Hybrid image encryption using advanced least significant bit algorithm,chaotic maps and DNA encoding for digital healthcare

Abstract

Security and privacy are major concerns in this modern world. Medical documentation of patient data needs to be transmitted between hospitals for medical experts opinions on critical cases which may cause threats to the data. Nowadays most of the hospitals use electronic methods to store and transmit data with basic security measures, but these methods are still vulnerable. There is no perfect solution that solves the security problems in any industry, especially healthcare. So, to cope with the arising need to increase the security of the data from being manipulated the proposed method uses a hybrid image encryption technique to hide the data in an image so it becomes difficult to sense the presence of data in the image while transmission. It combines Least Significant Bit (LSB) Algorithm using Arithmetic Division Operation along with Canny edge detection to embed the patient data in medical images. The image is subsequently encrypted using keys of six different chaotic maps sequentially to increase the integrity and robustness of the system. Finally, an encrypted image is converted into DNA sequence using DNA encoding rule to improve reliability. The experimentation is done on the Chest XRay image, Knee Magnetic Resonance Imaging (MRI) image, Neck MRI image, Lungs Computed Tomography (CT) Scan image datasets and patient medical data with 500 characters, 1000 characters and 1500 characters. And, it is evaluated based on time coefficient of encryption and decryption, histogram, entropy, similarity score (Mean Square Error), quality score (peak signal-to-noise ratio), motion activity index (number of changing pixel rate), unified average changing intensity, image similarity score (structure similarity index measurement) between original and encrypted images. Also, the proposed technique is compared with other recent state of arts methods for 500 characters embedding and performed better than those techniques. The proposed method is more stable and embeds comparatively more data than other recent works with lower Mean Square Error value of 4748.12 which is the main factor used to determine how well the data is hidden and cannot be interpreted easily. Also, it achieved a Peak Signal-Noise Ratio (PSNR) value of 71.34 dB, which is superior than other recent works, verifying that the image quality remains uncompromising even after being encrypted.

Keywords

Hybrid image encryption least significant bit algorithm arithmetic division operation canny edge detection algorithm chaotic maps DNA encoding

1 Introduction

Throughout the centuries, the healthcare industry has undergone significant changes and advancements, particularly in the area of digital technology. To encourage the adoption of new technology and ensure the maintenance of accurate medical records, guidelines and data standards have been developed. Electronic Health Records (EHRs) [22] have replaced paper medical reports, providing healthcare professionals with a digital collection of a patient’s medical history that can be easily accessed and shared across various networks. The secure storage and communication of patient data is crucial in enabling healthcare providers to make informed decisions that can save lives. It is therefore essential to recognize the vital role that accurate and up-to-date patient records play in determining the best course of action for patients. The crucial part of healthcare sector is to ensure data security, but challenging as the nature of the data is sensitive and vulnerable. The safeguarding of data involves implementing measures, technologies, and algorithms to fortify the security system and prevent theft, unauthorized access, and misuse.

Cryptography is a commonly used technique in the healthcare industry [14] to ensure privacy and secure communication. It primarily concerns authentication, non-repudiation, secrecy, and integrity. It employs a key to encrypt the data using several cryptographic techniques, rendering it unreadable. Based on the key, the cryptographic techniques may be symmetric and asymmetric. In symmetric encryption, a single key is employed for both encryption and decryption. Conversely, in asymmetric encryption, two distinct keys are used - a public key for encryption and a private key for decryption. However, one of the primary challenges with cryptographic methods is that encrypted messages tend to look suspicious [16] and attract the attention of attackers who may attempt to detect hidden data and illegally decrypt the message. To address this problem, a data hiding technique called steganography has been introduced. Steganography is a covert communication technique that involves hiding the existence of data by embedding hidden messages in any media. There are different steganography techniques [26] based on the way of embedding data within a cover image. An effective algorithm is capable of embedding a significant volume of data into a cover image without causing any distortion to its quality. Preprocessing techniques, such as compression algorithms, can be employed to the input data to facilitate the embedding of larger volumes of data without compromising the quality of the cover image. However, once the concealed data is detected, it can be decrypted with ease.

With the development of new algorithms, cyber-attacks techniques are also evolving, making it crucial to create a robust encryption scheme that is difficult to break. Chaotic maps, which are discrete maps, [39] usually take the form of iterated functions. Being called ”~chaos”™, it exhibits some kind of chaotic behavior which is used to generate keys for the image encryption. The chaotic nature of these maps makes it sensitive to initial condition which makes it is ideal for use as a key and is highly recommended by researchers due to its security, speed, efficiency, complexity, and low computational overhead. Another approach for ensuring end-to-end security in healthcare is DNA cryptography, which uses the biological structure of DNA to store and transmit the data which conceals the hidden data and becomes unnoticed as it relates the healthcare terminologies. The concept is to use the four-nitrogen base of DNA as code and the data is encoded into DNA sequence prior to transmission.

To maintain message confidentiality, it is crucial to not only develop robust cryptographic methods, but also to securely exchange keys for encryption purposes and to authenticate users. The use of cryptographic keys plays a vital role in safeguarding sensitive information, mitigating the risk of data breaches, and complying with legislation. However, the loss or theft of these keys can have serious consequences, including the loss of data and compromised system security. As a result, it is imperative for any security-focused organization to establish a comprehensive key management protocol. There are many Key Distribution Centers (KDCs) to manage keys efficiently and securely.

The proposed work has several primary contributions, which include:

The use of a six chaotic maps combination and LSB technique to strengthen the resilience of the process.

To maintain the quality of cover image, the LSB technique with Arithmetic Division is incorporated that divides the data into two parts: quotient and remainder, and then embedded in the LSB and image edges respectively. Also, it enables us to store a vast amount of data without compromising the quality of the cover image.

DNA sequence is generated using DNA encoding to transmit the embedded image securely.

Most of the recent works either provided data embedding feature using LSB with least number of chaotic maps or provided image security using DNA encoding. Also, they supported only limited number of characters embedding support into images. Whereas, the proposed technique combines LSB with arithmetic division, six chaotic maps and DNA encoding to provide larger number of characters embedding support and security during the transmission.

