Protection Of Electronic Health Records Dr Sujatha B K1
Mamtha Mohan1
Professor(TCE Department)
Assistant professor (ECE Department)
Ph no.9448963245 1
Ph no.8050000066
M S Ramaiah Institute of Technology, Bangalore. E-mail:
[email protected] Affiliation:Jain University
ABSTRACT— A reliable medical image management must
I. INTRODUCTION
ensure proper safeguarding of the Electronic health records.
This era has led to major technological innovation, internet
Safeguarding the medical information of the patients is a
being the forehand in it. The advent of internet being deployed
major concern in all hospitals. Digital watermarking is a
for many applications has made tremendous progress in wide
technique popularly used to protect the confidentiality of
spectrum of fields like medicine, health records, diagnosis etc.
medical information and maintaining them which enhances
Fast and secure access to patient’s records helps to save lives
patient health awareness. In this paper we propose a blend
with timely treatment in emergency situations. Therefore,
of Discrete Wavelet Transform (DWT) and Singular Value
anywhere anytime accessible online health-care or medical
Decomposition (SVD) for watermarking the EHR. The EHR is
systems play a vital role in daily life. It taps down this
further encrypted using key based encryption method for
advantage of the need for wide scale protected accessibility of
access control. The SVD is applied to the approximate and
health records to provide a safe and efficient way of sharing
vertical coefficients of the wavelet transform. The technique
and accessing the patient’s records. The EHRs(electronic
improves EHR protection and facilitates in accounting for
health records) is a methodical way of maintaining, securing
performance parameters like peak signal to noise ratio
and protecting health information of patients. It facilitates high
(PSNR) and Mean Square Error (MSE).Additionally the
quality patient treatment and awareness. EHR contains case
properties of
like
study, diagnosis, blood reports, medication, reports about
Normalized Cross-correlation(NC) ,wavelet energies and
various other conditions and others necessary undergoing
homogeneity between the pixel pairs of host and the encrypted
treatment. Moving to electronic health records is important to
watermarked EHR are exploited.
the modernization and revamping of our healthcare system,
Gray
Level
Co-occurrence
Matrix
Keywords— Discrete Wavelet Transform, Singular
but it poses great challenges in the areas of security, safety and
Value Decomposition, Watermarking, Peak signal to
privacy of patients records. Computerized medical records are
noise ratio, Mean Square Error, Normalized Cross
prone to potential abuses and threats. In the last few years,
Correlation, Electronic Health Record (EHR), Gray
thousands of human fraternity have faced compromises of
Level Co-occurrence Matrix(GLCM).
their health information due to the advent of security lapses at hospitals, insurance companies and government agencies. Various commercial companies make a living,buying and
selling doctor’s prescribing habits to the pharmaceutical
The public key encryption is employed to encrypt the records
companies.
and decrypted using cipher image. This is mainly used to
Additionally sensitive electronic data, especially when stored
enhance the security. It is an asymmetric cryptographic
by a third party, is vulnerable to blind subpoena or change in
protocol which is mainly based on the public key. The
user agreements. The main question here is how to provide a
watermarked EHR is encrypted using the public key and all
secure way of accessing and sharing of the medical records.
the users are shared a private key. The user trying to access the
Since the internet is prone to snooping by intruders we need to
EHR decrypts it using his private key. The strength lies in it
ensure the security of the medical records. Many solutions
being impossible for a properly generated private key to be
have been provided for electronic health records. The patient’s
determined from its corresponding public key. Thus the public
privacy is assured by access control, which verifies the
key may be published without compromising security,
person’s access permission in order to ensure security.
whereas the private key must not be revealed to anyone not
Encryption is also proposed along with restricting access .But
authorized
if the server only holds the decryption key that would be
signatures. Embedded encrypted patient data is stored in
catastrophic. So we propose a design which we refer to as an
central server of a health care provider and shared over the
optimal watermarking technique as a solution to secure and
internet which allows anywhere and anytime access of health
private storage of patient’s medical records. The optimal
records at ease. The electronic health records are secured
watermarking technique is based on SVD and DWT domain
through authenticated access, verifying the identity of the user
for gray-scale images .This has formed from the performance
and validating their access permissions included for their
parameters namely peak signal to noise ratio (PSNR), Mean
access. The proposed approach produces watermarked
Square Error (MSE). PSNR is the ratio between the maximum
encrypted patient data which protects their copyrights and
possible
of
avoids modifications. This further strengthens the security and
noise that
protects the patient’s records from any sort of compromise and
power
of
a
signal
and
corrupting http://en.wikipedia.org/wiki/Noise
the
power
to
read
messages
or
perform
digital
affects the fidelity of its representation. The high spreading of
threats.
broadband networks and new developments in digital
This Paper is organized as follows: Section 2 discusses about
technology has made ownership protection and authentication
the related work, Section 3 deals with the proposed
of digital data since it makes possible to identify the author of
methodology involving the algorithms for watermarking and
an image by embedding secret information to the host image.
extraction as well as the performance parameters considered.
