Data Confidentiality in Healthcare Monitoring Systems Based on Image Steganography to Improve the Exchange of Patient Information Using the Internet of Things

Recently, with the availability of fast and reliable Internet, the distance between a patient and a doctor is becoming unimportant. Physicians will be able to request the medical images of their patients regardless of the geographical area. However, a lot of challenges face such successful implementation. To facilitate remote diagnosis, patient electronic medical record (EMR), including medical images, that originates in one system needs to be exchanged either within the same organization or across different organizations. Steganography is the practice of concealing a secret message inside a cover medium. In this paper, steganography will be used to embed the patient's personal information securely and imperceptibly in their medical images to enhance confidentiality in case of a distant diagnosis. The security of the medical data is improved to maintain confidentiality and integrity using IoT. The least significant bit of the approximate coefficient of integer wavelet transform is proposed. The distortion between the cover image and stego-image is obtained by measuring the mean square error and PSNR, and normalized cross-correlation is utilized to estimate the degree of closeness between the cover image and stego-image.


Introduction
Communication through digitized media has been increasingly evident with the development of the Internet. All individual and commercial communication takes place on the Internet, where computerized media is the primary means. When sensitive data from businesses and organizations is shared, the security of the information is a major problem.
ey are required to keep their data safe from interfering eyes. However, everyone in today's world has access to the Internet, so there is a great danger in transmitting data digitally. e conservation of data during transmission is addressed in this way [1].
In today's world, the confidentiality of secret data is paramount, and advances in the security of computers have positioned steganography as a superior technique for acquiring secured data. Steganography is the method of concealing secret data in a message, audio file, picture, or video by embedding it in another image, audio file, video, or message [2][3][4]. It is used to keep sensitive information safe from hackers. Nowadays, the volume of data shared via the Internet is expanding. As a result, data security is considered a severe concern when data is communicated through the Internet [5,6].
In steganography methods, each pixel of the cover picture is hidden with an equivalent number of secret bits. e embedding alteration in the cover image is equal. Individual pixels in a digital picture, on the other hand, have complicated statistical connections. As a result, the picture quality is automatically lowered while modifications with equal number of bits are made in the cover image pixel [7]. Different adaptive embedding techniques have been included within the steganography method to address these concerns. Each pixel of the cover picture is embedded with a changing number of bits using this adaptive embedding approach. As a result, the majority of researchers concentrated on adaptive strategies to increase the safety of the steganography approach [8][9][10][11][12]. Additionally, the value of edge pixels is unaffected by modifications made during the embedding process. As a result, edge pixels can hold more hidden bits than smooth pixels. In most applications, metaheuristic algorithms are employed to tackle optimization problems [13][14][15][16].
To safeguard data, IoT transmitted data in the cloud through the Internet is employed. High data security is provided via complicated encryption and decryption technology. Only encryption and decryption are used, but data concealment offers a greater benefit [17,18].
For security difficulties in the data communication between two devices in an IoT network as shown in Figure 1, several security criteria such as authentication, integrity, and secrecy were applied. Attacks are classified as low-, medium-, high-, and extremely high-level attacks based on their behavior and threat level. e two types of image steganographic methods are spatial domain and frequency domain [19]. Spatial domain approaches deal with the direct change of picture pixels, and while they have a higher payload and imperceptibility, they are vulnerable to statistical assaults [20]. Frequency domain approaches, on the other hand, use changed coefficients resulting from different transformations such as DWT, DFT, and DCT for data embedding. ese approaches are more resistant to image processing assaults, but they are computationally demanding and have a small payload, making them unsuitable for real-time applications [21,22]. ere are two ways to classify steganography techniques. Steganography methods are classified into image steganography, video steganography, text steganography, audio steganography, and network steganography, depending on the kind of cover image [23]. Steganography techniques are divided into two categories based on the embedding domains: spatial domain techniques, such as least significant bit and pixel value differencing approaches; transform domain techniques, such as discrete cosine transform (DCT), discrete wavelet transform (DWT), and integer wavelet transform (IWT); spread spectrum systems; masking distortion systems; and filtering methods [24].