The remaining sections of this paper is organized as follows: Section II covers related works which includes steganography and DNA cryptography techniques to securely hide the information in an image. Section III covers proposed methodology which includes tailored hybrid image encryption technique to hide the patients health records into a medical image. Section IV analyses the experiments done on the proposed image encryption with 500, 1000, 1500 characters of patients health records. Section V evaluates the performance of the proposed model in comparison with other existing models over different parameters like encryption time, decryption time, entropy, Mean Square Error (MSE), Peak Signal-Noise Ration (PSNR). Finally, Section VI concludes the work with advantages, limitations and possible future directions.

2 Related Works

Steganography involves hiding a message within some form of media, such as images, audio, or video files. Different types of steganographic techniques have been developed over the years to ensure secure transmission of data [1, 4, 7, 25, 27, 29, 36]. Also, variety of image encryption techniques were proposed by lot of researchers [3, 6, 9, 11, 21, 30, 31]. Osunade et al. [27] proposed a method where Least Significant Bit (LSB) algorithm is used to conceal data in an image using only two colors (GREEN and BLUE), while minimizing any noticeable alterations to the image’s color scheme. But here only 8 bytes to 1024 bytes of data can be hidden. Mohamed et al. [25] suggested that, dividing the image into two segments can help to increase the capacity for embedding data, as well as improve the quality of the resulting stego image. This is achieved by using LSB substitution to modify certain bits in one of the image segments, while the other segment is used to determine which pixels have undergone changes. Omar et al. [7] presented a technique in which the k-least significant bits were utilized to conceal the secret image into the cover image. In the decryption process, a local entropy filter was utilized to detect and extract the concealed image from cover image, and an image quality enhancement technique was employed to improve the quality of the extracted image. Yigit and karabatak [36] suggested a modification in LSB approach, in which the final two bits were considered to hide the data to increase the amount of data that could be hidden. Bitmap Image File (BMP) files were used since they give an uncompressed 24-bit color picture file, so that there is no change in image size while the data is hidden. Abbas Darbani et al. [4] found that steganography in lossless BMP format was more challenging than lossy Joint Photographic Experts Group (JPEG) format. To mitigate the impact on the image quality problem, they proposed a method of steganography where the least significant bits in the discrete matrix were used for hiding data in the JPEG images. Jagan Raj et al. [15] devised a steganographic method that involves two stages and has a smaller footprint in the cover image. This approach produces better quality stego images than other LSB techniques currently in use. The embedding phase uses optimized secret messages, allowing for high-capacity embedding rates while improving the quality of the cover image.

For additional security purposes, methods like DNA cryptography, logistic and Chaotic maps are applied on data. Hraoui et al. [12] conducted a comparison between a conventional encryption method based on Advanced Encryption Standard (AES) and a logistic map. They concluded that the AES-based technique demonstrated superior security performance compared to the logistic map, that exhibited some weak periodic windows. However, the logistic map was found to be advantageous in terms of its low computational cost and ease of implementation, making it a feasible option for real-time communication systems that require image encryption. Nilesh Y. Choudhary et al. [35] uses a 2D Arnold map to encrypt a gray image by dividing into blocks. The Arnold map is used to permute images, and the permuted blocks are merged to produce partially encrypted images. This method is efficient because Arnold maps have the property of producing the same image after a certain number of iterations. Zhang et al. [38] suggested a method by exchanging rows and columns based on a logistic chaotic map, converting each pixel into four nucleotides, and then transforming them into base pairs using Chebyshev’s chaotic map. Each nucleotide is converted to its respective complementary base pairs and the resulting two-dimensional matrix is converted into an encrypted image. Babu et al. [21] presented a novel multiple-image encryption frame format based on a secure force algorithm, Arnold Cat map RSA algorithm, chaotic permutation techniques and DNA sequence encoding. Each segmented element of an image is encrypted using all four methods, which provides protection against various statistical and differential attacks. DNA sequence encoding is used to ensure secure and efficient figure administration. Additionally, the Arnold Cat Map is used to perform a transformation on randomly organized pixels in the image. To generate the final encrypted image, a random key is produced by merging all four blocks of the encrypted image, and then an XOR operation is performed using generated key and the combined encrypted image. Lidong Liu et al. [19] proposed a decryption approach that eliminates the need for transmitting large secret keys and synchronizing with plaintext images, thus preserving the image characteristics. However, this method is susceptible to noise and attacks. Sakshi Patel et al. [28] developed a modified version of the chaotic map that utilizes a 32-bit ASCII private key to generate a confusion matrix encoded with DNA. To enhance the security and privacy of the encryption, multiple chaotic maps [37] can be combined to create more robust and resilient chaotic maps with variable keys. The sine square logistic map was used to generate keys, which resulted in a highly secure encrypted image with minimal computational complexity. To evaluate the effectiveness of native chaotic and hybrid chaotic approaches for image encryption, Chaudhary et al. [2] used various parameters such as Peak Signal-Noise Ratio (PSNR), Number of changing pixel rate (NPCR), Unified average changing intensity (UACI), histogram and computation time. However, to address the limitations of these approaches, they proposed a new method that uses six different chaotic maps, namely Chebyshev map, Gaussian map, Henon map, Logistic map, Tent map, and Piecewise map. By XOR-ing these six different keys, the attributes of the image were not altered.

3 Proposed work

Over the past few years, the growth of cryptographic technology is tremendous, and experts have developed many algorithms and techniques to improve the security of information systems and information in transmission. By the phrase ”~no one solution fits all”™, individual cryptographic algorithms may have vulnerabilities when handling sensitive data like Electronic Health Records (EHRs). Input EHRs and associated medical images undergo the following processes, to convert them into a format that can be safely and quickly transferred between hospitals in the event of an emergency, and are transferred between different hospitals primarily for the purpose of interprofessional medical opinion. This work proposes an efficient hybrid encryption scheme to store and transmit patient’s EHR using LSB, chaotic maps and DNA encoding. The work comprises of three main modules: Embedding medical data into a medical image, image encryption using chaotic maps and DNA encoding. The dataset section provides the detail about medical images used and the input patient data. Figure 1. shows the overall workflow of the proposed work.

Fig. 1

Overall workflow of the proposed work.

