Improved authentication and computation of medical data transmission in the secure IoT using hyperelliptic curve cryptography

Data transmission is a great challenge in any network environment. However, medical data collected from IoT devices need to be transmitted at high speed to ensure that the transmitted data are secure. This paper focuses on the security, speed and load of transmission. To prove security, combined steganographic methods involving cryptographic algorithms are used. The proposed model begins by updating two entries, medical image data and medical report data. Digital imaging and communications in medicine image data hold the medical report data to be encrypted and transmitted over the network channel. Although the proposed work follows the conventional method of data transmission from encryption until transmission, an effort has been made to split up the given data without transmitting them as such. As a public cryptography mechanism, the algorithm is also capable of transmission during decryption. The method of this article is genuine in proving its secure actions during the transmission of medical data and medical images. The proposed method justifies its performance when tested in hiding medical transcription data of different sizes varying across 30, 45, 64, 128 and 256 bytes in sample images with an average PSNR ranging from 55 to 70 dB, an MAE averaging from 0.2 to 0.7, and an SSIM, SC and correlation coefficient averaging to 1. This research is proven to work well in a simulation environment, and the results prove the genuine nature of the proposed technique.


Introduction
The components of a conventional cryptographic system consist of plain text as input, cipher text as output and keys.The encryption process creates the cipher text, and the decryption process takes place on the receiver side to extract the plain text.Many cryptographic algorithms exist to secure data with encryption so that attacks are clearly handled.In addition to the security of transmission, the data load is to be considered so that the entire data will be transmitted in a compressed fashion occupying very little space.The third consideration of data transmission is reliability which is proven by various factors, such as scaling and cropping.When data transmission occurs, a huge number of mathematical calculations are required to protect the data on its journey.
The survey performed here shows related works that have been carried out in various fields relevant to this research article.It proceeds with a survey of data encryption and information hiding and continues with region of interest extraction to hide textual medical data in a medical image without distorting the original medical image.To provide security to medical data, cryptography plays an important role and hyperelliptic curve cryptography (HECC) is involved to utilize this process in the proposed work.Hence, to provide a mathematical base for hyperelliptic curve systems, some research articles related to HECC are also studied.Additionally, it becomes important to implement a secure data transmission system in handling healthcare data retrieved from IoT systems [1].
The research field of security mentions information hiding and information encryption as the major concepts.This research focuses on combining the core concepts of providing data security.Hybrid steganography exists this field of research to provide more secure transmission of medical data.The state-of-the-art includes the work that Jain et al. [2] proposed.The flow of the survey of existing research then moves on from choosing the best combination of encryption algorithms, to choosing the region of interest to hide data.Then, it proceeds with the major part of data transmission through hyperelliptic curve cryptography and finally to data compression.The initial and most commonly used cryptographic algorithms are taken into consideration.Preetha et al. [3], from a comparative 1 3 Improved authentication and computation of medical data… study of RSA and enhanced RSA based on the execution time, introduced RSA as an algorithm based on research through data from far in the past.Karthick et al. [4] discussed the performance excellence of triple DES, and Akash et al. [5] conducted a performance evaluation of DES and AES.
In information hiding, the basic concern is to hide the data so that the cover data are not disturbed.Interest is concentrated in a region of cover images and data involving medical information.Renuka et al. [6] proposed an algorithm to extract the region of interest (ROI) automatically using statistical moments to detect the region of interest from noisy medical images.Mousavi et al. [7] proposed a method to detect the region of interest (ROI) involving a series of stages, such as morphological image reconstruction, Gaussian low-pass filtering, thresholding and boundary detection.Additionally, the robustness of the proposed algorithm was shown.
Kazeminia et al. [8] proposed a histogram-based dispersion method, in which the region of interest is separated from the original medical image based on the effective compression ratio of statistical and spatial redundancies.Ye et al. [9] analyzed two methods for edge detection a combined threshold-based Sobel edge detector a fuzzy c-means region growing method for the purpose of ROI extraction from medical images and proved by analysis that both the methods surpass the performance of traditional edge detection methods.Shin et al. [10] also analyzed some edge detectors using object recognition tasks.
Hyperelliptic curve cryptography plays a major role in securing medical data to be embedded into medical images.Related to the mathematical foundations of hyperelliptic curve cryptography, Jan Pelzl et al. [11] provided a mathematical description of converting the elliptic curves that are familiar to computer scientists to hyperelliptic curves.Additionally, in his work [12] he expanded the explicit formula for genus 4, a major mathematical concept of hyperelliptic curves.Henri et al. [13], by his mathematical method of derivation provided sufficient growth of hyperelliptic curves from elliptic curves that were considered preliminary.To bridge the gap between mathematics and computer science, Padma et al. [14] encoded and decoded a message using the Koblitz method implemented in elliptic curve cryptography.