Related Work
For securing information in an IoT architecture, three-color picture steganography algorithms are proposed. e first and third techniques employ red, green, and blue components for information transmission, whereas the second technique utilizes green and blue components. e dynamic positioning techniques were developed by utilizing the shared secret key to hide data in the deeper layer of the image channels [25]. e performance analysis of the secret image steganography technique for the security of images and data is discussed. For image steganography, modified LSB replacement and data mapping algorithms have been developed. Initially, the secret picture was preprocessed utilizing the data mapping method which was used to embed the secret picture in the cover image. In general, the majority of LSB approaches did not rely on pixel correlation or picture content. As a result, it might be detected through RS analysis.
Consequently, some preventive measures are required to enhance the security of LSB-based steganography approaches. An edge detection procedure has been implemented in the LSB-based steganography approach to address this problem. Furthermore, before incorporating the secret images into the cover image, they should be encrypted to increase security. Due to the lack of encryption methods preceding the embedding process, the security level of several current approaches has been decreased [26]. e data secured in the fog cloud IoT system is presented. e user in one region of the system uploads substantial data using the suggested quantum steganography convention and then transmits the protected data to the fog cloud. e intended receiver receives the data from the mist cloud and extracts the anticipated material using the specified extraction method [27].
A method for encrypting any form of an image, particularly medical images, has been developed. ey intended to secure the integrity of electronic medical data while also sustaining its availability and authentication to ensure that only authorized personnel had access to it. In the first stage, the AES encryption technology was used. With seven values retrieved from the ear image as feature vectors, the ear print is also included. By delivering medical pictures over the Internet, the suggested approach increased the security of these images and protected them from unauthorized access [28]. e decision tree approach is used to present a novel method for shifting the medical data of the patient by securing the data into the medical cover image. e encryption is performed in the form of several chunks which are disseminated uniformly. Secret numbers are allocated to the cover image in the mapping method based on the breadthfirst search to insert the data. Before embeddings, the data was encrypted using the RSA technique [29].
A safe approach that can meet the security and protection requirements while also overcoming the SPECS flaws is offered. In addition, this demonstrates the plan's success through execution assessments in terms of confirmation postponement and transmission overhead [30]. Developments and security adjustments attempting to solve the vulnerabilities of the security approach are presented to overcome the instinctive safety flaws of the 2-factor substantiation method. e proposed security enhancements may be used with the 2-factor authentication approach to achieve a more secure and robust two-factor client verification through WSNs [31]. For embedding reasons, the interblock approach is utilized. Images in the JPEG format are referred to as host or stego-images. is approach is only used to conceal patient information in medical JPEG photos. e difference in the coefficients is calculated using discrete cosine transform (DCT) [32].
In today's healthcare systems, Internet of ings (IoT) devices play a critical role. To incorporate patient data in any cover media, 2D DWT is applied. For the cover photos, grayscale and color photographs are utilized. Text data is encrypted using standard methods before being embedded in the cover medium. Various numerical methods are used to validate the imperceptibility of the cover image [33,34].
On the segmented image, the region of the object and the reversible watermarking technique are employed. If image modes such as X-rays, magnetic resonance imaging (MRI), or computed tomography (CT) images have been tampered with or forged, the presented methodologies perform effectively to identify the tampering using the hash code. As medical systems are more prone to fabrication or manipulation, reversible watermarking methods are particularly useful [35,36].
Steganography is a method for embedding data in many images, as compared to classical steganography, which uses just one image at a time for embedding. In the event of an exceptional state in the communication media during data transmission, secret data bits can be recovered from many shares. Compressed JPEG images are extensively useful for communication channels. An intermediate image is constructed before transmitting it to the channel, which is near enough to the stego-image [37].
A novel method for securing secret data in a fingerprint image created from a hidden message is proposed. Unlike traditional steganography techniques, there is no requirement for a cover signal for the embedding process. e secret message is transferred to the polynomial, encoded at diverse points of polarities, and utilized as a portion of the hologram to generate the fingerprint image [38][39][40].

Methodology
is section discusses the proposed methodology for maintaining data confidentiality during the IoT distribution process. Because information is passed across numerous hops in the Internet of ings, data security is critical. Due to the ease with which data may be accessed, the mixture of various gadgets and the interconnections established through a multitude of data give space for privacy breaches in IoT. As a result, in such a case, data may be secured by a reliable encoding technique. Accordingly, this study proposes a dependable data transmission model for a safe IoT connection, as shown in Figure 2, which is a depiction of a setup in healthcare that uses an IoT dispersed structure.