3.1 Datasets

In this work, the required data are a medical cover image and an input text file containing the patient’s details. For the medical cover image, a high-resolution image with different dimensions are taken from 3 different medical image datasets. The datasets utilized in this work are CT Scan images of Lungs [23, 24], ChestX-ray14 [32] and Digital Imaging and Communications in Medicine (DICOM) Library [5]. The patient’s information is considered as sensitive data that needs to be protected from unauthorized access. The text file contains input data which consists of different files having different capacities around 500, 1000 and 1500 characters which includes the patient’s name, age, gender, date of birth, file Number, physician, exam, date of visit, clinical information, comparison, contrast, techniques, findings, impression, next visiting date.

3.2 Embedding medical data into a medical image

The process of maintaining and sharing patient medical records in registers as paperwork or electronic devices can add a lot of extra storage space and also make it difficult to retrieve the necessary data for a particular patient. Thus, the patient’s EHR is embedded in the patient’s relevant medical images for easy storage and sharing using image steganography. Image steganography aims to hide sensitive or confidential data in cover images to generate stego images. This technique was introduced to secretly exchange information and avoid the attention of intruders. Images are only vulnerable if an intruder is aware of the data’s existence, whereas image steganography completely hides the data’s existence [10]. The LSB algorithm is a well-known steganographic approach that conceals confidential data in the LSBs of a cover image. Nevertheless, this technique has limitations concerning the quantity of data that can be concealed and the image quality that can be maintained. Also embedding the data directly into the LSB can also be risky as it causes the attacker to identify the hidden data and if identified, it is easy to retrieve the data and misuse it for personal gain or to defame an organization.

Steganographic approaches can be categorized into temporal and spatial domains to overcome these limitations. Different methods can be used to embed messages in medical images. Temporal domain employs a method in which the image can be converted into frequency components using techniques like Fast Fourier Transform, Discrete Cosine Transform, or Discrete Wavelet Transform, and messages can be hidden in some or all of the transformed coefficients. Alternatively, the spatial domain method employed to insert the message bits into the LSB locations of coverage intensity pixels. This LSB method can embed a larger amount of data, but it is vulnerable to statistical attacks. Frequency domain techniques provide better resistance to noise and attacks. This proposed method employs innovative spatial domain techniques to increase the data capacity that can be hidden in medical images while maintaining high quality, ensuring the utmost security for patient’s medical images and EHR data. The proposed method first uses the Canny edge detection algorithm [8] to detect edges and non-edges in the related medical images, then divide the patient data into two parts (quotient and remainder) in parallel. Finally, the divided data is embedded in the coordinates of edges and non-edges respectively. The final representation of the original image and the stego-image after performing the above method is shown in Fig. 2.

Fig. 2

Image after embedding data.

3.2.1 Canny edge detection

Canny edge detection is a multistage algorithmic technique [34] that provides structural information about an image by detecting various edges in the image. Common criteria for an accurate edge detection algorithm include a low error rate in edge detection, the edge point should be localized on the center and edges marked only once in the image, and preventing image noise from generating false edges. The Canny edge detector performs many preprocessing steps to improve the accuracy. Images, regardless of their format, are converted to grayscale before processing begins. The image is first smoothed with a gaussian filter to reduce the noise present in the image, then the amplitude and direction of the gradients are maximally suppressed, and finally strong and weak edges are detected and connected using double thresholding. The algorithm 1 demonstrates the working of canny edge detector. Apart from image, it is used here to retrieve the coordinates of the edges and non-edges of the input medical image to perform an efficient LSB technique. The resultant coordinates are used to find the coordinates of the edges (strong) and non-edges (weak) by setting the threshold value.

Algorithm 1 Canny Edge Detection Algorithm

Require: MedicalImage M

1: M_g ← cvtColor (M) //Convert the medical image to grayscale image

2: M_g ← GaussianBlur (M_g) //Noise Reduction

3: gx, gy ← Sobel (M_g) //Calculate the gradient derivative of the image

4: mag, ang ← cartToPolar () Converting Cartesian to Polar coordinates

5: Perform Double Thresholding step

6: Setting the minimum and maximum thresholds

7: S1, S2← where (mag ≥ strongTh)

8: W1, W2← where (mag ≤ weakTh)

9: for i_x in range(width)

10: for i_y in range(height) do

11: gradMag ← mag [i_y, i_x]

12: if gradMag ≤ weakTh then

13: mag [i_y, i_x] ←0

14: els if strongTh > gradMag ≥ weakTh then

15: ids [i_y, i_x] ←1

16: else

17: ids [i_y, i_x] ←2

18: end if

19: end for

20: end for

21: mag← merge(mag, mag, mag)

22: return mag, S1, S2, W1, W2

3.2.2 Least significant bit algorithm using arithmetic division operation

Image steganography is the technique to hide the patient’s medical record in a related medical cover image. This technique is introduced to secretly exchange information and avoid the attention of intruders. Images are only vulnerable if an intruder is aware of the data’s existence, whereas image steganography completely hides the data’s existence. Since the patient’s medical information is sensitive and should not be misused or altered at any cost, LSB using arithmetic division method is used in the proposed work. First, the patient’s data is converted into binary format considering each byte of the binary EHR data individually and represented in terms of the quotient and remainder as two parts in memory. Since 1 byte consists of 8 bits, the arithmetic division operation takes 1 byte and divides it by the common divisor of 8, the optimal value gives a 5-bit sized quotient and 3-bit sized memory. Analyze the size of all byte quotients and determine the size pattern used to store all quotients according to size. Store the quotient data optimally by replacing the LSB of the image only on the weak threshold coordinates that are the non-edge coordinates detected in canny detector. Similarly, the remainder is embedded on the strong threshold coordinates that are the edges of the image detected by the canny edge detection algorithm. The main purpose of dividing the binary data into two parts and embedding them in two different parts is an effort to ensure imperceptibility by embedding high-capacity data and good stego image quality while at the same time serving as a countermeasure making it difficult to retrieve the hidden information. The steps for embedding the patient data into a medical image using the LSB with Arithmetic Division Operation is given in the algorithm 2.