Proposed model
This paper proposes a secure transmission model that may be applicable in an IoTbased healthcare system in which clinical data are secured in medical image.The proposed model, as shown in Fig. 1, involves the following consecutive steps: AES and Blowfish hybrid cryptography to encrypt the medical data, the Koblitz method to choose the embedding points, HECC to embed the encrypted clinical information into a medical image in the region of interest extracted by the Sobel edge detection method and then applying 5-level discrete wavelet transform (5-DWT) to compress the embedded image.Hence, the major optimization parameter of transmission, maximizing payload and security and minimizing storage space is attained.
The major techniques of encryption, Koblitz encoding, HECC (genus 4) and 5-level DWT will be discussed in what follows to give a brief idea of the concepts.

Medical data encryption
Medical data and the related medical image when transmitted over a transmission channel need to be secure enough to preserve the privacy of the patient.Hence, medical data that are given as an option on medical images such as X-rays, CT scans, MRI, are embedded into the medical images to make them more secure.The medical data to be embedded into the medical image are initially processed with two different encryption techniques.AES and Blowfish were selected.This combination was chosen since there are already works that prove that these two algorithms can withstand guessing attacks and have more beneficial results than other existing algorithms.
The results for some sample data that were fed into existing programming codes in MATLAB are shown here.To prove which the best algorithm to use is, the parameters discussed by Wahid et al. [15] are considered: encryption duration, decryption duration, memory utilization, avalanche effect, and average entropy.On interpreting the values of encryption duration and decryption duration, memory utilization measure should be found to be quite low accompanied by high avalanche effect and Improved authentication and computation of medical data… average entropy.The results for the option of the AES and Blowfish algorithms are shown in Tables 1 and 2.
The cryptographic scheme used here includes the procedures of encryption and decryption.During encryption, all medical data are converted to binary values and are divided into odd and even positions-M even and M odd .M even is encrypted by Blowfish, and M odd is encrypted by AES.
Then, the XOR operation is performed on the encrypted bits to obtain the binary matrix.The difference in segment encryption is obtained as the encryption of the even and odd bits, and this occurs with different public keys for M even and M odd encryptions.The private keys are the key generated from the biometric identity of a person [16][17][18], who is probably a clinical physician or a medical researcher trying to fetch the data stored in the medical image.This private key (P r ) identifies the person using the data and stored as a log information.The public key (P u S, P u R) is taken as the patient ID.
The results of the analysis based on traditional cryptographic algorithms and their performance analysis are depicted in Tables 1 and 2. Since the maximum entropy and avalanche effect are found with AES and Blowfish with respect to the data used in this work, these two algorithms are preferred for further proceedings.
As a result of executing the hybrid encryption algorithm, the cipher text of the medical data is obtained and is ready to be transmitted.

Region of non-interest extraction
Even though the points of a curve are described by the Koblitz method, these curves are to be placed in a location on a medical image such that it does not affect the medical image.This becomes an important criterion to be met when dealing with medical image watermarking or medical image data embedding.Since in the proposed system, the medical data in text format received as feedback from the medical supervisor are to be embedded into a medical image and stored, it becomes quite important that the embedding process should take place in the areas where the medical data are null, so that on embedding or extracting the medical data, the medical details in the medical image are not spoiled or disturbed.There are many edge detection operators that are used in MATLAB for ROI extraction.By working with sample images, it is possible to determine the best operator to use.The polymorphic edge detection operators considered for comparison are Prewitt, Sobel, Canny, Robert's, zero crossing and LoG.The images considered in the medical scenario are grayscale images, and the same are used for comparison.The optimal performance result justifies Sobel operator for use in further experimentation as shown in Tables 3, 4.