Steganography.
A steganography approach based on encryption is proposed for conveying secret data. Normally, a digital image consists of disparate picture parts known as pixels. As a cover picture, a grayscale image and a color image are utilized in this work. As a result, a picture is represented by a large array of bytes. Image encryption, embedding phase, quality improvement, and extraction phase are the four key aspects of the system proposed. is programme is commonly used in photos, although the method's characteristics are generally stated in some figures, including hash marking. Steganography protects against unauthorized users and illegal copyrights.
Steganography is a progression in which secret data is hidden so that its presence cannot be recognized. is is why steganography is sometimes referred to as "covered writing." e goal of steganography is not only to secure the encryption but also to hide it so that no one can detect or determine the presence of the hidden secret data.
is system or technique aims to hide the presence of any secret data. e person who is not permitted to push for knowledge access should not even know if any secret information is available.
e basic components of steganography are the message, the carrier and the stego-key. e message is that the secret text, image, video, or audio has to be safeguarded using the steganography process. e carrier is the path or medium through which the key and, hence, the covered message are sent. e stego-key is the password by which confidential data is protected and exposed as shown in Figure 3.

Image Encryption.
In the encryption process, the hidden image is processed with binary by plane decomposition which is utilized to decompose the image into binary bit planes. e image is represented with binary planes in this method for a decimal number which is given as (1) e grayscale image has the pixel value in the range of 0 to 255, which is decomposed into binary bit values. With the support of secret key binary, keystreams are generated. ese binary keystreams enter the two stages of the encryption model. e piecewise linear chaotic map is utilized to produce the keystreams, which are represented as Journal of Healthcare Engineering e control parameter is represented as δ, and x i provides the initial condition of piecewise linear chaotic mapping. e set of the hidden image δ is encrypted with the initial value x 0 e keystream X � x 1 , x 2 , . . . . . . x n is converted into the integer sequence X 1 (i).
e bit-plane decomposition utilizes the keystream into bit planes to obtain the binary sequence k 1 and k 2 in which the bits are arranged from a higher bit plane to a lower bit plane.

Diffusion Stage.
e diffusion stage is performed by the following steps.
(1) Elements M are added as (2) Cyclic operation is performed in M 1 to obtain the matrix M 11 , and M 1 element is shifted right by S 1 bits.   Journal of Healthcare Engineering (3) First element M 1 is encrypted, and the key of the first element M 1 is given as (4) Element P is given as (5) Cyclic operation is performed in M 2 to obtain the matrix M 22 , and M 2 element is shifted right by S 2 bits. (6) First element M 22 is encrypted, and the key of the first element M 2 is given as

Confusion
Stage. e steps performed in confusion matrix are given as follows.
(1) e elements P 1 and P 1 are added as given below: (2) e keystreams k 1 and k 2 are generated using the secret key k(x, δ). e initial value a 0 is generated using the following: (3) e chaotic sequence is generated as (4) e integer sequence X 1 and X 2 is given as (5) e row vector R 1 is obtained by encrypting the swapping elements P 1 and P 2 : (6) e row vector R 2 is obtained by encrypting the swapping elements P 1 and P 2 : temp � P 2 (i), (7) R 1 and R 2 are the row vectors which are transformed into n×m images to obtain the secret image.

Embedding Process.
e embedding approach involves some cover image and secret image preparation, as well as secret key extraction and data hiding. e cover picture was chosen based on certain conclusions drawn from earlier steganography studies. is should be done carefully so that the superiority of the stego-image created after hiding is preserved. Certain pixels or blocks are chosen from the cover image using a random key. Before embedding, the secret picture is compressed and encrypted. Compression reduces the quantity of data to be hidden, while encryption improves security. Even if the primary concern of steganography is exploited, data should not be exposed. e secret image is compressed using a sophisticated wavelet-based compression algorithm. Simple bit operations like AND and OR are used to encrypt data. After that, the secret picture is transformed into a bitstream, referred to as the secret data. e LSB approach is utilized to disguise the secret data in the chosen pixels. It is sufficient to swap the final two bits if the quantity of data is less. Otherwise, the secret data is swapped for the least significant 3 bits of the chosen pixels to generate the stego-image as shown in Figure 4.

LSB Domain Algorithm.
e algorithm for hiding a hidden text in an image is the LSB. e LSB embedding technique uses the secret text bitstream to be hidden to substitute the LSBs of the pixels in the cover picture. Because deviations in the LSBs of pixels do not produce variation in the image, the stego-image is virtually identical to the cover image. e pixel value I(a, b) of LSB is similar to message bit which is embedded in I(a, b), and it remains unchanged. e stego-image is obtained as follows: where m is the next bit for embedding each pixel by changing a bit. e pixels of an image must be adjusted to incorporate a hidden message. It is hard to differentiate between the cover image and the stego-image. is approach often generates significant distortion in the cover image when the number of hidden bits for each pixel reaches three. ere are many steganographic to be utilized to mitigate the distortion induced by LSB replacement. Adaptive approaches alter the number of concealed bits in each pixel, resulting in a higher image quality compared to systems that rely only on LSB replacement. However, this comes at the expense of lowering the embedding capacity.