Algorithm 2 Least Significant Bit Algorithm using Arithmetic Division Operation

Data: Strong Edges (S1, S2), Weak Edges (W1, W2) PatientData P Output: Patient Data Embedded Image (E)

P_B ← binaryconvert (P)

for i in range(size (P_B)) do

q_i, r_i ← ArithmedicDivision (P_{B
_i}, 8)

store (q_i, W1_i, W2_i) //stores quotiet into weak threshold coordinates.

store (r_i, S1_i, S2_i) //stores remainder into strong threshold coordinates.

end for

3.3 Encryption using Chaotic Maps

Since steganography is susceptible to passive attacks, encryption is performed on stego images to protect data embedded medical images from passive attacks. Chaotic encryption is used to make it complex enough that an intruder cannot break the algorithm and crack the keys used for encryption. This data encryption process is proposed for its great encryption quality of confusion and diffusion implemented by the ability to generate control parameters and combine multiple maps. A chaotic map is a discrete map, usually in the form of an iterative function, used to generate keys for encrypting images, called “chaos”. It exhibits non-deterministic behavior, is highly sensitive to initial conditions, is difficult to crack, and is suitable as a key. The field of chaos theory, as explored in [18], is a mathematical discipline that has widespread applications in various fields such as meteorology, economics, and philosophy. Several image encryption methods have been developed using one-dimensional, two-dimensional, or multiple chaotic systems, concatenated chaotic maps, and other related techniques. Chaos ciphers are gaining more attention than other ciphers due to their lower mathematical complexity and better security. It is highly recommended by researchers for its computational efficiency, speed, cost, complexity, and computational overhead. Six one-dimensional maps are utilized in this work to generate keys for encrypting medical images, providing a fast, cost-effective, and secure method of encryption.

The six chaotic maps are the Chebyshev map, the Gaussian map, the Henon map, the logistic map, the Tent map and the piecewise map. Generate six different keys and XOR the keys with the medical image separately to get the encrypted image. These six maps are specifically chosen for this work because with thousands of chaotic maps available these six are the most stable maps and when used together maintains uniformity in encrypting each pixel in the image. Thus, the results of the encrypted image are uniform or a flat pattern, which can be seen when generating a histogram of the encrypted image which also implies that the encryption prevents attacks. Using the control parameters and the initial conditions the keys for each map are generated and generated six different keys are XORed with the stego-medical image separately to get the encrypted image. The explanation, benefits of each map used and their equations are discussed in the following sections.

3.3.1 Chebyshev map

In infinite computation for accuracy, the Chebyshev map can generate aperiodic and chaotic sequences with real values of infinite length. It allows the users and trustworthy servers to efficiently eliminate the need of key management produce the shared encryption key and agree on a session key. The Chebyshev map is a potential cryptographic primitive with good chaotic qualities such as mixing and ergodicity. The created chaotic sequences have acceptable statistical distribution properties since the mean is 0. It provides a reasonable level of security and randomness owing to its chaotic property. The Chebyshev map [21] is defined in Equation 1, $x_{n + 1} = cos (n * arccos (x_{n}))$ (1)

3.3.2 Gaussian map

The Gaussian map, which is named after Johann Carl Friedrich Gauss, is an iterative chaotic map that is non-linear. This map is similar to logistic map as it maps the bell-shaped Gaussian function. The Gauss map [33] is defined in Equation 2, $x_{n + 1} = e^{- α x^{2} n} + β$ (2) where α and β are the real numbers with α >0 (otherwise, unbounded and not chaotic) and with β in range -1 ≤ β ≤ 1 shows the strongest chaotic property. Here the value of α is 4.7 and β is 0.6.

3.3.3 Henon map

Henon chaotic system is advantageous on aspects like conditions for initial value and sensitive to system parameter that fulfills important requirements of cryptography. Henon map has a rich variety of dynamic behaviors, including periodic orbits, quasi-periodic orbits, and chaotic attractors, depending on the choice of parameters. The value of the Henon Chaotic map is intermittent or converges to a periodic orbit and is used to generate the random binary number for circular right-bit shift. The Henon map [3] is defined in Equation 3 and 4, $x_{n + 1} = 1 - α x_{n}^{2} + y_{n}$ (3) $y_{n} = b * x_{n}$ (4) The transformation function operates on a point (x_n, y_n) located in a two-dimensional plane. The function modifies the coordinates of the original point to produce a new point. The state variables are represented by x_n and y_n, while the control parameters are denoted by α and b. Specifically, the value of α is set at 1.4, while b is 0.3.

3.3.4 Logistic map

The logistic chaotic map refers to a specific type of mathematical function that produces a series of numerical values, known as the "orbit" of the map. This sequence is determined by Equation 5, $x_{n + 1} = λ x_{n} (1 - x_{n})$ (5) where x_n+1 is the n^th number in the orbit, x_n is the previous number in the orbit, and λ is a constant called the "control parameter". The orbit of the map is dependent on the initial value of x, called seed. Logistic chaotic maps [13] can exhibit chaotic behavior, meaning that small changes in the initial seed or the control parameter can lead to large changes in the long-term behavior of the orbit. Here λ is known as a bifurcation parameter that has a range from 0 to 4, thus the chaotic nature is obtained when “λ”ž ranges from 3.5694 <λ≤ 4. Here the value is chosen as 2.

3.3.5 Tent map

The tent map, as a component of the non-smooth piecewise chaotic map, exhibits strong chaotic behavior for a control parameter value of μ = 2. Equation 6 defines the Tent map [17], $z_{n + 1} = {\begin{matrix} μ z_{n}, for z_{n} < 0.5 \\ μ (1 - z_{n}), for z_{n} \geq 0.50 \end{matrix}$ (6)

Assuming a control parameter value of μ = 2, the chaotic Tent map has an initial condition of z_n, and μ is restricted to the range of [0, 2].

3.3.6 Piecewise map

A piecewise linear chaotic map is utilized to dilute each pixel of the picture. It shuffles the locations and diffuses the values of pixels in a plain picture at the same time. Therefore, the proposed work not only delivers robust encryption outcomes, but also showcases its capacity to resist brute-force and statistical attacks by employing a large key space. The method is highly responsive to even minor changes in the image and key, rendering it resilient to differential attacks. Piece wise map [20] is defined in Equation 7, $x_{n + 1} = (x_{n} - (x_{n} / q)) / q$ (7) Here the value of q is 0.4.

