SPECIAL SECTION ON INFORMATION SECURITY SOLUTIONS FOR TELEMEDICINE APPLICATIONS

Received February 20, 2018, accepted March 17, 2018, date of publication March 21, 2018, date of current version May 2, 2018.
Digital Object Identifier 10.1109/ACCESS.2018.2817615

Secure Medical Data Transmission Model

for IoT-Based Healthcare Systems
MOHAMED ELHOSENY 1 , GUSTAVO RAMÍREZ-GONZÁLEZ2 , OSAMA M. ABU-ELNASR3 ,
SHIHAB A. SHAWKAT4 , ARUNKUMAR N5 , AND AHMED FAROUK 1
1 Faculty of Computers and Information, Mansoura University, Mansoura 35516, Egypt
2 Department of Telematics, University of Cauca, Popayán 190001, Colombia
3 Computer Science Department, College of Computes and Information, Mansoura University, Mansoura 35516, Egypt
4 Directorate of Education, Salah Al-Din, Iraq
5 School of Electrical and Electronics Engineering, SASTRA University, Thanjavur 613401, India

Corresponding author: Mohamed Elhoseny ([email protected])

ABSTRACT Due to the significant advancement of the Internet of Things (IoT) in the healthcare sector,
the security, and the integrity of the medical data became big challenges for healthcare services appli-
cations. This paper proposes a hybrid security model for securing the diagnostic text data in medical
images. The proposed model is developed through integrating either 2-D discrete wavelet transform 1 level
(2D-DWT-1L) or 2-D discrete wavelet transform 2 level (2D-DWT-2L) steganography technique with a
proposed hybrid encryption scheme. The proposed hybrid encryption schema is built using a combination of
Advanced Encryption Standard, and Rivest, Shamir, and Adleman algorithms. The proposed model starts by
encrypting the secret data; then it hides the result in a cover image using 2D-DWT-1L or 2D-DWT-2L. Both
color and gray-scale images are used as cover images to conceal different text sizes. The performance of the
proposed system was evaluated based on six statistical parameters; the peak signal-to-noise ratio (PSNR),
mean square error (MSE), bit error rate (BER), structural similarity (SSIM), structural content (SC), and
correlation. The PSNR values were relatively varied from 50.59 to 57.44 in case of color images and from
50.52 to 56.09 with the gray scale images. The MSE values varied from 0.12 to 0.57 for the color images and
from 0.14 to 0.57 for the gray scale images. The BER values were zero for both images, while SSIM, SC,
and correlation values were ones for both images. Compared with the state-of-the-art methods, the proposed
model proved its ability to hide the confidential patient’s data into a transmitted cover image with high
imperceptibility, capacity, and minimal deterioration in the received stego-image.

INDEX TERMS Cryptography, DWT-1level, DWT-2level, encryption, healthcare services, Internet of

Things, medical images, steganography.

I. INTRODUCTION be authorized personnel. The two main algorithms used for

IoT creates an integrated communication environment of data encryption in this work are the Advanced Encryp-
interconnected devices and platforms by engaging both vir- tion Standard (AES) and the Rivest-Shamir-Adleman (RSA)
tual and physical world together [1]. With the advent of algorithm [18]. AES is a symmetric cipher where the same
remote digital healthcare based IoT systems, the transmis- key is used on both sides [19]. It has a fixed message block
sion of medical data becomes a daily routine. Therefore, size of 128 bits of text (plain or cipher), and keys of length
it is necessary to develop an efficient model to ensure the 128,192, or 256 bits. When longer messages are sent, they
security and integrity of the patient’s diagnostic data trans- are divided into 128-bit blocks. Apparently, longer keys make
mitted and received from IoT environment [2]–[8]. This goal the cipher more difficult to break, but also enforce a longer
is carried out using steganography techniques and system encrypt and decrypt process. On the contrary, the RSA is
encryption algorithms together to hide digital information a public key algorithm, which widely used in business and
in an image [9]–[16]. personal communication sectors [20]. It has the advantage of
Cryptography is another term for data encryption [17]. having a variable key size ranging from (2-2048) bits.
Encryption cryptography is the process of encoding mes- The primary research in hiding data started with steganog-
sages in a way that hackers cannot read it, but that can raphy, which refers to the science and art of hiding

2169-3536
2018 IEEE. Translations and content mining are permitted for academic research only.
20596 Personal use is also permitted, but republication/redistribution requires IEEE permission. VOLUME 6, 2018
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
M. Elhoseny et al.: Secure Medical Data Transmission Model for IoT-Based Healthcare Systems