IWT
Technique. An integer data set is transformed into another integer data set using the IWT. When the data is hidden in the coefficients of the wavelet filters used in the Journal of Healthcare Engineering DWT, any method that is most effective for the floatingpoint values of the pixels that should be integers may result in the loss of the hidden data, failing the data hiding system. To avoid difficulties with wavelet filters' floating-point accuracy when the input data is an integer, such as in digital images, the output data will no longer be an integer, preventing perfect recreation of the original image and preventing information loss through forward and inverse transforms. e lifting technique is one of the approaches for performing the IWT. IWT can transform integer wavelet coefficients from pixel values and recreate the image from integer wavelet coefficients due to its numerical advantages. e image pixel is decomposed into four subband wavelets of DWT: LL, LH, HL, and HH. e coefficient of the image is approximated using the LL subband, vertical details of the image are given using the LH subband, horizontal details of the image are examined in HL, and diagonal details of the image are given using the HH subband. e coefficient of IWT is computed as e coefficient of inverse IWT is given as e steps of the algorithm for the embedding process in IWT-LSB are as follows: (1) e cover image is read.
(2) e secret image is read.
(3) IWT is applied for the cover image. (4) Change the LSB of the coefficient image by the secret image. (5) Until the secret data is completely hidden in the cover image, step 4 will be continued. (6) Inverse IWT is applied. (7) Stego-image is obtained.

Image Quality Enhancement.
e obtained stego-image from the embedding process is of insufficient quality. As a result, a processing procedure on the unique intelligent system is required. is phase is necessary for reducing the chances of numerical identification and other types of image modification attempts.

Hybrid Fuzzy Neural
Network. An HFNN with a backpropagation learning method is employed to improve the image quality in this study. In general, neural networks resemble HFNNs that are inhomogeneous. e neural network refers to a framework that can simulate how the human brain learns. e stego-image transformed to binary bit values in order to identify free bits and bits that encompass secret bits. A buffer is built to keep the free bits that are not used by the embedding process. e stego-and cover images are also used to extract statistical and perceptual attributes. e statistical and visual aspects are represented by the chi-square probability and the Euclidian norm. e HFNN is provided with the free bits buffer as well as the two characteristics of the stego-image. e cover attributes are then compared to the HFNN outputs. By integrating the changed free bits with the secret bits, a new stego-image is created if the outputs match the characteristics of the cover image.
e HFNN weights are modified using a backpropagation learning process.
(1) e stego-image is generated by hiding the secret image, which is implemented using embedding approach algorithms. (2) Features are extracted from the stego-image using the feature extraction technique.  (3) A buffer is generated which is not utilized in the steganographic algorithm. e secret is not hidden in the buffer bit, and it is called free bits. (4) e statistical and visual measure of the stego-image is measured using a fuzzy neural network with backpropagation. e statistical and visual measure of the stego-image is measured using a fuzzy neural network with backpropagation. erefore, for an updated stego-image, the free bit buffer, statistical and visual measures are contained in the output layer. (5) e output of the fuzzy network is compared with the cover image. e stego-image is formed if the output and cover image get matched and the output with the free bit is used by assembling the other bit in which the secret image is hidden. Otherwise, step 4 is repeated.
e input layer, rule layer, fuzzification layer, inference layer, and defuzzification layer are the five layers that compose HFNN as shown in Figure 5. e input neurons of the HFNN were trained using five layers of backpropagation, using inputs from the free bits buffer, and numerical and graphic characteristics. Following that, all hidden neurons in the fuzzification layer get the inputs. e membership function is used to perform fuzzy process on input characteristics at this layer. For the excellent approximation of input space, the Gaussian membership function is utilized. By altering the parameter values, this bell-shaped function produces several membership functions for the input characteristics. e strength of fuzzy rules is determined in the rule layer using the logical AND operator. On fuzzy inference, the inference layer executes OR operations. e HFNN output will emerge from the defuzzification layer. e weight parameter is used to link the nodes of all layers in HFNN. e dimensionality reduction is a feature extraction in which set of features is transformed form stego image. e statistical features are obtained from chi-square probability and visual features from the Euclidean norm. Let x and y be the input, then linguistic input variable A 1 , A 1 and B 1 , B 1 . e linguist state is given as u j1 , u j1 , u j1 , u j1 are the parameter set. e logical operator is used to strength the output of the network layer: e defuzzification is obtained as the normalization: e error function is given by e desired output is represented as y.