Koblitz method of encoding and HECC
Now, the data seem to be secure, but during transmission, more attacks are also possible, which should be resisted.The Koblitz method of encoding helps to convert the cipher text CT md into points on a curve, which prevents the attacker from guessing the data.Hyperelliptic curves are used to carry the points that contain the CT md .These hyperelliptic curves provide a secure mechanism against attacks, whereby it is difficult to guess the positions of points containing the ciphertext medical data.Additionally, these hyperelliptic curves are capable of encrypting and decrypting the points on the curve but not the embedded data.
Until this study, there has been no proven research on polynomial-time algorithms to find a large number of points on an arbitrary curve.Since the aim is to find points on the curve that are related somehow to the plain text CT md and not just random points on the curve, probabilistic algorithm should be used to reduce the chance of failure.Hence, encoding and decoding become important feature steps in encryption and decryption, which directs in the choice of the Koblitz method of encoding [19].

Improved authentication and computation of medical data…
Assuming that the medical data or the records are to be stored in a centralized database for ease of access, these data will be monitored and administered by the central medical database authority (CMDA).The primary responsibility of the CMDA is to receive, store, authenticate the requester, and distribute data for purposes of analyzing of patient or central access.
These are the curve points that have encoded messages and a position on the curve to embed them as a hyperelliptic curve.Elliptic curves form the base for hyperelliptic curves as special elliptic curves with genus ≥ 2. However, the level of security provided by HECC is the same as that of ECC for a smaller key size than that of ECC [20].The arithmetic procedures of the hyperelliptic curve cryptosystem include addition and doubling in the Jacobian of the curve in the group law.The algorithm for the group operation given by Cantor [21] involved many improvements in the computation of group operations.Recently, many attempts [22] have been made and succeeded in the development of an efficient group operation for HECC.For example, the existing Cantor algorithm was represented explicitly as Harley's algorithm and this approach succeeded obtaining subexpressions and explicit formulas for hyperelliptic curves of genus 2 and was extended by Lange and others.
The works of Lange [23], Matsuo et al. [24], Miyamoto et al. [25], and Takahashi [26] contributed to developing more efficient algorithms for performing group operations.Group operations, in the form used in hyperelliptic curves, underperform with respect to speed, when compared to elliptic curves.
Arithmetic operations on polynomial rings were performed in the Cantor algorithm, whereas general arithmetic operations in the integer ring were carried out by explicit methods.Nagao et al. [27] showed the steps needed to perform group operations on the Jacobian divisor class group of hyperelliptic curves.Intersecting points of the Jacobian variety curve and hyperelliptic curves [28] were made into a group.
Additionally, there exists a myth that HECC involving curves with higher genera creates complexity when compared to cryptosystems involving lower genus elliptic curves.Pelzl et al. [29,30] proved that genus-4 hyperelliptic curves (HECC) can outperform genus-2 HECC for low cost security applications and that a performance measure such as genus-3 HECC could be used.The security of genus-4 HECC was also proven.Conventional hyperelliptic curves use the Cantor algorithm and subexpression algorithm to perform basic calculations.
The explicit formulas take the coefficient of the Mumford representation [31] of Cartesian points as input.Since the Koblitz encoding method creates points for curves, HECC is used to fetch the points and not data, and it carries data in the form of points encoded over curves.

Image compression
Based on the HECC curve points, Koblitz encoding is performed to embed the secret data into the medical image.The embedded image undergoes a 5-level discrete wavelet transform to reduce the storage size of the embedded image.Since the discrete wavelet transform can be added as a means of image compression Huffman lossless coding is used.The higher frequency regions contain information on the edge components, whereas the lower frequency regions are divided again into higher and lower frequency regions.The decomposition of these bands is specified in the algorithm that takes the encoded image to the 5-level decomposition.The adopted combined frequency domain transform works with 5-level compression of the discrete wavelet transform implemented on the face image as shown in Fig. 2. The watermarking scheme employed is an optimization approach to provide high security and reduced storage space.Security is ensured in the proposed system using image compression.To enhance security, 5-level DWT is used.The performance measures include PSNR and MAE.The watermark embedding and extraction procedure exhibits much better results, which are proven by better PSNR and MAE values as shown in Table 4.The proposed method has proven to withstand attacks with an appreciable level of improvement.A watermarking scheme with high resistance to attacks was developed and implemented.

Improved authentication and computation of medical data…
As a result of executing the hybrid decryption algorithm, the cipher text of the medical data is converted into plain text and the medical image is retrieved for use.