information within an image. The benefit of steganography is information to ensure that authorized people only can access
that it can be utilized to transmit classified messages without the information. First, the AES encryption technique was
the fact of the transmission being detected. The DWT has a applied on the first part. The ear print also embedded in this
tremendous spatial localization, frequency spread, and multi- work, where seven values were extracted as feature vector
resolution characteristics, which are matching with the theory from the ear image. The proposed technique improved the
of forms in the human visual system. This paper implements security of medical images through sending them via the
both 1-level and 2-level of DWT steganography techniques internet and secured these images from being accessed via
that operate on the frequency domain. It split up the image any unauthorized person.
into high and low iteration parts. The high iteration part Abdelaziz et al. [5], surveyed the analysis of the security
contains edge information, whereas the low iteration part is vulnerabilities and the risk factors detected in mobile medi-
frequently divided into high and low iteration parts [21]. cal apps. According to risk factor standards, these apps can
The purpose of the steganography is not only preventing be categorized into remote monitoring, diagnostic support,
others from knowing the hidden information, but also remov- treatment support, medical information, education and aware-
ing the suspicion in having hidden information. The message ness, and communication and training for healthcare workers.
is a confidential document to be transmitted and camou- Eight security vulnerabilities and ten risks factors detected
flaged in the carrier so that it becomes difficult to detect. by the World Health Organization (OWASP) mobile security
There are two main aspects in any steganography system, project in 2014 have been analyzed.
which are steganography capacity and imperceptibility. How- Razzaq et al. [9], proposed a fused security approach based
ever, these two properties are confusing with each other on encryption, steganography, and watermarking techniques.
because it is tough to increase capacity while maintaining It decomposed into three stages; (1) encrypting the cover
the steganography imperceptibility of a steganography sys- image using XOR operation, (2) embedding process done
tem. Furthermore, there are still limited methods of conceal- using least significant bits (LSBs) for generating the stego-
ing information for use with data transfer communication image, then (3) watermarking the stego-image in both spatial
protocols, which can be unconventional but their future is and frequency domains. Experimental results proved that
promising. proposed method was very much efficient and secured.
This paper aims to improve the security of medical data Jain et al. [11], proposed a new technique for trans-
transmission based on the integration between a steganogra- ferring the patient’s medical information into the medical
phy technique and a hybrid encryption scheme to get a highly cover image by hiding the data using decision tree concept.
secured healthcare system. The coding is done in the form of different blocks that evenly
This paper is organized into five sections includ- distributed. In concealment, secret code blocks are assigned
ing this section. Section 2 illustrates the related works; to the cover image to insert the data by the mapping mecha-
Section 3 explains the proposed model and their algorithms; nism based on breadth-first search. RSA algorithm was used
Section 4 provides the experimental results and their discus- to encipher the data before embeddings.
sions, and Section 5 summarizes the main conclusions. Yehia et al. [12], surveyed various healthcare applications
based on wireless medical sensor network (WMSN) that can
II. RELATED WORKS be implemented in IoT environment. Also, discussed the
Shehab et al. [2], surveyed a comprehensive study on security security techniques that used for handling the security issues
issues in IoT networks. Various security requirements such of healthcare systems especially hybrid security techniques.
as authentication, integrity, confidentiality were discussed. Zaw and Phyo [13], presented an algorithm based on
A comparative study of different types of attacks, their behav- dividing the original image to the group of blocks, where
ior, and their threat level that categorized into low-level, these blocks are arranged in the form of turns using a trans-
medium-level, high-level, and extremely high-level attacks formation algorithm. After that, the transformed image is
and suggested possible solutions to encounter these attacks encryption using the Blowfish algorithm. It was found that the
were provided. correlation decreases and the entropy increase by increasing
Bairagi et al. [3], proposed three color image steganogra- the number of blocks through using smaller block sizes.
phy approaches for protecting information in an IoT infras- Sreekutty and Baiju [21], proposed a medical integrity
tructure. The first and third approaches use three (red, green, verification system to improve the security of medical image.
and blue) channels, while the second approach uses two The proposed system mainly decomposed into two stages:
(green and blue) channels for carrying information. Dynamic 1) the protection and 2) the verification. Through the pro-
positioning techniques have been used for hiding information tection stage, the binary form of the secret data is embedded
in the deeper layer of the image channels with the help a in the high-frequency part (HH) within the cover image using
shared secret key. 2D Haar DWT frequency domain technique. Through the ver-
Anwar et al. [4], developed a technique to secure any type ification stage, the extraction algorithm is applied to retrieve
of images especially medical images. They aimed to main- the original cover image and secret data.
tain the integrity of electronic medical information, ensuring Bashir et al. [22], proposed an image encryption technique
availability of that information, and authentication of that based on the integration of shifted image blocks and the

VOLUME 6, 2018 20597

 M. Elhoseny et al.: Secure Medical Data Transmission Model for IoT-Based Healthcare Systems

FIGURE 1. The proposed framework for securing the medical data transmission.

FIGURE 2. The decomposition process of DWT-2L.

FIGURE 3. DWT decomposition of an image. (a) DWT-1L. (b) DWT-2L.

basic AES. The shifted algorithm technique is used to divide

the image into blocks. Each block consists of many pixels,
and these blocks are shuffled by utilizing a shift technique Yin et al. [24], proposed an image steganography approach
that moves the rows and columns of the original image in such based on Inverted LSB (ILSB) technique for securing the
a way to produce a shifted image. This shifted image is then transmitted face images from the IP camera as the IoT device
used as an input image to the AES algorithm to encrypt the to the home server in the LAN network. The local home server
pixels of the shifted image. serves as a processing power node for the encryption of the
Muhammad et al. [23], presented an efficient, secure stego images before transmitting them to the cloud and other
method for RGB images based on gray level modifica- devices for further processing.
tion (GLM) and multi-level encryption (MLE). The secret Seyyedi et al. [25], proposed a secure steganography
key and the secret data are encrypted using MLE algorithm method based on encrypting the confidential information
before mapping it to the gray-levels of the cover image. Then, using the symmetric RC4 encryption method and embed-
a transposition function is applied to the cover image before ding it within the cover image based on the partitioning
data hiding. The usage of transpose, secret key, MLE, and approach with minimal degrading of the quality. The cover
GLM adds four different levels of security to the proposed image is partitioned into a predefined 8 × 8 blocks. Each
algorithm, making it very difficult for a malicious user to block is manipulated using Integer Lifting Wavelet Trans-
extract the original secret information. form (ILWT) method, then TSO (Tree Scan Order) is applied

20598 VOLUME 6, 2018

M. Elhoseny et al.: Secure Medical Data Transmission Model for IoT-Based Healthcare Systems

TABLE 1. Results of PSNR and MSE values using 2D-DWT-1L and

2D-DWT-2L for color images.