The Properties of Gray Level Co-occurrence Matrix like
A discussion of the experimental results is done in Section 4.
Normalized Cross-correlation(NC) ,wavelet energies and
Section 5 discusses the conclusion and the future work.
homogeinity between the pixel pairs of host and the encrypted
II. RELATED WORK
watermarked EHR are exploited for enhancement of the EHR security.
There are a number of potential applications under the
The algorithm embeds the watermark by modifying the
Umbrella of privacy-preserving data sharing and processing.
singular values of the host records. It is followed by singular
There has been considerable research at de-identification of
value decomposition and packing of values and encrypting.
medical record information[1] and de-identification of clinical
Encryption schemes with strong security properties will
records[2].Various other attempts on de-identification of visit
guarantee that the patient's privacy is protected (assuming that
records have been done [3],But these do not include the entire
the patient stores the key safely).
medical records.
Giakoumaki et al proposed a wavelet transform-based
yielding better performance. Haar wavelet is a sequence of
watermarking, the drawback is that it is related only to
rescaled "square-shaped" functions which together form
medical images and not the entire records and also medical
a family or basis. The Haar wavelet's mother wavelet
images are overwritten which may be unacceptable in
function
can be described as
diagnosis. We plan to work on digital watermarking, which would help ensuring the privacy and security of digital media and safeguard the copyright, and hiding the ownership identification [4]. Watermarking is a process that embeds data
Its scaling function
can be described as
into a multimedia object to protect the ownership to the objects [5] Encryption is a solution to secure and private storage of patient’s medical records [8]. The hierarchical
2.Singular Value Decomposition (SVD)
encryption system partitions health records into a hierarchical structure, each portion of which is encrypted with a
Singular Value Decomposition is a matrix factorization
corresponding key. The patient is required to store a root
technique. The SVD of a host image is computed to obtain
secret key, from which a tree of sub keys is derived [6].
two orthogonal matrices U, V and a diagonal matrix the
III. METHODOLOGY Watermark embedding
process is performed on the new matrix S+kW to get Uw, Sw and Vw, where k is the scale factor that controls the strength of the watermark embedded to the original image. Then, the
Host Image
watermarked record Fw is obtained by multiplying the
Encrypted watermarked image
Watermarked image(DWT w +SVD)
matrices U, Sw, and VT . The steps of watermark embedding are summarized as follows:
Cover image(EHR)
Encrypted watermarked image
watermark W is added to the matrix S. Then, a new SVD
Recovered vv watermark Decrypted watermarked image
i. The SVD is performed on the original image (F matrix). F = USVT
IDWT+SVD dewatermark
(1)
ii. The watermark (W matrix) is added to the SVs of Reconstructed original image
Watermark extraction
the original image (S matrix). D = S + kW
(2)
iii. The SVD is performed on the D matrix. Fig1: Block diagram of the proposed methodology.
1.Discrete Wavelet Transform
D = UwSwVT
(3)
iv. The watermarked image (Fw matrix) is obtained using the
Wavelet transform has the capacity of multi-resolution
modified SVs (Sw matrix).(EHR)
analysis. Embedding of a watermark is made by modifications
Fw = USwVT
of the transform coefficients using haar wavelet. The inverse
v.Apply the key based encryption to the watermarked EHR.
transform is applied to obtain the original record. The host
The medical images considered are MRI images. Initially the
image is decomposed into four sub-bands namely LL, LH,
watermark is embedded in the image by setting an initial value
HL, and HH. A hybrid DWT-SVD based watermarking
of scaling factor α. Using gray level Co-occurrence matrix the
scheme which is further encrypted using key based encryption
properties like wavelet energies,homogeinity and cross
method is developed that requires less computation effort
correlation of the medical document(pixel pairs) are measured
(4)
The original record is reconstructed from the encrypted EHR by extraction of the watermark. Optimum value of scaling
(6)
factor is found by iteration of the above and tabulating the
Here, MAXI is the maximum possible pixel value of the
results to obtain desirable values of PSNR, MSE, NC,wavelet
image=255.
energies and homogeinity.
