Human Genome in a Smart Card

Mete Akg

un, Bekir Erg

uner, A. Osman Bayrak and M. S¸amil Sa

gıro

glu

ITAK B

ILGEM UKEAE (National Research Institute of Electronics and Cryptology), 41470, Gebze, Kocaeli, Turkey

Keywords:

Human Genome, Smart card, Disease Risk Test.

Abstract:

Gene sequencing costs have fallen considerably with advancements in technology in the last 10 years. This

cost will be reduced even further in the following years. That means personal genomics will be very possible

in the near future. How and where genetic information is stored is the biggest problem for individuals. Fur-

thermore, privacy of genomic data has a great importance because it can be used to identify an individual and it

contains privacy-sensitive data. Therefore, privacy-preserving methods for the use of genomic data should be

developed. In this paper, we present a method for storing some parts of human genome needed for the disease

risk computation in a smart card. Our method uses variations that separate any genome from the reference

genome. It selects the variations corresponding to the exonic regions and ﬁlters them according to their variant

quality and genotype quality. We show that our method can reduce a single whole human genome to the size

small enough to be stored in a smart card without abusing the genomic privacy of individuals. Furthermore,

we also propose a simple system in which genomic smart cards are used to perform privacy preserving disease

risk test.

1 INTRODUCTION

Next Generation Sequencing technologies are paving

the way to individual genomics and personalized

medicine. In the near future, decline in the cost of

genome sequencing would break the bond between

doctor and patient. That means every person will want

to have their own genome without contacting doctors.

Storage and retrieval of personal genomes would ap-

pear to be a major problem. Gene sequences provide

information about the current health status of people

they belong to. In addition, they carry information

about the health problems that were experienced in

the past and may occur in the future. Therefore, the

privacy of gene sequences is of great importance.

High-throughput sequencing methodologies are

capable of producing large amount of sequencing

data. This data may occupy tens or even hundreds

of gigabytes of disk space. Although individual ge-

nomics is not common, the storage of these large

amount data is already a big problem for researchers

and clinicians. Genomic science shows the disposi-

tion of a person to any disease stems from variations

in its genome. Clinicians can use genomic variations

for diagnosis and preventive medicine. Genomic data

provides opportunities for substantial improvements

in diagnosis and preventive medicine. In particular, it

has been shown that an individuals predisposition to

a disease depends on genomic variations. There are

some companies such as 23andMe (23andme, 2013)

and Counsyl (Counsyl, 2013) that offer the calcula-

tion of genetic risk for some diseases.

Containing private and sensitive personal data

such as genetic predisposition to certain diseases, in-

dividual genomic data attracts broad interest of a large

variety of health-care stakeholders like pharmaceuti-

cal and private insurance companies, physicians and

on line direct-to-consumer service providers. To pro-

tect the privacy of his genomic data, an individual

might not want to disclose this sensitive data to other

parties. For example, while getting a risk assessment

service for a speciﬁc disease from a medical unit, a

patient should be able not to reveal any further infor-

mation apart from that speciﬁc disease to either the

medical unit or the physician. The only information

that the medical unit or the physician might get would

be the risk level. Thus, for accuracy of genetic pre-

disposition tests, protecting genomic data privacy of

individuals is crucial as well as using the correct and

complete data.

In this paper, we present a method for storing im-

portant parts of human genome needed for the dis-

ease risk computation in a smart card. Our method

uses variations that separate any genome from the ref-

erence genome. It selects the variations correspond-

ing to the exonic regions and ﬁlters them according to

310

Akgün M., Ergüner B., Bayrak A. and Sa

gıro

glu M..

Human Genome in a Smart Card.

DOI: 10.5220/0004799903100316

In Proceedings of the International Conference on Health Informatics (HEALTHINF-2014), pages 310-316

ISBN: 978-989-758-010-9

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

their variant quality and genotype quality. It indexes

the previously known variations by using by using

Single Nucleotide Polymorphism (dbSNP) database

(Sherry et al., 2001). Furthermore, our method deals

with the variations that are speciﬁc to the person. We

show that our method can reduce a single whole hu-

man genome to the size small enough to be stored in

a smart card. We also propose a system in which ge-

nomic smart cards are used to perform privacy pre-

serving disease risk test.

2 RELATED WORK

In (Reber and Perttunen, 1997), a method storing at

least a portion of human genome on on the machine-

readable storage medium was proposed. This method

proposes the storage of raw genome in a lossless man-

ner. However, the storage limits and the data process-

ing method are not speciﬁed.