FIGURE 4. The synthesis process of 2D-DWT-2L.

FIGURE 5. Color and gray format of the dataset.

to each manipulated block to identify proper location of

confidential information.
Khalil [26], proposed a method that studies the medi-
cal image quality degradation when hiding data in the fre-
quency domain. The secret plaintext was encrypted using
RC4 encryption before the embedding process. The Discrete
Fourier Transform (DFT) was applied to transfer the cover
image into the frequency domain by decomposing it into its Sajjad et al. [29], proposed a domain-specific mobile-
sinusoidal (sine and cosine) fundamental components in dif- cloud assisted framework for outsourcing the medical stego-
ferent frequencies. The results indicated that the quality of the images to cloud for selective encryption. The visual saliency
image is exceptionally degraded when embedding data close detection model has been used for detecting the region of
to the low-frequency bands (DC) and this effect decreases interest (ROI) from the transmitted image. The directed-
in the upper-frequency bands. edge steganography method has been used for embedding
Abdel-Nabi and Al-Haj [27], proposed a crypto- the detected ROI in the cover image and producing the stego
watermarking approach based on AES standard encryption image which sent to cloud for selective encryption.
algorithm and reversible watermarking data hiding technique Parah et al. [30], proposed a secure high capacity scheme
to secure medical images. The results proved that the pro- capable of securing an electronic patient record (EPR) con-
posed approach achieves both the authenticity and integrity cealed in medical color images for an IoT based healthcare
of the images either in the spatial domain or the encrypted system. Two address vectors namely MAV (Main Address
domain or both domains. Vector) and CAV (Complementary Address Vector) have
Li et al. [28], proposed a secret image sharing scheme been generated as pseudorandom addresses to address the
compatible with an IoT- cloud framework for embedding the pixel locations for further embedding. It utilizes LSB method
secret image shares. The proposed scheme composed of two for concealing EPR using two/three RGB bit planes.
modules; shadow images generation module for generating
the secret shares based on the Shamir’s polynomial, and shar- III. THE PROPOSED MODEL
ing key formulation module for embedding the secret image This paper proposes a healthcare security model for secur-
shares into the cover image based on a 24-ary notational ing a medical data transmission in IoT environments.
system. The proposed model composes of four continuous processes:

VOLUME 6, 2018 20599

 M. Elhoseny et al.: Secure Medical Data Transmission Model for IoT-Based Healthcare Systems

(1) The confidential patient’s data is encrypted using a pro- transformation using low-pass and high-pass filters along the
posed hybrid encryption scheme that is developed from both rows of the image; then the result decomposed along the
AES and RSA encryption algorithms. (2) The encrypted columns of the image [21]. Fig. 2 elucidates this process.
data is being concealed in a cover image using either Fig. 2 illustrates the elemental decomposition process for
2D-DWT-1L or 2D-DWT-2L and produces a stego-image. Cj (n, m) image of a size N × M in four decomposed sub-
(3) The embedded data is extracted. (4) The extracted data band images which are referred to a high-high (HH), a high-
is decrypted to retrieve the original data. Fig. 1 shows the low (HL), a low-high (LH), and a low-low (LL) frequency
general framework of our proposed model for securing the bands. Fig. 3 shows the effect of the decomposition process
medical data transmission at both the source’s and the desti- on the image.
nation’s sides. The proposed model implements the steganographic
scheme. The steganographic scheme Ŝ = {fη, fη−1 , C, S, T}
A. DATA ENCRYPTION SCHEME is composed of embedding and extraction processes.
The proposed model implements the cryptographic scheme. The embedding process takes a cover image C and a secret
The cryptographic scheme Ĉ = {fη, fη−1 , C, S, T} is com- text message T as input and generates a stego-image S.
posed of encryption and decryption processes. Throughout While the extraction process inversely extracts the embedded
the encryption process, the plain text T is divided into odd part message. It can be mathematically modeled as given in the
Todd and even parts Teven . The AES is used to encrypt Todd following equations below.
using a secret public key s. The RSA is used to encrypt Teven
Ŝ = {fη, fη−1 , C, S, T} (5)
using a secret public key m. The private key x that used in the
decryption process at the receiver side is encrypted using AES S = {fη (C, T)} (6)
n o
algorithm and sent to the receiver in an encrypted form to T = fη−1 (S) (7)
increase the security level. The encryption process can be
mathematically modeled as given in the following equations Throughout the embedding process, the secret text is trans-
below. formed into an ASCII format and then divided into even and
odd values. The odd values are concealed in vertical coef-
C = 0 EAES ,0 ERSA , Todd , Teven ,Ṫodd , Ṫeven , s, m, x (1)

ficients mentioned by LH2. The even values are concealed
Ṫodd = 0 EAES (Todd , s)