3.4 DNA encoding

Due to the massive quantity of data created during encryption, there is an exponential increase in demand for an efficient way of digital data storage. DNA-based data storage technologies meet this enormous need for data storage. A data storage solution with a storage density in the order of magnitude is provided by DNA. There are also other encoding techniques like Binary encoding, ASCII, Base64, QR and Morse code. But DNA is beneficial in terms of high storage capacity, stability, fidelity, durability and security with less data loss, it is used in the encoding. DNA provides a scalable, error-free, random-access data storage technology. It doesn’t require much room to store a lot of information. It provides an extremely high level of data protection while being stored. A DNA sequence has four nucleobases: T (Thymine), C (Cytosine), A (Adenine), and G (Guanine). C and G are complementary to each other, while T and A are complementary to each other. DNA coding is used to map the nucleotide sequences that make up a DNA strand. Regarding binary, 1’s and 0’s complement each other. So, these binary digits (in the form of 1s and 0s) are changed into the letters A, C, G, and T in order to store binary data in DNA. If the DNA sequence represents an 8-bit gray image, then each pixel has a length of 4. Ongoing beyond the DNA encoding rules by encoding the obtained bitstream [10101100], [GGTA] is obtained. Where 00 = A, 11 = T, 01 = C, 10 = G. From this example, it can be concluded that one pixel with a gray value of 172 can be represented by a DNA-encoded bit with the sequence [GGTA]. Now, from 4 types of DNA nucleic acids, 24 types of combinations can be made. However, there are only eight combinations that follow the complementarity principle. In this work, one among the eight-complementarity encoding rules denoting A, T, C, and G as 00, 11, 01, and 10 is used respectively, which is shown in Table 1. The overall steps of the proposed technique is given in the algorithm 3.

Table 1
DNA Encoding Rule

CODE RULE

A 00

C 11

T 01

G 10

CODE	RULE
A	00
C	11
T	01
G	10

Algorithm 3 Proposed Hybrid Image Encryption Algorithm

Data: MedicalImages M, PatientData P Output: Stegnographic Image

Edge Detection

mag, S1, S2, W1, W2 ← CannyEdgeDetection (M) //Using the Algortihm 1

[2.]

Embed Patient Data using Least Significant Bit Algorithm using Arithmetic Division Operation

E ← EmbedPatientData (S1, S2, W1, W2, P) //using Algorithm 2

[3.]

Encryption using Chaotic Maps

CM ← CalculateChebyshevmap () // using Equation 1

GM ← CalculateGaussianmap () // using Equation 2

HM ← CalculateHenonmap () // using Equation 3

LM ← CalculateLogisticmap () // using Equation 5

TM ← CalculateTentmap () // using Equation 6

PM ← CalculatePiecewisemap () // using Equation 7

EI ← Encrypt (E, CM, GM, HM, LM, TM, PM)

DNA Encoding

D ← LDNAEncoding (EI) // using the Table 1

Finally, to reduce the total size of the output file for an faster transmission a bz2 compression algorithm is used here. bz2 is a file compression format that uses the Burrows-Wheeler method which is well-known for its excellent compression ratio and speed. Additionally, the bz2 format includes a built-in error detection mechanism, which allows it to detect and correct errors that may occur during the compression or decompression process. The sequence of data after DNA encoding is compressed as a zip file and transmitted.

4 Experimental analysis

Common cryptographic approaches employ keys that are so difficult for intruders to find out. The complexity of breaking a secure system now grows with chaotic key generation. In order to assist the medical team and patients who are handling medical data securely a hybrid image encryption strategy using an advanced least significant bit algorithm, chaotic map and DNA encoding has been developed in this study. The LSB algorithm provides a secure substitution of image pixel values, while chaotic maps add a level of randomness to the encryption process. Six chaotic maps are utilized in this system to encrypt medical data sequentially with six keys generated in a specific order with predefined initial values. The reason to use six distinct encryption keys is to increase the integrity and robustness of the system. DNA encoding adds an extra layer of complexity and security to the encryption, for secure and fast transmission, to store maximum data and minimize the loss of data during transmission. It also reduces the size of the data to be transmitted hence increasing data transmission speed. After assessing the system using standard criteria, highly encouraging results were obtained.

The following measures are being used to assess the proposed method’s privacy, robustness, and efficiency:

4.1 Time coefficient of encryption and decryption

Selecting an encryption method involves considering the duration required for encoding and decoding data, which impacts the system’s effectiveness and productivity. The duration of encryption and decryption relies on several variables, including the complexity of the encryption algorithm, the amount of data being encrypted or decrypted, and the processing capability of the device or system being employed.

4.2 Histogram

A histogram is a form of graphical representation of an image that illustrates the proportionate frequency of different gray levels present within it, which offers an overall understanding of its visual appearance. In Fig. 3, both the original image and its encrypted counterpart are presented side by side, each displaying their respective histograms.

Fig. 3

The histograms of the original and encrypted image were compared in this analysis.

4.3 Entropy

Entropy pertains to the level of unpredictability or disorder present within the cover image utilized to conceal confidential data. The degree of entropy can be assessed through different methods. One such method is Shannon entropy, which quantifies the amount of data contained in the image. An image with high entropy exhibits greater randomness, allowing for the embedding of additional information without affecting its visual features considerably. The entropy value can be calculated by using the Equation 8, $H (x) = \sum_{i = 0}^{n - 1} pb (x) {log}_{2} pb (x)$ (8) where H is the entropy, pb(x) is the probability of the occurrence of pixels in the image.

4.4 Similarity score

The similarity score between two images can be evaluated using the MSE. This metric quantifies the difference between the pixel values of the two images being compared. It is determined by adding the squared differences between the pixel values of the two images, and then dividing it by the total number of pixels. The MSE calculated in Equation 9 is used to assess the quality of a steganographic system by comparing the original image to the steganographically modified image. A lower MSE means that the disparity between the two images is less. $MSE = (\frac{1}{WH}) * \sum_{i = 0}^{W} \sum_{j = 1}^{H} {[M (i, j) - N (i, j)]}^{2}$ (9) Where M is the pixel value of the original and N is the pixel value of the encrypted images. (i, j) is the location of pixel at that iteration, W and H are dimensions of the image.

