, is presented for GA.It can maximize the true positive rate and lessen the rate of error in choosing useful features.In various images such as medical images, the represented features of the image are only the ones used to make better and more accurate detection.Thus, it is important to extract suitable features of the image rate in addition to improving quality.They are used: Here, to create a stronger key other than the one resulted from the previous process, evolutionary optimization is applied.The same standards followed in the previous step are used here in the process of development in (4) as an objective function.Anyhow, it is conducted iteratively.Four new arbitrary solutions are generated at the beginning of this process.Thus, they represent the solutions of the population, (4).Using (4), each one of the four proposed solutions is simultaneously evaluated till we find the best or better solutions after some generations.The process of optimization continues till sufficient merit of the fitness function is achieved for any of these candidate solutions.During the successive generations, it is expected that the solutions that have high ranks will be exploited for the encryption of the message.
Figure 2 shows the key generation process with the use of GA in encryption.Firstly, the length of the key must be identified to determine the working parameters of the GA before starting the process of optimization.The proposed GA parameters are as: The plain or data text will be transformed into a certain format of image as an outcome of this action.The text to be encrypted will be hidden in that image.To view the text by the receiver, the image must be segmented first into blocks.The concealed key will be exploited to make some adjustments to each color component.Hackers will face a big problem in breaking the encryption and stealing the information will be impossible.Figure 3 explains the general framework for concealing the secret data by the proposed systems.These figures represent the main steps performed on both sides by the sender and receiver.

Extraction and decryption text
The encoded image will be loaded onto the ciphertext along with the underlying in-formation for that text.The image will then be converted to a grayscale image to facilitate further processing of the image.This is to extract sensitive information en-coded within the image.It is more efficient than using the native RGB image format.While, the grayscale image contains less information than the RGB image.Thus, this transformation is a TELKOMNIKA Telecommun Comput El Control  Evolutionary programming approach for securing medical images using … (Shihab A. Shawkat) 1291 way to extract features that only relate to valuable information about the encrypted data.In the context of our proposed method, the decoding process represents the extracting operation of the original message.The decoding first starts by splitting the 1-D element matrix frame and retrieving the final element representing the message length.Next, the location of the key is extracted with two of the removing the appended zeros.After that, all the added zeros of the hidden infor-mation are deleted.Here, the location of the message must be discovered by the extraction of the key and the secret information via its location.Eventually, it can be extracted by the frame hash number.To transform the binary array of the key of the information to decimal numbers is so important.
Later, the image is passed to another major step, which is the process of dividing only the image specific parts, where the secret message or information is stored.This image is then parsed into two components.The image index is the typical block identifier, where the hidden text fragments will be extracted.At this moment, the information required for the hidden text is returned and the cover image from which the text is extracted is compared with the original image, so that we can evaluate the image accuracy.The method proposed is subject to standard evaluation measures in the field to avoid possible attacks on the ciphertext.Figure 4 shows the mechanism for extracting the text from the cover image using the proposed method.

EVALUATION OF THE PROPOSED METHOD 4.1. Dataset and used parameters
The illegal obtaining or changing the personal data of the patient may result in the violation of the patient's personal rights and change the course of the treatment process.The way patients' personal information and data are stored and transferred must be preserved because they are vulnerable to illegal appropriation or violation.This may cause harm to the patients.In this study, been used cover objects, 15 images of (512×512) size type of computed tomography (CT) and magnetic resonance (MR), chosen randomly from the library of digital imaging and communications in medicine (DICOM) [26].The integrity and security of medical data have become a big chthe patient's personal information and message generated by the doctor's comment are stored in the images.Our aim with this work to mask the secrete data of patients into a cover image transmitted with high ability, imperceptibility and in order to get new dataset (grayscale and colour) images of chosen randomly (15 images), The used images of the test are as shown in Table 1.It shows the gray and color datasets versions.
Our suggested model is implemented with MATLAB R2021b software operating on a PC with 8 GB of RAM, a 2.27 GHz Intel® Core™ i5 CPU, and Windows 10.The computation of selected statistical values specifies the proposed-model quality of security.These scales compute the comparison between the stegoimage and the original image (cover).The achieved results will be assessed on images of both color and grayscale with various text sizes.Here, the evaluation of performance depends on four statistical parameters structural similarity index measure (SSIM), mean square error (MSE), peak signal to noise ratio (PSNR), and bit error rate (BER) [27].

PSNR
PSNR is typically employed to calculate the proportion between the maximum potential power of signs and the power of corrupting distortion that influences image accuracy.Mainly, the ratio is used to differentiate between the quality of a stego-image containing hidden data to that of an empty cover image.In ( 5) is used to determine the PSNR value.
MSE MSE is simply the average squared disparity between a cover image and a stego-image.A lower MSE value means fewer variations between both images.In ( 6) is used to determine the MSE value.

SSIM
It is an index that gauges how structurally similar two images are.Its range of values is from -1 to 1.The SSIM of two almost similar photographs is close to 1.As a result, it is employed as a technique for determining how similar two images are.To calculate the SSIM between (1) and (2) at a certain pixel .The SSIM value is calculated using the in (7).