(2) in diagonal coefficients specified by HH2. The algorithm that
Ṫeven = ERSA (Teven , m)
0
(3) is used in the embedding procedure by evolved 2D-DWT-2L
is described below.
Ẋ = EAES (x, s)
0
(4)
The algorithm used in the encryption procedure is described Algorithm 2 Embedding 2D-DWT-2L Algorithm
below. Inputs: cover image, a secret message (main_cipher and s).
Output: stego image.
Algorithm 1 Hybrid (AES & RSA) Algorithm
Begin
Inputs: secret plain Stext message.
1. Convert the secret message in ASCII Code as asciiMsg
Output: main_cipher message, key s
2. Divide asciiMsg to odd and even
Begin
3. Scan the image row by row as img
1. Divide plain msg into two parts (Odd_Msg, Even_Msg)
4. Compute the 2D wavelet for the first level by harr filter
2. Generate new AES key s
that generates (LL1), (HL1), (LH1), and (HH1)
3. EncOdd = AES-128 (Odd_Msg, s)
5. Compute the 2D wavelet for the second level by harr
4. Generate new RSA key (public = m) and (private = x)
filter that generates (LL2), (HL2), (LH2), and (HH2)
5. EncEven = RSA (Even_Msg, m)
6. Loop
6. Build FullEncTxt by inserting both EncOdd and
6.1 Hide odd values in vertical coefficient, set
EncEven in their indices
LH2(x,y) = odd values
7. EncKey = AES-128 (x, s)
6.2 Hide even values in vertical coefficient, set
8. Compress FullEncMsg by convert to hashs
HH2(x,y) = even values
9. Compress EncKey by convert to hashs
7. End Loop
10. Define message empty main_cipher = ""
8. Return Stego image
11. main_cipher = Concatenate (FullEncMsg, EncKey)
End
12. Return main_cipher and s
End
C. EXTRACTION PROCEDURE
After incorporating the text into the cover image, the
B. EMBEDDING PROCEDURE 2D-DWT-2L technique is carried to extract the secret mes-
In this process, A Haar-DWT was implemented. Throughout sage and retrieve the cover image. The extraction algorithm
Haar-DWT, 2D-DWT-2L can be formulated as a consecutive is described in Algorithm 3.

20600 VOLUME 6, 2018

M. Elhoseny et al.: Secure Medical Data Transmission Model for IoT-Based Healthcare Systems

Algorithm 3 Extraction algorithm Algorithm 4 Hybrid Decryption (AES & RSA) Algorithm
Inputs: stego image Inputs: main_cipher (secret) message, key
Output: Retrieved secret message and original cover image Output: secret (plain, text) message.
Begin Begin
1. Scan the stego image row by row 1. Divide main_cipher into two parts; HashedTxt and
2. Compute the 2D wavelet for the first level by harr filter HashedKey
3. Compute the 2D wavelet for the second level by harr filter 2. FullEncMsg = Decompress (HashedTxt)
4. Prepare msg = "" 3. EncKey = Decompress (HashedKey)
5. Loop 4. x = Decrypt_AES-128 (EncKey, s)
5.1 Extract the text embedded in vertical coeffi- 5. EncOdd = Split (FullEncMsg, odd)
cient, set odd values =LH2(x,y) 6. EncEven = Split (FullEncMsg, even)
5.2 Extract the text embedded in vertical coeffi- 7. Odd_Msg = Decrypt_ AES-128 (EncOdd, s)
cient, set even values = HH2(x,y) 8. Even_Msg = Decrypt_ RSA (EncEven, x)
6. End Loop 9. Define main_plain message
7. msg = Append (odd values, even values) 10. Loop on All Char
8. Compute idwt2 for the constructed approximation that 10.1 If odd Insert odd characters into odd indices
generates the original image within main_plain message
9. Return msg as a retrieved secret message and original 10.2 Else Insert even characters into even indices
cover image within main_plain message
11. End of Loop
12. Return main_plain (text) message
Once the secret text message has been extracted, the cover End
image is synthesized from the reconstructed approximation
by calling the idwt2 for the second level and then for the TABLE 2. Results of PSNR and MSE values using 2D-DWT-1L and
2D-DWT-2L for gray scale images.
first level [21]. Figure 4 elucidates the basic DWT synthesis
process.

D. DATA ENCRYPTION SCHEME

Decryption refers to the process of converting the encrypted
data back to the user in a well-known format; which is
the reverse of the encryption process. The same key used
by the sender has to be used over the cipher-text through-
out the encryption process. The decryption process can be
mathematically expressed as given in the following equations
below.

Ĉ = {0 E−1 , 0 E −1 , Todd , Teven , Ṫodd , Ṫeven , s, x (8)

0 AES RSA

x = EAES Ẋ, s (9)
RSA ((Ṫeven , x)}
Teven = {0 E−1 (10)
Todd = { EAES Ẋodd , s
0 −1


(11)

The proposed decryption algorithm is described in

Algorithm 4.