4.5 Quality score

The quality score of a signal, commonly known as peak signal-to-noise ratio (PSNR), is a measure of the ratio between the maximum power that a signal can have and the amount of noise present in the signal. This ratio is often expressed in decibels (dB). The PSNR values are calculated using the Equation 10 which is used to evaluate the difference between the original image and the steganographically modified image. A higher PSNR value indicates a smaller difference between the two images. $PSNR = 20 * (\frac{255}{\sqrt{MSE}}) dB$ (10)

4.6 Motion activity index

The motion activity index, commonly known as number of changing pixel rate (NCPR) is a measure of the percentage of the pixels that are dissimilar between two given images. Here the comparison is done between the original image to a steganographically modified one to assess the quality of a steganographic system. A lower NCPR score implies that there is a smaller variation between the two images. The variables A1 and A2 are associated with the pixel values of the original and encrypted images, respectively. $D (x, y) = {\begin{matrix} 0 if A 1 (x, y) = A 2 (x, y) \\ 1 if A 1 (x, y) \neq A 2 (x, y) \end{matrix}$ (11) $NPCR = \sum \frac{D (x, y)}{P} * 100 %$ (12) Where P represents the total number of pixels

4.7 Unified average changing intensity (UACI)

Here we calculate the mean absolute difference of the pixel values of the two images to get the unified mean change intensity. UACI value can be calculated according to the Equation 13. This method evaluates the quality of the steganographic system by comparing the original and steganographically modified images. $UACI = \sum_{i, j} \frac{| A 1 (i, j) - A 2 (i, j) |}{L * P} * 100 %$ (13)

Here, P refers to the total number of pixels in the encrypted image, while L is the maximum pixel value that is compatible with the image.

4.8 Image similarity score

The image similarity scores, commonly known as Structure similarity index measurement (SSIM) is based on the perception of human vision and is intended to reflect the perceived quality of an image. To evaluate the similarity between two images, the SSIM uses Equation 14 to compare the luminance, contrast, and structure of the images. This index is designed based on the fact that the human visual system is more sensitive to certain types of image differences than others. It considers the statistical dependencies between the pixels in the images to provide a more comprehensive evaluation of their similarity. $SSIM (i, j) = \frac{(2 μ_{i} μ_{j} + A_{1}) (2 σ_{ij} + A_{2})}{(μ_{i}^{2} + μ_{j}^{2} + A_{1}) (σ_{i}^{2} + σ_{j}^{2} + A_{2})}$ (14) Where μ_iμ_j is the average of i, j. σ_ij is the covariance of i, j. A₁ and A₂ are two constants to stabilize division with a weak denominator.

5 Results and discussion

The results analysis always helps in making the right choice towards a successful outcome. To show the result analysis of the proposed method is better, several medical images of different sizes were taken to analysis different performance metrics on each of them. The advantage of the proposed technique is to embed high capacity data in a medical image and transmit it. For users to utilize this method to a vast scope, data files of different capacities are taken and embedded in different medical images and various metrics were applied to prove that neither the quality of the image nor the content of the data retrieved is deviated from the input. The purpose of doing this analysis is to prove the flexibility of the proposed method, that it suits data files of different size and still get the expected results. The images used for this comparison are Chest XRay image of dimension 368 x 372 taken from [23] dataset, the Knee MRI image of dimension 512 x 512 and Neck MRI image of dimension 512 x 319 taken from [5], Lungs CT Scan image of dimension 512 x 512 taken from [32]. Tables 2–4 shows the metric comparison for images that embed 500 characters, 1000 characters and 1500 characters respectively.

Table 2
Evaluation measures for 500 characters

Image Chest Knee Lungs Neck

Metrics XRay MRI CT Scan MRI

Dimension 368×372 512×512 512×512 512×319

Encryption time (s) 10 18 16 12

Decryption time (s) 12 19 18 13

Entropy 7.99 7.99 7.94 7.99

MSE 4748.12 4748.12 2178.19 4071.56

PSNR 71.34 74.48 74.37 72.15

UACI 42.83 36.26 38.14 35.07

NPCR 99.61 99.61 98.83 99.63

SSIM 0.99998 0.99999 0.99999 0.99999

Image	Chest	Knee	Lungs	Neck
Dimension	368×372	512×512	512×512	512×319
Encryption time (s)	10	18	16	12
Decryption time (s)	12	19	18	13
Entropy	7.99	7.99	7.94	7.99
MSE	4748.12	4748.12	2178.19	4071.56
PSNR	71.34	74.48	74.37	72.15
UACI	42.83	36.26	38.14	35.07
NPCR	99.61	99.61	98.83	99.63
SSIM	0.99998	0.99999	0.99999	0.99999

Table 3

Evaluation measures for 1000 characters

Image	Chest	Knee	Lungs	Neck
Metrics	XRay	MRI	CT Scan	MRI
Dimension	368×372	512×512	512×512	512×319
Encryption time (s)	16	18	17	11
Decryption time (s)	13	20	20	12
Entropy	7.99	7.99	7.94	7.99
MSE	8809.60	4749.29	4589.08	7794.13
PSNR	68.54	71.40	71.35	69.25
UACI	42.83	36.26	38.14	35.07
NPCR	99.61	99.61	98.83	99.63
SSIM	0.99996	0.99998	0.99996	0.99998

Table 4

Evaluation measures for 1500 characters

Image	Chest	Knee	Lungs	Neck
Metrics	XRay	MRI	CT Scan	MRI
Dimension	368×372	512×512	512×512	512×319
Encryption time (s)	9	16	26	11
Decryption time (s)	8	18	18	12
Entropy	7.99	7.99	7.94	7.99
MSE	13046.40	6973.26	6793.97	11351.39
PSNR	66.87	69.77	69.68	67.60
UACI	42.83	36.26	38.14	35.07
NPCR	99.61	99.61	98.83	99.63
SSIM	0.99992	0.99997	0.99991	0.99997

