A healthcare. A typical E-Healthcare system consists

A Fuzzy Logic based Bit Plane
Complexity Segmentation steganography to secure Electronic Health Records

Abstract

            Electronic
Healthcare or E-Healthcare (EH) is a paperless management promises to speed up
typical bureaucracy of healthcare. A typical E-Healthcare system consists of
many component and subsystem, such as appointment, routine clinical notes,
picture archiving, lab and radiology orders, etc., are vulnerable to security
threats. Cryptology and steganography are generally used to ensure medical data
security. For this reason, a Fuzzy Logic based Bit Plane Complexity
Segmentation (FL-BPCS) steganography is combined with AES cryptography and
Huffman lossless compression is proposed to secure patient data’s. In this,
Electroencephalogram time series, doctor’s comment and patient information are
selected as hidden data and Magnetic Resonance image are used as cover image.   

Keywords: E-Healthcare,
steganography, FL-BPCS, Electroencephalogram, cryptology

I.                  
Introduction

The
privacy and security of EH information falls into two categories. First, inappropriare
releases from authorized users who intentionally or unintentionally access or
disseminare information violation of EH system policies or EH computer system
can be break by outsiders. Second, open disclosure of patient health
information to parties against to the interests of specific individual patient
or invadere a patient’s privacy. These falls arises from the flow of data
across the EH system amongst and between providers, payers and secondary users,
with or without the patient’s knowledge are conceptually quite differ or
require different counteracts and interventions.

Technical
obstaculum of the intruders includes the use of firewalls to isolate internal networks
with strong encryption based authentication and authorization. There is no
known obstaculum for external networks like Denial-of-Service. Nevertheless,
technical countermeasures cannot be cures all security threats. Obstaculum such
as encryption, steganography and authentication are the only effective ways to
counter EH security threats against internet interface.       

     Steganography
is used for covert communication. The embedding algorithm will convert the
cover medium into stego medium by embedding secret data into it. The inverse
process of embedding is done to extract the secret data. Imperceptibility,
security, capacity, robustness, embedding complexity are the steganography
factors that has to be considered. Image steganography is developed according
to its use in medical fields to communicate between patient as well as
communication between doctor’s and laboratory people’s to hide secret messages.
Steganography is to avoid drawing attention to the transmission of hidden
medical information. If suspicion is raised, steganography and cryptography is
planned to achieve the security of secret medical data’s. Both are
complementary to each other and provide better security, confidentiality and
authenticity.

Eiji
Kawaguchi et al. proposed a Bit Plane Complexity Segmentation (BPCS) to
increase the embedding capacity and also to overcome the short comes of
traditional steganography techniques such as Least Significant Bit (LSB),
Transform embedding, Perceptual masking techniques. But the issue is embedded
secret data can be retrieved by using Difference Image Histogram (DIH).

Karakis
et al. proposed a Fuzzy Logic based Least Significant Bit (FL-LSB) to reduce
the insecurity of LSB planes. By using fuzzy the LSB planes were chosen to
embed the secret data’s. But the issue is low embedding capacity and the stego
image is invalid. This technique can be offered by using the following
sections. In section 2, the brief description of BPCS, cryptology as well as lossless
compression technique are presented, in section 3, the work methodology is
presented, in section 4, results are presented, in section 5, the discussion
and analysis are done, and finally the work is concluded.

II.               
Brief
description of BPCS, Cryptology and lossless compression technique

Bit
Plane Complexity Segmentation for embed and extract

            This
steganography method makes use of the human vision. The cover image is divided
into informative region and noise-like region and the secret data is hidden in
noise blocks of vessel image without degrading image quality. The data is
hidden in both Most Significant Bit (MSB) as well as LSB planes.

 

Cryptology
with security concern

            Cryptology
uses the Advanced Encryption Standard (AES) to encrypt the secret data to form
second security layer. It comprises of a series of linked operations, some of
which involve replacing the inputs by specific outputs (Substitution) and
others involve shuffling bits around (Permutations). It treats the 128 bits of
a plaintext block as 16 bytes as a matrix. It has built-in flexibility of key
length, which allows a degree of future-proofing against progress in the
ability to perform exhaustive-key searches.

Lossless
compression technique to reduce the size

            Lossless
compression technique uses the Huffman to reduce the secret data size. This
method takes a symbol (bytes) and encodes them with variable length codes
according to the statistical probabilities. A frequently used symbol will be
encoded with couple of bits, while symbol that are rarely used will be encoded
with more bits.