IV. EXPERIMENTAL RESULTS AND EVALUATION

A. SIMULATION ENVIRONMENT
The implementation of our proposed model was carried out
using the MATLAB R2015a software running on a personal
computer with a 2.27 GHz Intel (R) Core (TM) I3 CPU, 8 GB
RAM and Windows 7 as the operating system.
The calculation of specific statistical metrics determines
the quality of the proposed security model. These metrics
calculate the ratio between the original image and the stego Mean Square Error (MSE), Bit Error Rate (BER), Struc-
image. The obtained results were evaluated based on six tural Similarity (SSIM), Structural Content (SC), and
statistical parameters; the Peak Signal to Noise Ratio (PSNR), Correlation.

VOLUME 6, 2018 20601

 M. Elhoseny et al.: Secure Medical Data Transmission Model for IoT-Based Healthcare Systems

TABLE 3. Results of the six statistical parameters obtained from TABLE 4. Results of the six statistical parameters obtained from
performing the (2D-DWT-2L) with hybrid (AES and RSA) scheme on color performing the (2D-DWT-2L) with hybrid (AES and RSA) scheme on gray
images with different text sizes. images with different text sizes.

PSNR calculates the imperceptibility of the stego- BER calculates the probability that a bit will be incorrectly
image [31]. The higher the value of PSNR of stego image received due to noise [33]. It is the number of bits received
reveals a higher quality of stego image or a higher impercep- in error divided by the total number of bits transferred.
tibility of hidden message. The PSNR is calculated according The BER is calculated using the following equation:
to the following equation:
BER = Errors/Total Number of Bits (14)
I2
 
PSNR = 10 log10 (12) SSIM measures the structural similarity between two
MSE images [34]. Its value ranges between −1 and 1. When two
images are nearly identical, their SSIM is close to 1. The fol-
where I represent the maximum possible value of the pixel
lowing formula is used to compute the SSIM between two
in the image (e.g., for a gray-scale image the maximum value
sequences seq1 and seq2 at a given pixel P:
is 255) and MSE is the mean square error.
MSE calculates the magnitude of average error between the 2 ∗ mu1 (p) mu2 (p) + c1 2 ∗ cov (p) + c2
SSIM = ×
original image and stego-image [32]. The MSE is computed mu1 (p)2 +mu2 (p)2 + c1 s1 (p)2 +s2 (p)2 + c2
as depicted below: (15)
1 Xn Xm where, mu1 (P) and mu2 (P) are the mean value of seq1 , and
MSE = 2
(Xij −Yij )2 (13) seq2 computed over a small XY window located around
[R × C] i=0 j=1
P. While, s1 (P) and s2 (P) are the standard deviations of seq1,
where, R and C are the numbers of rows and columns in the and seq2 computed over the same window. And cov(P) is the
cover image, Xij , is the intensity of Xij the pixel in the covariance between seq1 and seq2 which are computed over
cover image, and Yij is the intensity of Yij the pixel in stego the same window. C1 = (K1 ∗ L)2 is a regularization constant,
image. C2 = (K2 ∗ L)2 is a regularization constant, K1 and K2 are

20602 VOLUME 6, 2018

M. Elhoseny et al.: Secure Medical Data Transmission Model for IoT-Based Healthcare Systems

FIGURE 6. Histogram of the cover image before and after applying the proposed model on color images with different text sizes (15, 30, 45 and 55 bytes).

regularization parameters (must be >0), and L is a dynamic according to the following equation:
range of the pixel values. PM PN
j=1 (y (i, j))
2
SC is a correlation-based measure, where it also mea- i=1
SC = PM PN (16)
j=1 (x (i, j))
sures the similarity between two images [35]. It is calculated 2
i=1

VOLUME 6, 2018 20603

 M. Elhoseny et al.: Secure Medical Data Transmission Model for IoT-Based Healthcare Systems

FIGURE 7. Histogram of the cover image before and after applying the proposed model on color images with different text
sizes (100, 128 and 256 bytes).

where, x(i, j) represents the original image and y(i, j) repre- by using the following equation:
sents the distorted image.
Correlation determines how much two signals or vectors xy−
P
P

xn y
P
are similar or different in phase and magnitude when two Correlation = q q
P 2 P 2

P 2
P
2
sets of data are strongly linked together [36]. It reaches its n x − x n y − y
maximum when the two signals are similar. It is calculated (17)

20604 VOLUME 6, 2018

M. Elhoseny et al.: Secure Medical Data Transmission Model for IoT-Based Healthcare Systems

FIGURE 8. Histogram of the cover image before and after applying the proposed model on gray-scale images with different text sizes (15, 30, 45 and
55 bytes).

where n is the number of pairs of data, X is the input image, ferent lengths and hiding them in both color and gry scale
and Y is the stego image. images. The hidden message was analyzed before being
transmitted and after being received by the expected recipient.
B. SECURITY ANALYSIS That is to guarantee less distortion occurs within to the origi-
In this study, the comparisons were conducted between the nal cover file after concealing the secret text. Our model was
original medical cover image and the stego-image. The pro- applied to both datasets; DME eyes dataset [37] that represent
posed model was tested using different messages with dif- the first three images, and DICOM dataset [38] that represent

VOLUME 6, 2018 20605

 M. Elhoseny et al.: Secure Medical Data Transmission Model for IoT-Based Healthcare Systems

FIGURE 9. Histogram of the cover image before and after applying the proposed model on gray-scale images with different
text sizes (100, 128 and 256 bytes).