The efficacy of cryptography depends on how effectively the data is concealed using the encryption key. Larger keys are preferred to enhance key strength. To ascertain the superiority of the proposed method over other existing techniques, Table 5 illustrates a comparative analysis of previous studies with the proposed system. With differing input data file sizes in the proposed method, the input data file of 500 characters is considered in this comparison. The proposed method takes 10 and 12 seconds as encryption and decryption time respectively. While Elamir and Mabrouk [6] encrypted 100 characters getting 8.3 seconds as encryption time and 10.03 seconds as decryption time. This shows the proposed method is more stable and embeds comparatively more data. MSE value is the main factor metric used to determine how well the data is hidden and cannot be interpreted easily commonly which is low as 4748.12. Babu et al. [21] proposed his idea with four encryption method namely, DNA Sequence Encryption, Secure force algorithm, RSA encryption and Arnold Cat Map Encryption getting entropy value as 7.99. While the proposed hybrid method yields entropy value similar to it. Further evaluations of the proposed work, in comparison to the methods outlined in refs. 19, 32, 33, 34, and 35, demonstrate noteworthy enhancements in entropy, UACI, and NPCR, providing compelling evidence of the heightened level of security and safety achieved through the proposed algorithm. The SSIM value is on a scale from –1 to 1, with a value of 1 indicating that the images are indistinguishable, and a value closer to –1 indicating greater dissimilarity. The proposed approach produced a value of 0.9999, confirming that the original image and the stenographic image are almost identical. Ashwak et al. [1] combined Least Significant Bit with Chaotic system reporting PSNR value of 45.9012 db. A higher PSNR value indicates that the image quality has suffered less degradation during encryption. In the proposed work, a PSNR value of 71.34 dB was achieved, which is superior to those of other comparable systems, verifying that the image quality remains uncompromised even after being encrypted. The utilization of a hybrid encryption method can safeguard confidential medical information from unauthorized modifications and access, thereby upholding the confidentiality and security of patients’ personal health data.

Table 5

Comparison of results with other studies

Algorithm	Enc. Times(s)	Dec. Times(s)	Entropy	MSE	PSNR	SSIM	UACI	NPCR
Chaotic map + DNA encoding + diffusion [38]	-	7.422	-	-	-	28.2970	99.7017
LSB + Chaotic system [1]	-	-	7.84	-	45.9	-	-	-
Chaotic map + DNA coding [9]	-	-	7.98	16,138	6.05	0.0067	33	99.9
DNA Sequence Encryption + Secure Force + RSA encryption + Arnold Cat Map Encryption [21]	3.23	-	7.99	9131	5.8	-	38.89	99.984
Chaotic map encryption [37]	7.98	-	7.84	10417	7.95	-	32.82	98.698
LSB + Chaotic + DNA [6]	8.3	10.03	7.94	13,743	26.7	0.0067	33.124	99.61
Proposed Method	10	12	7.99	4748.12	71.34	0.9999	42.83	99.61

6 Conclusion

Ensuring the confidentiality and safety of patient data is a crucial matter within the eHealth sector. This is due to the sensitive nature of the information being stored and shared, as well as the potential consequences if this information were to fall into the wrong hands. A secure technique for e-health care can help ensure that the information remains confidential and protected at all times. In the proposed study, the hybrid image encryption system using an advanced Least significant Bit algorithm, chaotic maps, and DNA encoding was tested on medical images and comparisons were made with other existing techniques to make a detailed analysis. The results were advantageous to the proposed system stating its effectiveness in providing robust and secure encryption of medical data in medical images for transmission in digital healthcare. The proposed technique is tested with only with maximum of 1500 characters. But, for real time applications 1500 characters are not sufficient. So, in future wavelet based steganography can be incorporated to increase the number of characters to be embedded into a medical image.

This approach can be adopted to applications in a variety of fields, including blockchains, cybersecurity, and the Internet of Things. In blockchains, the technique can be used to create a secure and tamper-proof ledger of patient information, ensuring that the data is accurate and cannot be altered without detection. In Cybersecurity, this can be used to secure the transmission of patient information to mobile locations without physical effort, while the Internet of Things can be used to connect medical devices and gather data from patients, with the secure technique helping to protect the privacy and security of information.

References

Alabaichi

Ashwaq

, Al-Dabbas

Maisa’a Abid Ali K.

and Salih

Adnan

, Image steganography using least significant bit and secret map techniques, International Journal of Electrical and Computer Engineering (IJECE) 10(1) (2020), 935.

Chaudhary

Nirmal

, Shahi

Tej Bahadur

and Neupane

Arjun

, Secure image encryption using chaotic, hybrid chaotic and block cipher approach, Journal of Imaging 8(6) (2022), 167.

Chen

Jinlin

, Wu

Yiquan

, Sun

Yeguo

and Yang

Chunzhi

, Image encryption algorithm using 2-order bit compass coding and chaotic mapping, Symmetry 14(7) (2022), 1482.

Darbani

Abbas

, AlyanNezhadi

Mohammad M.

and Forghani

Majid

, A new steganography method for embedding message in JPEG images. In 2019 5th Conference on Knowledge Based Engineering and Innovation(KBEI). IEEE, 2019.

DICOM. Dicom medical library. https://www.dicomlibrary.com/, 2016.

Elamir

Mona M.

, Al-atabany

Walid I.

and Mabrouk

Mai S.

, Hybrid image encryption scheme for secure e-health systems, Network Modeling Analysis in Health Informatics and Bioinformatics 10(1) (2021).

Elharrouss

Omar

, Almaadeed

Noor

and Al-Maadeed

Somaya

, An image steganography approach based on k-least significant bits (k-LSB). In 2020 IEEE International Conference on Informatics, IoT, andEnabling Technologies (ICIoT). IEEE, 2020.

Fernandes

Aisha

and Jeberson

Wilson

, Covert communication using arithmetic division operation, Procedia Computer Science. 45 (2015), 354–360.

Gasimov

V.A.

and Mammadov

J.I.

, DNA-based image encryption algorithm, IOP Conference Series: Materials Science and Engineering 734(1) (2020), 012162.

10.

Gulia

Sarita

, Mukherjee

Saurabh

and Choudhury

Tanupriya

, A nextensive literature survey on medical image steganography, CSITransactions on ICT 4(2-4) (2016), 293–298.

11.

Hassan

Rosilah

, Pepic

Selver

, Saracevic

Muzafer

, Ahmad

Khaleel

and Tasic

Milan

, A novel approach to data encryption based on matrixcomputations, Computers, Materials and Continua 66(2) (2021), 1139–1153.

12.

Hraoui

, Gmira

, Jarar

A.O.

, Satori

and SaaidiBenchmarking

, AES and chaos based logistic map for image encryption.In 2013 ACS International Conference on Computer Systems andApplications (AICCSA). IEEE, 2013.

13.

Yuping

, Zhu

Congxu

and Wang

Zhijian

, An improved piece wise linear chaotic map based image encryption algorithm, The Scientific World Journal. 2014 (2014), 1–7.

14.