MSE is defined as
3.1 Algorithm to embed watermark into cover image (7)
Steps: i. Read the cover image & watermark EHR.
Cross-correlation is a measure of similarity of two waveform
ii.Apply DWT to cover image to obtain approximation,
as a function of a time-lag applied to one of them.
horizontal, vertical, diagonal DWT coefficients(LL, HL, LH,
For continuous functions ‘f’ and ‘g’, the cross-correlation is
HH) Calculate the approximate DWT coefficient by adding
defined as
the watermark record using Cal(i,j)=ca1(i,j)+(α*watermark)
dt
(5)
(8)
iii. Where Ca1 & ca1 are the modified & original
Where f* denotes the complex conjugate of ‘f’ and‘ ’ is the
approximation coefficients and α is a scaling value as set to10.
time lag.
iv. Apply SVD to the decomposed sub-bands(LL) and (HL)
3.Grayscale co-occurrence matrix(GLCM)
and Encrypt the decomposed record.
GLCM is an m x n x p array of valid gray-level co-occurrence
v.Find the Inverse DWT and decrypt the watermarked record.
matrices.graycoprops normalizes the gray-level co-occurrence
vi. Increment α, apply SVD and inverse DWT.
matrix (GLCM) so that the sum of its elements is equal to 1.
3.2 Algorithm for Watermark Extraction
The energy of each sub bands of the EHR is calculated as
i. Extract the watermark from vertical & approximation DWT coefficient as per the equation WN= (SN-S)/ α; where α=10
‐dimensional DWT, to obtain the first level
(9)
ii. Apply two
Homogenity measures the closeness of the distribution of
decomposition of the watermarked image. LL1, HL1, LH1,
elements in the GLCM to the GLCM diagonal.
HH1 iii.Decrypt the watermarked record and calculate the
(10)
performance parameters.
3.4 Key based Encryption
3.3 Performance parameters
The public key encryption is employed to encrypt the records
The peak signal-to-noise ratio (PSNR) is used to measure the
and decrypt using cipher image. This is mainly used to
quality of reconstructed records. PSNR is the ratio between
enhance the security. It is an asymmetric cryptographic
the maximum possible power of a signal and the power of
protocol which is mainly based on the public key. The
corrupting noise that affects the fidelity of its representation.
watermarked EHR is encrypted using the public key and all
Because many signals have a very wide dynamic range, PSNR
the users are shared a private key. The user trying to access the
is usually expressed in terms of the logarithmic decible scale.
EHR decrypts it using his private key. It is asymmetric
PSNR is most easily defined via the mean squared error
cryptography in terms that where a key used by one party to
(MSE). Given a noise-free m×n monochrome image I and its
perform either encryption or decryption is not the same as the
noisy approximation K.
key used by another in the counterpart operation.
The PSNR is defined as
IV.RESULTS
The medical images considered are MRI obtained from the
range of found to give unrealistic PSNR values of 0 or
database “MRI Images”.
infinity. Hence 62 to 80 dB while the MSE is in the range of
Fig 2(a): watermark embedding
Recovered_ watermark
Watermark+de crypted image
Reconstructed
Original
Watermarked
Watermark
Cover image
image
Image
Image
+encrypted
+watermarked
image
image
MRI Image
MRI Image
0.0015 to 0.048. With increasing values of scaling factor. Fig 2(b): watermark extraction
Wavelet energies of the sub bands ranges from 0.4949 to 0.5461 and homogeneity has a decreasing gradient from
Table 1: performance parameters
0.9870 to 0.8147 portrayed in TABLE 2.The watermarked EHR is encrypted using the public key and all the users are
Scaling
PSNR
factor
in dB
MSE
NC
shared a private key as shown in fig 2(a). The user trying to access the EHR decrypts it using his private key. This is portrayed in fig 2(b).