the last two images. Fig. 5 shows the color and gray versions images as illustrated in table 1 and table 2 respectively.
of these dataset. The PSNR values were relatively reached to 57.44 and
In this model, the text is encrypted by using the 56.39 in case of DWT-2L for color images and gray-scale
evolved hybrid encryption scheme that was previously images respectively. While the PSNR values were relatively
explained. Then it is being embedded using either reached to 56.13 and 55.43 in case of DWT-1L for color
2D-DWT-1L or 2D-DWT-2L stenography techniques. It was images and gray-scale images, respectively. The MSE values
found that DWT-2L gives better PSNR and MSE results using DWT-2L varied from 0.12 to 0.57 for the color images
compared with DWT-1L in case of both color and gray scale and from 0.14 to 0.57 for the gray images. While the MSE

20606 VOLUME 6, 2018

M. Elhoseny et al.: Secure Medical Data Transmission Model for IoT-Based Healthcare Systems

TABLE 5. Comparing the performance of our proposed model with Anwar [3] A. K. Bairagi, R. Khondoker, and R. Islam, ‘‘An efficient stegano-
et al. approach graphic approach for protecting communication in the Internet of Things
(IoT) critical infrastructures,’’ Inf. Secur. J., Global Perspect., vol. 25,
nos. 4–6, pp. 197–212, 2016.
[4] A. S. Anwar, K. K. A. Ghany, and H. El Mahdy, ‘‘Improving the security
of images transmission,’’ Int. J. Bio-Med. Inform. e-Health, vol. 3, no. 4,
pp. 7–13, 2015.
[5] A. Abdelaziz, M. Elhoseny, A. S. Salama, and A. M. Riad, ‘‘A machine
learning model for improving healthcare services on cloud computing
environment,’’ Measurement, vol. 119, pp. 117–128, Apr. 2018. [Online].
Available: https://doi.org/10.1016/j.measurement.2018.01.022
[6] M. Paschou, E. Sakkopoulos, E. Sourla, and A. Tsakalidis, ‘‘Health Inter-
values by using DWT-1L were ranged from 0.16 to 3.43 for net of Things: Metrics and methods for efficient data transfer,’’ Simul.
the color images; and from 0.18 to 3.49 for the gray scale Model. Pract. Theory, vol. 34, pp. 186–199, May 2013.
[7] M. Sajjad et al., ‘‘Raspberry Pi assisted face recognition framework
images. for enhanced law-enforcement services in smart cities,’’ Future Gen-
There was no variation between the studied images in the erat. Comput. Syst., to be published, doi: https://doi.org/10.1016/j.future.
BER, where its values were zero for both images. The SSIM, 2017.11.013
[8] P. Kumar and H.-J. Lee, ‘‘Security issues in healthcare applications using
SC, and Correlation were almost equal to one with all of wireless medical sensor networks: A survey,’’ Sensors, vol. 12, no. 1,
the studied color and gray images as illustrated in table 3 pp. 55–91, 2012.
and table 4 respectively. Figures (6, 7, 8 and 9) show the [9] M. A. Razzaq, R. A. Shaikh, M. A. Baig, and A. A. Memon, ‘‘Digital
image security: Fusion of encryption, steganography and watermarking,’’
histograms of the cover image before and after applying Int. J. Adv. Comput. Sci. Appl., vol. 8, no. 5, pp. 224–228, 2017.
the proposed model on both color and gray scale images [10] N. Dey and V. Santhi, Intelligent Techniques in Signal Processing for Mul-
sequentially with different text sizes. timedia Security. New York, NY, USA: Springer, 2017, doi: 10.1007/978-
3-319-44790-2.
[11] M. Jain, R. C. Choudhary, and A. Kumar, ‘‘Secure medical image steganog-
C. COMPARING THE RESULTS AGAINST ANOTHER raphy with RSA cryptography using decision tree,’’ in Proc. 2nd Int. Conf.
Contemp. Comput. Inform. (IC3I), Dec. 2016, pp. 291–295.
APPROACH [12] L. Yehia, A. Khedr, and A. Darwish, ‘‘Hybrid security techniques for
The performance of our model was compared with another Internet of Things healthcare applications,’’ Adv. Internet Things, vol. 5,
pp. 21–25, Jul. 2015.
technique developed by Anwar et al. [4] on 256×256 pixel [13] Z. M. Zaw and S. W. Phyo, ‘‘Security enhancement system based on the
medical color image using 18-byte text size. Table 5 shows integration of cryptography and steganography,’’ Int. J. Comput., vol. 19,
the obtained PSNR and MSE values from applying our model no. 1, pp. 26–39, 2015.