Jayanthi

Prisilla

and Iyyanki

Muralikrishna

, Cryptography in the healthcare sector with modernized cyber security. In Quantum Cryptography and the Future of Cyber Security, pages 163–183. IGI Global, 2020.

15.

Jayapandiyan

Jagan Raj

, Kavitha

and Sakthivel

, Enhanced least significant bit replacement algorithm in spatial domain of steganography using character sequence optimization, IEEE Access. 8 (2020), 136537–136545.

16.

Kester

Quist-Aphetsi

, Nana

Laurent

, Pascu

Anca Christine

, Gire

Sophie

, Eghan

Jojo M.

, and Quaynor

Nii Narku

, A cryptographic technique for security of medical images in health information systems, Procedia Computer Science. 58 (2015), 538–543.

17.

Khan

Muhammad Asif

and Jeoti

Varun

, Modified chaotic tent map withimproved robust region. In 2013 IEEE 11th MalaysiaInternational Conference on Communications (MICC), IEEE. 2013.

18.

Kumar

Vijay

and Girdhar

Ashish

, A 2d logistic map and lorenz-rossler chaotic system based RGB image encryption approach, MultimediaTools and Applications 80(3) (2020), 3749–3773.

19.

Liu

Lidong

, Lei

Yuhang

and Wang

Dan

, A fast chaotic image encryptionscheme with simultaneous permutation-diffusion operation, IEEEAccess. 8 (2020), 27361–27374.

20.

Hui

, Wang

Xiaoteng

, Fei

Zongming

and Qiu

Meikang

, The effects of using chaotic map on improving the performance of multi objective evolutionary algorithms, Mathematical Problems in Engineering. 2014 (2014), 1–16.

21.

Babu

, Image encryption using chaotic maps and DNA encoding, Journal of Xidian University 14(4) (2020).

22.

Al Mamun

Abdullah

, Azam

Sami

and Gritti

Clementine

, Block chain-based electronic health records management: A comprehensive review and future research direction, IEEE Access 10 (2022), 5768–5789.

23.

Mirsky

Yisroel

, Lungs ct-scan. https://doi.org/10.24432/C5J318, 2019.

24.

Mirsky

Yisroel

, Mahler

Tom

, Shelef

Ilan

and Elovici

Yuval

, Ct-gan: Malicious tampering of 3d medical imagery using deep learning, 2019.

25.

Mohamed

Marghny H.

and Mohamed

Loay M.

, High capacity image steganography technique based on LSB substitution method, Applied Mathematics & Information Sciences 10(1) (2016), 259–266.

26.

Nashat

Dalia

and Mamdouh

Loay

, An efficient steganographic technique for hiding data, Journal of the Egyptian Mathematical Society 27(1) (2019).

27.

Osunade

, I.A. , Enhancing the least significant bit (LSB) algorithm for steganography, International Journal of Computer Applications 149(3) (2016), 1–8.

28.

Patel

Sakshi

, Bharath

K.P.

and Muthu

Rajesh Kumar

, Image encryption decryption using chaotic logistic mapping and dna encoding, 2020.

29.

Saračeević

Muzafer

, Adamović

Saša

, Miškovic

Vladislav

, Maček

Nemanja

and Šarac

Marko

, A novel approach to steganography based on the properties of catalan numbers and dyck words, Future Generation Computer Systems 100(2019), 186–197.

30.

Saračević

Muzafer

, Sharma

Sudhir Kumar

and Ahmad

Khaleel

, A novel block encryption method based on catalan random walks, Multimedia Tools and Applications 81(25) (2021), 36667–36684.

31.

Selimović

Faruk

, Stanimirović

Predrag

, Saračević

Muzafer

, Selimi

Aybeyan

and Krtolica

Predrag

, Authentication based on the image encryption using delaunay triangulation and catalan objects, Acta Polytechnica Hungarica 17(6) (2020), 207–224.

32.

Summers and Ronald , Chestxray14. https://nihcc.app.box.com/v/ChestXray-NIHCC, 2019.

33.

Suryadi

M.T.

, Satria

Yudi

and Prawadika

Luqman N.

, An improvement on the chaotic behavior of the gauss map for cryptography purpose susing the circle map combination, Journal of Physics: Conference Series 1490(1) (2020), 012045.

34.

Ziqi

, Ji

Xiaoqiang

, Wang

Meijiao

and Sun

Xiaobing

, Edge detection algorithm of medical image based on canny operator, Journal of Physics: Conference Series 1955(1) (2021), 012080.

35.

Choudhary

Nilesh Y.

and Gupta

Ravindra K.

, Partial image encryption based on block wise shuffling using arnold map, International Journal of Computer Applications 97(10) (2014), 33–37.

36.

Yigit

Yildiray

and Karabatak

Murat

, A stenography application for hiding student information into an image. In 2019 7th International Symposium on Digital Forensics and Security (ISDFS). IEEE, June 2019.

37.

Yousif

Bedir

, Khalifa

Fahmi

, Makram

Ahmed

and Takieldeen

Ali

, Anovel image encryption/decryption scheme based on integrating multiple chaotic maps, AIP Advances 10(7) (2020), 075220.

38.

Zhang

Jian

, Fang

DongXin

and Ren

Honge

, Image encryption algorithmbased on DNA encoding and chaotic maps, Mathematical Problemsin Engineering. 2014 (2014), 1–10.

39.

Zia

Unsub

, McCartney

Mark

, Scotney

Bryan

, Martinez

Jorge

, AbuTair

Mamun

, Memon

Jamshed

and Sajjad

Ali

, Survey on image encryption techniques using chaotic maps in spatial, transform and spatiotemporal domains, International Journal of InformationSecurity 21(4) (2022), 917–935.

Hybrid image encryption using advanced least significant bit algorithm,chaotic maps and DNA encoding for digital healthcare

Abstract

Keywords

1 Introduction

2 Related Works

3 Proposed work

3.2 Embedding medical data into a medical image

3.2.2 Least significant bit algorithm using arithmetic division operation

3.3 Encryption using Chaotic Maps

3.3.1 Chebyshev map

Table 1 DNA Encoding Rule CODE RULE A 00 C 11 T 01 G 10

4.1 Time coefficient of encryption and decryption

4.2 Histogram

References

Table 1
DNA Encoding Rule

CODE RULE

A 00

C 11

T 01

G 10