III.            
Proposed
work

The
proposed system aims to use EEG signals and MR images that are obtained from
same patients and to embed more data with EEG into MR images of same patient.
For this reason, the MR images and EEG of epilepsy patients are collected from
the Department of Neurology at Bonn University. 12 females and 11 males were
included (age: 18-65 years; mean age: 35 ± 7.7 years).

 

The
embedded message was combined with the patient’s information, doctor’s
comments, and EEG file header information and segmented EEG data. The patient’s
information (patient name, patient ID, patient birth date, patient gender,
patient age, patient weight, patient address, study date, study time, study ID,
study modality, study description, series date, series time, and series
description) were separately selected from the meta-header of each of the DICOM
files. An EEG was also recorded from 21 multiple electrodes that are placed on
the scalp using an International 10-20 system. EEG records take about 20-40
minutes for the diagnosis or treatment. EEG file headers have information on
record such as time points, number of electrodes, sampling interval, starting
time, and name of electrodes. In Table 1, EEG data is stored using short data
structure as 4 byte (The Capacity of EEG Data=The Number of Electrodes*Time
Points of EEG*4). Furthermore, MATLAB were used to code the methods and
analysis.

 

Least
Significant Bit (LSB) and Most Significant Bit (MSB) embedding is a simple and
fast strategy in steganography. It has high imperceptibility and embedding
capability. Hence, this study proposes new methods to modify LSB and MSB
embedding using medical data. The analyses consist of two stages: embedding and
extracting, respectively. Initially, the patient’s information is obtained from
DICOM series of epilepsy patient. The EEG data is segmented according to the
size of these DICOM images.

 

An
image is sampled by pixels. In the gray-scale image, pixels have gray level
intensities. In color images, pixels are also represented by three component
intensities, being red (R), green (G), and blue(B). A similarity measure is the
similarity degree between two groups or between two objects. In image
processing, the similarity measure of two pixels is used with distance
information in Euclidean color space. Demirci proposed a similarity-based
method for edge detection. Furthermore, Pixel-Value Differencing (PVD) or
Adjacent Pixel Difference (APD) methods determine embedding pixels in
histogram-based steganography. These methods have high embedding capacity and
PSNR values.

 

The
main idea of this method is to generate a new image whose pixels have double
values at the interval 0 1. This similar image of cover image is used to
determine pixels for the embedding message. In this method, if the values of
similar pixels are higher than the determined threshold (Th) by trial and
error, they are selected to hide the message. The neighboring pixels of the
image (P1, P2, …, P9 ) using the 3×3 window have three color component (R, G,
B).

 

The gray level differences of color components are calculated
the neighboring pixels of the stego image. The color distance of pixels are
calculated by the Euclidean norm. Thesimilarity values of pixels are founded.
Similarly, the coordinates of the pixels are determined between the similarity
values of pixels and threshold values. The hidden message is extracted using
the coordinates of the stego image’s pixels.

 

 

 

 

 

 

Fig. : Data Flow
Diagram for overall process

 

In message pre-processing stage, lossless
compression methods, which are LZW (Lempel–Ziv–Welch) and Huffman Compression,
are used to increase message capacity. Furthermore, LZW and Huffman Compression
methods also ensure the complexity of the message. To increase security, the
compression message is encrypted by the Rijndael symmetric encryption algorithm
using a 128-bit key. Secondly, the proposed methods, which are
fuzzy-logic-based Bit Plane Complexity Segmentation (FL-BPCS), select MSBs and
LSBs of image pixels with using the differences in gray levels of the pixels.
Finally, the selected MSBs and LSBs of the pixels are altered with the message
bits in stego images. These processes are simultaneously run with all DICOM
series to decrease computational time.

 

The
extracting message stage requires stego-DICOM images and a stego-key, which is
the authentication key for decryption. Firstly, the proposed methods give the
pixels coordinates, which have an embedded message. These pixels are used to
gather the message. Secondly, the obtained message is decrypted and
decompressed. Finally, the patient’s information, segmented EEG, and the
doctor’s comments are displayed in a GUI (Graphical User Interface) screen. All
hidden EEG data can be also gathered from the DICOM series. The comparison
results of the proposed algorithm are evaluated by PSNR (peak signal-to-noise
ratio), MSE (mean square of error), SSIM (structural similarity measure),
between the cover, and the stego-DICOM series.

 

IV.            
Results

The proposed system parameters
are used to evaluate the performance of the data
hiding techniques.

Peak Signal to Noise Ratio
(PSNR): The PSNR is generally used to measure the quality of stego image
in decibels (dB). Eq.1, gives
the expression for PSNR in which ICmax is the maximum pixel
value of the cover image and MSE is the mean square error:

(1)

Where:

(2)

In Eq. 2, x and y are the image
coordinates, M and N are the dimensions of the image, ISxy is
the generated stego-image and ICxy is the cover image.