[14] R. K. Gupta and P. Singh, ‘‘A new way to design and implementation of
as compared with those resulted by [4]. The results obtained hybrid crypto system for security of the information in public network,’’
after applying the models on 256∗256 color medical images Int. J. Emerg. Technol. Adv. Eng., vol. 3, no. 8, pp. 108–115, 2013.
with text size 18 bytes. It was found that our proposed model [15] S. A. Laskar and K. Hemachandran, ‘‘High capacity data hiding using LSB
steganography and encryption,’’ Int. J. Database Manage. Syst., vol. 4,
had a higher PSNR value and a smaller MSE value that no. 6, p. 57, 2012.
reveals the higher performance of our proposed model. [16] L. Yu, Z. Wang, and W. Wang, ‘‘The application of hybrid encryption
algorithm in software security,’’ in Proc. 4th Int. Conf. Comput. Intell.
Commun. Netw. (CICN), Nov. 2012, pp. 762–765.
V. CONCLUSION [17] S. F. Mare, M. Vladutiu, and L. Prodan, ‘‘Secret data communication
A secure patient’s diagnostic data transmission model using system using steganography, AES and RSA,’’ in Proc. IEEE 17th Int. Symp.
Design Technol. Electron. Packag. (SIITME), Oct. 2011, pp. 339–344.
both color and gray-scale images as a cover carrier for health- [18] A. K. Mandal, C. Parakash, and A. Tiwari, ‘‘Performance evaluation of
care based IoT environment has been proposed. The proposed cryptographic algorithms: DES and AES,’’ in Proc. IEEE Students’ Conf.
model engaged either 2D-DWT-1L or 2D-DWT-2L steganog- Elect., Electron. Comput. Sci. (SCEECS), Mar. 2012, pp. 1–5.
[19] S. F. Mjolsnes, Ed., A Multidisciplinary Introduction to Information Secu-
raphy and hybrid blending AES and RSA cryptographic tech- rity. Boca Raton, FL, USA: CRC Press, 2011.
niques. The experimental results were evaluated on both color [20] R. L. Rivest, A. Shamir, and L. Adleman, ‘‘A method for obtaining digital
and gray-scale images with different text sizes. The perfor- signatures and public-key cryptosystems,’’ Commun. ACM, vol. 21, no. 2,
pp. 120–126, 1978.
mance was assessed based on the six statistical parameters [21] M. S. Sreekutty and P. S. Baiju, ‘‘Security enhancement in image steganog-
(PSNR, MSE, BER, SSIM, SC, and correlation). Compared raphy for medical integrity verification system,’’ in Proc. Int. Conf. Circuit,
to the state-of-the-art methods, the proposed model proved Power Comput. Technol. (ICCPCT), Apr. 2017, pp. 1–5.
[22] A. Bashir, A. S. B. Hasan, and H. Almangush, ‘‘A new image encryp-
its ability to hide the confidential patient’s data into a trans- tion approach using the integration of a shifting technique and the AES
mitted cover image with high imperceptibility, capacity, and algorithm,’’ Int. J. Comput. Appl., vol. 42, no. 9, pp. 38–45, 2012.
minimal deterioration in the received stego-image. [23] K. Muhammad, J. Ahmad, H. Farman, Z. Jan, M. Sajjad, and
S. W. Baik, ‘‘A secure method for color image steganography using
gray-level modification and multi-level encryption,’’ TIIS, vol. 9, no. 5,
REFERENCES pp. 1938–1962, 2015.
[24] J. H. J. Yin, G. M. Fen, F. Mughal, and V. Iranmanesh, ‘‘Internet of Things:
[1] A. Darwish, A. E. Hassanien, M. Elhoseny, A. K. Sangaiah, and Securing data using image steganography,’’ in Proc. 3rd Int. Conf. Artif.
K. Muhammad, ‘‘The impact of the hybrid platform of Internet of Things Intell., Modelling Simulation (AIMS), Dec. 2015, pp. 310–314.
and cloud computing on healthcare systems: Opportunities, challenges, [25] S. A. Seyyedi, V. Sadau, and N. Ivanov, ‘‘A secure steganography method
and open problems,’’ J. Ambient Intell. Humanized Comput., to be pub- based on integer lifting wavelet transform,’’ IJ Netw. Secur., vol. 18, no. 1,
lished, doi: https://doi.org/10.1007/s12652-017-0659-1 pp. 124–132, 2016.
[2] A. Shehab et al., ‘‘Secure and robust fragile watermarking scheme [26] M. I. Khalil, ‘‘Medical image steganography: Study of medical image
for medical images,’’ IEEE Access, vol. 6, pp. 10269–10278, 2018, quality degradation when embedding data in the frequency domain,’’ Int.
doi: 10.1109/ACCESS.2018.2799240. J. Comput. Netw. Inf. Secur., vol. 9, no. 2, p. 22, 2017.