BER
The number of errors made concerning the total sum of signals sent is known as BER.The performance of the system improves when the percentage of errors decreases and vice versa.Multiple disturbances during transmission have an impact on the signal, and this could be immediately translated into the number of errors in a string of a transmitted number of bits.With the use of ( 8), the definition regarding BER could be reduced to the following straightforward formula. =         ⁄ (8)

Experimental results
The technique presented in the section 3 was put to use in analysis metrics of used visual quality, where creating a simulation environment is essential to validate the performance of the proposed algorithms.In this study, the patient's personal information and message generated by the doctor's comment are stored in the images.The GA is used twice, in text encryption and in choosing the ideal block for hiding.Also, the STD chooses the ideal block for hiding.Whereas, working on the main goal of obfuscating images is to preserve the privacy of the contents of the image, which is highly desirable in the privacy request.It specifies the match index of the original block (before hiding) to the selected best block.Image pixels and message bits are  Evolutionary programming approach for securing medical images using … (Shihab A. Shawkat) 1293 exchanged sequentially.The best cluster is randomly selected using the GA and STD methods.The method, depending on optimizing and taking benefits of the similarities between pixels, has been proposed.High image quality can be achieved by using stego-image and increasing the data amount to be hidden.Analysis metrics of visual quality such as PSNR, MSE, SSIM, and BER are adopted to assess the performance of the method proposed.Figure 5 Experimental results of the three proposed techniques are presented with the images from various sources utilized for evaluating our suggested techniques.With both conventional and medical images, we use the recommended method.Thus, we observe the performance of the proposed smart technologies.The similarity-based method proposed by Elhoseny et al. [20], which saves the diagnosis text data in a medical image using a hybrid security paradigm.The suggested methodology encrypts sensitive data first, then conceals the outcome in a cover image.To conceal the various text sizes, a color image is utilized as cover images (128,256) bytes.It is different from the one that proposed by Karakus and Avci [21], Anand and Singh [28].The method we proposed is of different capacities of texts, (1000 characters between 142 words and 250) and (5000 characters between 714 and 1250 words).It uses words to mask the secrete data in medical images.The model proposed demonstrates its efficiency to mask the confidential data of the patient in the cover image sent with high non-perceiving ability, amplitude, and minimal degradation in the received groove image.The comparison results are shown in Table 2, Table 3, Table 4, and Table 5.
Table 2 shows the comparative performance of the outcomes of the four obtained statistical parameters after applying the suggested technique with Elhoseny et al. [20] on color images for (128,256) bytes.In these cases, the results are quite close to each other.The data are concealed, and efficient outcomes are obtained.
The suggested approach offers greater flexibility than the existing schemes, when the new system is compared to several well-known contemporary designs.As previously mentioned, evaluation criteria like SSIM, MSE, PSNR, and BER are exploited to assess the optical robustness and efficiency of the suggested technology.The highest PSNR of the scheme proposed is found in comparative studies, where no attack is louder than 75.7678 dB in Table 3, 68.6890 dB in Table 4, and 70.3197 in Table 5.The greatest MSE values are closer to one and are 0.0012331 and 0.0042543 in Table 3 and Table 4.The SSIM merits are nearer to one, as shown in the tables, ensuring that the proposed framework is not perceived.6 will show how the proposed technique compares in terms of speed while encrypting and decrypting color pictures (Lena image size: 512×512).Our suggested approach exhibits rather acceptable efficiency when compared to algorithms for similar environments, despite the fact that the prior environments were different.It demonstrates that because of how quickly it can encrypt and decrypt data, the technique we've shown is recommended for real-time applications.

CONCLUSION
The security and integrity of medical data has become a major challenge for healthcare services applications.it is challenging to find out the color component in the cover image for embedding in order to meet research objectives and withstand different kind of attacks.so, the main challenge faced while implementing secure, robust and imperceptible technique is combining cryptography and hidden text.Any illegal user is easily able to access important data and change their original owners in the current Internet domination scenario.As compared to others, our proposed and suggested technique outperforms many security models of recent techniques in pro-tecting diagnosis data in terms of imperceptibility and robustness.This is done to improve the embedding steganography and obtain high similarity between the cover image and the secret bit stream in a medical image with text.A robust GA and STD based scheme have been developed that provide better noncognition and potency.Investigations into the effectiveness of our plan for various imageprocessing activities reveals that the suggested method is resilient to simultaneous assaults.The suggested model is created to make sure that the stego-image (image after embedding) keeps as much of its original quality and distinctiveness as possible.In order to test the performance of the proposed method, visual quality analysis metrics such as PSNR, MSE, SSIM and BER have been used.Proposed model proved its ability to hide the confidential patient's data into a transmitted cover image with high imperceptibility, capacityty, and minimal degradation.The results show that the recom-mended strategy yields the best results in terms of nonvisual perception and solves the problems of classification confusion and improves the quality of the image.In future, Although, we believe that the research and contributions conducted in this paper are essential.It is possible to rely on other intelligent techniques to generate the key, such as the use of neural networks.In addition to the possibility of integrating more than one encryption algorithm and benefiting from the GA by generating the key to it, and the use of the well-known camouflaging function that achieves the required specifications.With encrypting images of various sorts and sizes, making it suitable for dependable and feasible cryptographic usage.

Figure 1 .
Figure 1.Conceptual diagram of embedding algorithm

Figure 3 .
Figure 3. Embedding the secret data into the source image

Figure 4 .
Figure 4. Extraction the secret data into the source image (a), and Figure 5(b) shows the algorithm's operation and experimental implementation.

Figure 5 .
Figure 5. Comparisons between the previous and posterior predictors visually (before hiding) the bestchosen block's matching index in dataset (a) color images and (b) grayscale images

Table 2 .
As compared to other methods for execution time (in seconds) Evolutionary programming approach for securing medical images using … (Shihab A.Shawkat)1295

Table 3 .
Size of the images used before and after embedding (1000-5000) character data hiding