Structural similarity (SSIM)
index: The SSIM is a method for finding the similarity
between cover image and the stego image. It is a perception-based model that
considers image degradation as perceived change in structural information. The
SSIM measure between two images IC and IS is
represented in Eq. 3, where,  is
the average of IC,  is
the average of  is
the variance of  is
the variance of  is
the covariance between IC and IS and k1,
k2 are two the variables used to stabilize the division with
weak denominator.

(3)

 

Fig. : Cover image and stego image

       

Fig. : Histogram of Cover image and
stego image

V.               
Discussion
and Analysis

The
graph is created based on the embedding capacity. In Fuzzy Logic based Least
Significant Bit the embedding capacity is mentioned as low, because to embed
the secret data it occupies only the LSB positions. In Bit Plane Complexity
Segmentation the embedding capacity is mentioned as low, when compared to the
Fuzzy Logic based Least Significant Bit but normally the embedding capacity is
high when compared to the primitive LSB because to embed the secret data it
occupies both MSB as well as LSB. In Fuzzy Logic based Bit Plane Complexity
Segmentation the embedding capacity is high because to embed the secret data
the red, green, blue channel were used with fuzzy.

Fig. : Embedding Capacity

The
graph is created based on the performance evaluation parameters PSNR, MSE,
SSIM. In Fuzzy Logic based Least Significant Bit the PSNR is achieved high,
because to embed the secret data it occupies only the LSB positions and MSE is
achieved low when compared to Bit Plane Complexity Segmentation and SSIM is
achieved high. In Bit Plane Complexity Segmentation the PSNR is achieved as low
because to embed the secret data it occupies both MSB as well as LSB and MSE is
achieved high and SSIM is low because MSB is more sensitive. While changing
that pixel values with another pixel values must be matched. In Fuzzy Logic
based Bit Plane Complexity Segmentation the PSNR is low because to embed the
secret data MSB as well as LSB is used. MSE is low and SSIM is high because it
uses red, green, blue channels with fuzzy rules.

Fig. : Performance Parameters

The
graph is created based on the steganalysis. Visual attacks involve observing
the unusual patterns and noisy blurred regions in some places of the stego
image. A statistical method
called RS steganalysis for detection of LSB embedding uses dual statistics
derived from spatial correlation of an image. Histogram based steganalysis techniques detect the existence of
secret data from smoothness of the stego image histogram. Similarly, a targeted
active steganalysis technique is implemented for HS embedding using the change
in the characteristics of histogram during data embedding.  In
Fuzzy Logic based Least Significant Bit the visual attacks, RS statistical
attack, Sample Pair Analysis is achieved 
high, because to embed the secret data it occupies only the LSB
positions but Difference Image Histogram is achieved low. In Bit Plane
Complexity Segmentation the visual attacks, RS statistical attack, Sample Pair
Analysis is achieved low, because to embed the secret data it occupies both MSB
as well as LSB but Difference Image Histogram is achieved high. In Fuzzy Logic
based Bit Plane Complexity Segmentation the visual attacks, RS statistical
attack, Sample Pair Analysis is achieved low, because to embed the secret data
the red, green, blue channel were used with fuzzy rules but Difference Image
Histogram is achieved high.

Fig. : Detection of Embedded
Data’s

Conclusion

In medical information system, medical data is
easily captured when being storing, receiving or transmission through computer
network and Internet. Cryptology and steganography are generally used to ensure
medical data security. For this reason, this study proposes new algorithm,
Fuzzy Logic-based Bit Plane Complexity Segmentation (FL-BPCS) to secure medical
data. EEG signals and MR images of epilepsy patients are used to combine
multiple medical signals into one file format. The embedding messages are
composed of EEG signals, doctor’s comment, and patient information in file
header of DICOM images. The messages are secured by using Huffman lossless
compression methods and Rijndael symmetric algorithm with 128 bit-key to
prevent the attacks.

The capacity of proposed algorithm is higher than
the result of similar studies in literature. According to the obtained result,
the proposed method ensures the confidentiality of the patient’s information.
The FL-BPCS method hides EEG signals, patient’s information and doctor’s
comment in the pixels of MR images. It also reduces data repository and
transmission capacity of the patients’ multiple medical data. In future, the
embedding and extracting of medical data from an cloud storage with multiple
records can be implemented for flexibility to access the patient data for diagnosis.
To access the patient data the grant privilege is given to make read and write
permission for authorized patient’s as well as doctor’s, nurse with laboratory
people’s.