Experimental results and evaluation
The implementation of the proposed model was carried out using MATLAB R2010, which runs on a personal computer with a 2.27 GHz Intel Core™ 13 CPU, 8 GB RAM and the Windows 10 operating system.
The data used throughout the experiment are MRI images and medical prescriptions obtained from MRI scanning centers and doctors, which account for 95 images that were personally collected on request.
Additionally, the medical transcription does not belong to the corresponding MRI images, and the identities of the patients are not known and were not revealed by the providers.
The images used were not the exact images obtained, which were fabricated in Photoshop to create original images as shown in Fig. 3.The performance of Fig. 3 Sample images the proposed system was proven by the statistical metrics.The five metrics considered for proving the quality of the proposed system were PSNR, MAE, SSIM, SC, and correlation.A quantitative index, the peek signal-to-noise ratio (PSNR), was employed to determine the imperceptibility of the watermarked image.The PSNR value is directly proportional to the quality of the steganographic image and the imperceptibility of the hidden message.A larger PSNR value indicates that the watermarked image resembles the original image more closely, meaning that the watermarking method makes the watermark more imperceptible.Generally, if the PSNR value is greater than 35 dB, the watermarked image is within acceptable degradation levels, i.e., the watermark is almost invisible to the human visual system.
The mean average error (MAE) calculates the magnitude of the average error between the original and the steganographic images.A lower MAE reveals that  Improved authentication and computation of medical data… the extracted watermark resembles the original, where the extracted watermarked image is further degraded by attacks.
If a method has a lower MAE, then it is considered to be more robust.The structural similarity index measurement (SSIM) is used to measure the structural similarity between two images, which ranges between −1 and + 1, showing whether the images are identical.Correlation was used calculate the similarity in signals producing a maximum value of similarity.Structural content (SC) is a correlation based measure that also shows similarity in the content of the images.
Watermarking is also one of the major contributions of the proposed method.The efficiency of any watermarked image is found by its PSNR value.The PSNR values obtained at various approximation levels are shown in Fig. 4. The embedded results with a text size of 128 bytes are listed in Fig. 5.The PSNR values of different approximation levels are depicted in Fig. 6.The implementation results as shown proves the imperceptibility of the watermarked image and are acceptable.
The proposed method is tested against several strategies: an enhancement attack-Gaussian filtering, noise addition-salt and pepper, scaling-50 and 75%, cropping, and compression attack.These are shown in Table 5.
In addition, watermarks can be extracted from the other image processing attack with lower MAE values, and the proposed method is more robust than the previous methods.The implementation results show that the imperceptibility of the watermarked image is acceptable.The presented method was tested by most of the common image processing attacks, such as Gaussian filtering of different sizes as an enhancement attack, adding salt and paper noise, scaling with two common factors: 50 and 75%, cropping, and compression attack.Specifically, in the case of adding noise and enhancement attack, proposed method shows a significant improvement in robustness.In as much as, the watermarks can be extracted from the other image processing attack with lower MAE values, proposed method proves to be robust as shown in Table 5 The texts embedded were sample texts of varying sizes ranging from 30 to 254 bytes.Table 5 shows the mean absolute error for the images when  an attack was performed, proving the robustness of the method for varying text sizes when tested on 8 different images.The results were not obtained from the automated programming strategy; rather, they were given as input manually, and the values were obtained.

Conclusions
The proposed system involves medical data that are encoded and embedded into a medical image, which is compressed and transmitted over a secure channel of cryptographic transmission.The structural similarity index measurement (SSIM) and structural content (SC) are approximately 1, which means that the image before the medical data are encoded does not change after retrieving the medical data, indicating that the transmitted medical image is not distorted.Additionally, the PSNR values indicate the high quality of the compressed and reconstructed images.For secure and comfortable transmission, cryptographic methods and algorithms are used that provide better performance when tested as individual media.Not all stateof-the-art methods have been compared and this method can be tested on real-time data transmission, which is specifically applicable in IoT systems.

Fig. 1
Fig. 1 Proposed medical data transmission method

Fig. 4
Fig. 4 PSNR of watermarked images for different approximation factors of Fig. 3a

Fig. 6
Fig. 6 Embedded images tested with a text size of 128 bytes

Table 4
Statistical parameters of the steganographic and original images for varying text size

Table 5
Experimental results of attacked images