Extraction
Process. e extraction process is used to extract the hidden image from the embedded process in adjustable order. e cover image used in the first step is not used to extract the secret image. e data is provided by the LSB, and the procedure was enhanced using the secret key.

Algorithm for Extraction Process in IWT-LSB
(1) Stego-image is read.
(3) IWT is applied for stego-image. (4) e secret data is extracted for the approximate image coefficient of stego-image. (5) Until the secret data is extracted, step 4 will be continued. (6) Inverse IWT is applied. (7) Image gets extracted. (8) Extract the secret data.

IWT-LSB Algorithm for Grayscale Image.
e proposed hybrid IWT-LSB technique for grayscale images involves an embedding phase and an extraction phase as shown in Figure 6. In the embedding phase, the grayscale cover image is transformed using IWT, then the secret text is embedded in the LSBs of the cover image's coefficients, and finally the inverse IWT is used to construct the stego-image. Without knowing anything about the original image, the hidden text might be extracted throughout the extraction process. e hidden secret text is retrieved from the LSBs of the filtered stego-image's coefficients, and the inverse IWT is useful for creating the extracted image. e 512 × 512 bitmap grayscale images are utilized as cover images for the hybrid IWT-LSB algorithms on grayscale images. e size of the estimated coefficient images after performing the IWT is 256 × 256, which implies that a secret text with up to 8 and 192 digits may be secured. To improve the system's robustness and secure the message from external impacts such as noise, compression, and filtering, the secret data is placed in the LSBs of the estimated coefficient images of the cover images.

IWT-LSB Algorithm for Color Image.
e color image is split into R, G, and B components in the proposed methodology, and the three components are employed to hide data as shown in Figure 7. In the embedding phase, the RBG component is used to decompose the color of cover image in which the IWT transform is used with the signature of the user and the secret data is embedded with actual Journal of Healthcare Engineering length in LSB and coefficient image is approximated for the red component of cover image. e image's approximated green and blue components coefficient are used to secure the secret data. en, the inverse IWT is applied to each component once the embedding process is complete, and then these components are recombined to generate a stego-image. e hidden data may be extracted during the extraction process without knowing anything about the original image. e R, G, and B components are decomposed from the noisy stego-image in which median filter is utilized for filtering and then transformed using IWT. e LSB of the approximated image coefficient is used for extracting the actual length of the secret image by utilizing the G and B components, and the inverse IWT is utilized to extract the original image.  Figure 6: Proposed embedding and extraction process for grayscale image steganography.  Parameters such as MSE, NCC, and PSNR are used for the performance evaluation. e error among the cover image and stego-image is examined using MES and PSNR and compared with the existing techniques. e cover image is represented as I C , and stego-image is represented as I S . e mean square error is calculated using the following equation:

Result and Discussion
e peak to signal noise ratio is given as e number of rows and columns is represented as n and m, and I max is the maximum hold of the original image. e similarity and dynamic extent of the cover image and secret image are quantified using the mean square error. e mean square error in the total number of pixels in color and grayscale image is given in Figure 8. e PSNR of the suggested algorithm compared with the existing technique is given in Figure 9. e PSNR regulates the difference in the dynamic range of invisibility in cover image and secret image in which the value is greater than   Journal of Healthcare Engineering 9 53 dB in grayscale and color image compared to the existing techniques. e degree of closeness between the cover image and stego-image is obtained using NCC which is shown in Figure 10. e degree of closeness is obtained after embedding the data in the secret image.

Conclusion
is paper used image steganography to securely and imperceptibly embed the patient's personal information in their medical images to enhance confidentiality in case of distant diagnosis. e least significant bit of the approximate coefficient of integer wavelet transform is proposed. is technique is analyzed for grayscale image and color image. IWT is utilized to hide the secret image in LSB in the grayscale image, while IWT with R, B, and G component is used for hiding the secret image in color image. e distortion between the cover image and stego-image is obtained by measuring the mean square error and PSNR, and the degree of closeness between the cover image and stego-image is estimated by utilizing the normalized cross-correlation. e result shows that the IWT-LSB technique can hide secret data with large length with better MSE, PSNR, and NCC.

Data Availability
e data used to support the findings of this study are included within the article.

Conflicts of Interest
e author declares that she has no conflicts of interest to report regarding the present study.