8
80.38
0.00059
0.0213
9
74.984
0.0021
0.0233
V.CONCLUSION AND FUTURE WORK
10
71.08
0.0051
0.0253
The watermarking can be used to provide proof of the
11
67.06
0.0104
0.0264
12
65.07
0.0175
0.0273
right person. The watermark has been inserted without
14
62.825
0.0339
0.0260
interfering with the documents usefulness. In this paper we
authenticity of EHR that is to say that the medical information of one patient has been issued from the right source and to the
have used DWT+SVD techniques to calculate PSNR, MSE, Table 2.subbands parameters
NC, wavelet energies and the homogeneity which are the properties of gray level co-occurrence matrix of different
Subbands
Energy
Homogeneity
LL
0.4949
0.9870
HL
0.5421
0.8628
LH
0.5137
0.8243
paper contributes in utilizing SVD generous properties along
HH
0.5461
0.8147
with hiding, protecting and safeguarding EHR which is an
medical records.SVD is very aptly used with DWT. It has been tested for different images and PSNR, MSE, NC are calculated for each image.Matlab R2013a has been used. This
asset to all the medical fraternity and the most important the The images have been watermarked using EHR and encrypted.
patient’s awareness regarding their medical health. This paper
On experimentation the scaling factors below 8 and above 14
also introduces new trends and challenges in using SVD and
were the scaling factor range is fixed in the range of 8 to 14.
key based encryption in image processing applications. This
From TABLE 1 it can be observed that PSNR values are in the
paper opens many tracks for future work in using SVD as an
enhances
37, Part 6, December 2012, pp. 723–729._c Indian
confidentiality, security and authenticity of medical health
Academy of Science Conference on Artificial Intelligence in
imperative
tool
in
signal
processing
and
Computer Science ,Malaysia (2013): 147-156.
records by encrypting the medical records using public key based encryption that assures confidentiality and authenticity.
[9]
and efficient health data management through multiple
REFERENCES [1]
watermarking on medical images.” Med Bio Eng Comput
Burr T, Klamann R, Michalak S, Picard R,Generation of Synthetic BioSense Data. Los Alamos National Laboratory
44 (2006)): 619-631. [10]
Report LA-UR-05-7841 2005. [2]
Proceedings in Computer
Clifford G, Automated de-identification of free-text
[12]
and Its Applications,
4689 international journal of
scientific progress and research (ijspr) volume - 03, number - 02, 2014. [5]
O P Singh, Satish Kumar, G R Mishra, Charu Pandey, Vartika Singh, Study of Watermarking International Journal of Scientific and
Research Publications, Volume
2, Issue 10, October 2012 1 ISSN 2250-3153 [6]
Josh Benaloh, Melissa Chase, Eric Horvitz, and Kristin Lauter, Patient Controlled Encryption: Ensuring Privacy of Electronic Medical Records, CCSW’09, November 13, 2009, Chicago, Illinois, USA.
[7]
Li Weng, Member, IEEE, Laurent Amsaleg, April Morton, and Stéphane Marchand-Maillet , A Privacy-Preserving Framework for Large-Scale Content-Based Information Retrieval, IEEE transactions on information forensics and security, vol. 10, no. 1,
[8]
january 2015
A LAVANYA∗ and V NATARAJAN, “Watermarking
patient data in encrypted medical images,” Sa¯ dhana¯ Vol.
Venkata kishore, Dr.T.Ramashri. “SVD based Robust
International
Inform Assoc 2009, 16(5):670-682.
Kumar, Study of Different Digital Watermarking Schemes
Akshay Kumar Gupta and Mehul S Raval. “A robust and
Image Watermarking using Particle Swarm Optimization,
method for the de-identification of health data. J Am Med
Manoranjan Kr Sinha, Dr. Rajesh Kumar Rai, Prof. G.
Information
replacement. Sadhana Vol. 37, Part 4 (2012): 425-440.
Roffey T, Bottomley J, A globally optimal k-anonymity
[4]
&
secure watermarking scheme based on singular values
El Emam K, Dankar FK, Issa R, Jonker E, Amyot D, Cogo E, Corriveau JP, Walker M, Chowdhury S, Vaillancourt R,
Science
Technology (CS & IT) (2011): 155-157. [11]
Decision Making 2008, 8:32. [3]
“SVD
UsingWavelet Transform, Computer Science Conference
Long W, Szolavits P, Moody G, Mark R,
medical records. BMC Biomedical Informatics and
S.Ramakrishnan, T.Gopalakrishnan, K.Balasamy.
Based Robust Digital Watermarking For Still Images
Neamatullah I, Douglass M, Lehman L, Reisner A, Villaroel M,
A. Giakoumaki, S. Pavlopoulos D, Koutsouris. “Secure
Journal
on
Electronic
&
Electrical
Engineering, Volume 05, and Number 7 (2009): 1-6 [13]
www.mathworks.com
[14]
www.ece.ncsu.edu/imaging/Archives/ImageDataBase