VOLUME 6, 2018 20607

 M. Elhoseny et al.: Secure Medical Data Transmission Model for IoT-Based Healthcare Systems

[27] H. Abdel-Nabi and A. Al-Haj, ‘‘Efficient joint encryption and data hiding OSAMA M. ABU-ELNASR received the
algorithm for medical images security,’’ in Proc. 8th Int. Conf. Inf. Com- M.S. degree in computer vision and the
mun. Syst. (ICICS), Apr. 2017, pp. 147–152. Ph.D. degree in empowering software testing
[28] L. Li, M. S. Hossain, A. A. A. El-Latif, and M. F. Alhamid, ‘‘Distor- process for improving software quality from
tion less secret image sharing scheme for Internet of Things system,’’ Mansoura University in 2009 and 2014, respec-
in Cluster Computing. New York, NY, USA: Springer, 2017, pp. 1–15, doi: tively.
https://doi.org/10.1007/s10586-017-1345-y
Since 2014, he has been a Lecturer with the
[29] M. Sajjad et al., ‘‘Mobile-cloud assisted framework for selective encryp-
Computer Science Department, College of Com-
tion of medical images with steganography for resource-constrained
devices,’’ Multimedia Tools Appl., vol. 76, no. 3, pp. 3519–3536, 2017. puters and Information, Mansoura University.
[30] S. A. Parah, J. A. Sheikh, F. Ahad, and G. M. Bhat, ‘‘High capacity and His research interests include information security,
secure electronic patient record (EPR) embedding in color images for IoT IoT, machine learning, and software quality management.
driven healthcare systems,’’ in Internet of Things and Big Data Analytics
Toward Next-Generation Intelligence. Cham, Switzerland: Springer, 2018,
pp. 409–437.
[31] N. F. Johnson and S. Jajodia, ‘‘Exploring steganography: Seeing the
unseen,’’ Computer, vol. 31, no. 2, pp. 26–34, Feb. 1998.
[32] F. A. Jassim. (2013). ‘‘A novel steganography algorithm for hiding
text in image using five modulus method.’’ [Online]. Available: https://
arxiv.org/abs/1307.0642
[33] Wikipedia Contributors. Mean Absolute Error. Accessed: Mar. 1, 2018. SHIHAB A. SHAWKAT received the B.Sc. degree
[Online]. Available: https://en.wikipedia.org/wiki/Mean_absolute_error in computer Science from the University of Tikrit
[34] Z. Wang, A. C. Bovik, H. R. Sheikh, and E. P. Simoncelli, ‘‘Image quality in 2009 and the M.Sc. degree in computer science
assessment: From error visibility to structural similarity,’’ IEEE Trans. from Mansoura University in 2017.
Image Process., vol. 13, no. 4, pp. 600–612, Apr. 2004. He is currently an Assistant Teacher with the
[35] C. S. Varnan, A. Jagan, J. Kaur, D. Jyoti, and D. S. Rao, ‘‘Image quality Directorate of Education in Salah Al-Din, Ministry
assessment techniques in spatial domain,’’ Int. J. Comput. Sci. Technol.,
of Education, Iraq. His research interest lies
vol. 2, no. 3, pp. 177–184, 2011.
in image processing and computer security and
[36] E. A. Silva, K. Panetta, and S. S. Agaian, ‘‘Quantifying image similarity
using measure of enhancement by entropy,’’ Proc. SPIE, Mobile Multi- media digital files like steganography and cryptog-
media/Image Process. Military Secur. Appl., vol. 6579, pp. 65790U-1– raphy.
65790U-12, Jan. 2007.
[37] H. Rabbani, M. J. Allingham, P. S. Mettu, S. W. Cousins, and S. Farsiu,
‘‘Fully automatic segmentation of fluorescein leakage in subjects with
diabetic macular edema,’’ Investigative Ophthalmol. Vis. Sci., vol. 56, no. 3,
pp. 1482–1492, 2015.
[38] F. J. McEvoy and E. Svalastoga, ‘‘Security of patient and study data asso-
ciated with DICOM images when transferred using compact disc media,’’
J. Digit. Imag., vol. 22, no. 1, pp. 65–70, 2009.
ARUNKUMAR N received the B.E., M.E., and
Ph.D. degrees in electronics and communica-
tion engineering with specialization in biomedical
engineering. He has a strong academic teaching
MOHAMED ELHOSENY received the Ph.D. and research experience of more than 10 years
degree in computer and information sciences in SASTRA University, India. His research interest
from Mansoura University, Egypt (in a sci- areas include image processing, biomedical engi-
entific research channel with the Department neering, and information security.
of Computer Science and Engineering, Univer-
sity of North Texas, USA). He is currently an
Assistant Professor with the Faculty of Com-
puters and Information, Mansoura University.
He has authored/co-authored over 65 International
Journal articles, Conference Proceedings, Book
Chapters, and two Springer brief book. His research interests include network
security, cryptography, machine learning techniques, Internet of Things, and
quantum computing. He is a TPC Member or Reviewer in over 30 Inter-
national Conferences and Workshops. Furthermore, he has been reviewing AHMED FAROUK received the M.S. and
papers for over 20 International Journals. His Ph.D. thesis was awarded the Ph.D. degrees from Mansoura University in 2009
best Ph.D. thesis prize (2016) at Mansoura University. and 2015, respectively. He is especially well
known for seminal contributions to theories of
quantum mechanics, communication and cryp-
tography. He has published 50 papers in reputed
and high impact journals like Nature Scientific
GUSTAVO RAMÍREZ-GONZÁLEZ is currently Reports, Physical Review A, and 14 book chapters.
a Professor with the Department of Telematics His research interests include quantum communi-
Engineering, Universidad del Cauca, Colombia. cation, quantum cryptography, quantum machine
His research interests include image process- learning and quantum information processing. He has worked for vari-
ing, secure communication, and machine learn- ous conferences at different levels from reviewer to Organizer-Chairman.
ing. He has published several research papers His continuous support for reviewing process lead to award 29 Certificates of
in reputed journals. He served as a Guest Editor appreciation for generous support and for invaluable efforts in the reviewing
for several special issues at many journals, such as process. He received the Best Researcher for 2015 and 2016 by the Al-Zahra
Computers and Electrical Engineering and Clus- College for Women, Oman.
ter Computing.

20608 VOLUME 6, 2018