In the literature, there are many solutions for pro-

tecting the privacy of genomic data used in genetic

tests. Troncoso-Pastoriza et al. (Troncoso-Pastoriza

et al., 2007) proposed a protocol for secure pat-

tern matching by evaluating automata in an oblivious

manner. The protocol is developed for secure DNA

matching and provides security in semi-honest set-

ting. The communication complexity is linear in the

size of input alphabet and the number of states of the

ﬁnite state machine.

Adida and Kohane (Adida and Kohane, 2006) pro-

posed a system GenePING for secure storage of large,

genome-sized datasets. GenePING is developed by

extending the PING (Riva et al., 2001; Simons et al.,

2005) personal health record system. The authors

claim that an attacker accessing to the raw GenePING

storage can not ﬁnd any relation between patient and

genomic data points.

Blanton and Aliasgari (Blanton and Aliasgari,

2010) proposed a solution for secure DNA searching

also using secure automata evaluation. Proposed pro-

tocol introduce improvements on (Troncoso-Pastoriza

et al., 2007) reducing communicating parties work.

This work proposes a secure outsourcing of computa-

tion protocol which uses external service provider and

modiﬁed multi-party protocol.

Jha et al. (Jha et al., 2008) presents privacy

protecting implementations on genomic computations

such as sequence comparisons and distance calcula-

tions using secure two-party communication proto-

col. For edit distance they developed three protocols.

First protocol uses Yao’s garbled circuits while sec-

ond combines garbled circuits with secure computa-

tion with shares. To overcome performance and scal-

ability issues they hybridized the ﬁrst two protocols

into the third.

Bruekers et al. (Bruekers et al., 2008) proposed

a solution, in semi-honest attacker model, for limited

DNA-based operations like identity, ancestor and pa-

ternity tests, based on Short Tandem Repeat. Pro-

vided solution is based on homomorphic encryption

and its complexity highly depends on the number of

errors to be tolerated.

Taking into account the fully-sequenced human

genome, Baldi et al. (Baldi et al., 2011) proposed

protocols, for paternity tests, personalized medicine

and genetic compatibility tests, based on private set

operations technique. The main aim of this work is to

provide privacy protecting mechanisms to individu-

als, who have their genomic data, for getting serviced

for genetic tests from authorized parties.

Canim et al. (Canim et al., 2012) proposed to

use cryptographic hardware for secure storage, share

and query of genomic data. As a tamper-proof hard-

ware, secure coprocessors are employed for process-

ing genomic data owned by health organizations e.g.

hospitals. Data reside, in encrypted form, in data

storage servers which are assumed to be untrusted.

Proposed solution only addresses the genomic data

owned by health organizations and use potentially ex-

pensive tamper-proof hardware. As the authors agree,

the proposed protocol cannot provide privacy in the

case where information is extracted from the query

results.

Ayday et al. (Ayday et al., 2012) proposed a pri-

vacy protecting method, for medical tests and person-

alized medicine using genomic data, based on homo-

morphic encryption. The authors evaluate that per-

sonal genomic data is quite sensitive to be left to in-

dividuals own and propose to store personal genomic

data, in encrypted format, in a storage and process-

ing unit which can be in the control of governments,

non-proﬁt organizations or private companies, such as

cloud storage service providers.

Focusing on disease risk tests, Ayday et al. (Ayday

et al., 2013) proposed a system to provide privacy-

protecting methods, based on homomorphic encryp-

tion, for genomic, clinical and environmental data

storage and process. Like (Ayday et al., 2012), this

work also proposes to use storage and processing unit

to store sensitive data in encrypted form and disease

risk tests are performed by authorized institutions us-

ing homomorphic encryption technique and secure in-

teger comparison.

HumanGenomeinaSmartCard

311

3 BACKGROUND

3.1 Smart Health Cards

Worldwide, some health care organizations imple-

mented smart card applications to store and track

heath records of patients. These applications have

some advantages over traditional paper-based or

computer-based systems:

• More security and privacy for patient data

• Less fraud in healthcare

• Secure transfer of medical records

• Secure access to medical records in case of emer-

gency

• Secure platform for implementing other health-

care applications

• Secure computation module

• Interoperability

• Low implementation cost

• Modular solution

In healthcare applications, the major issue is not

the conversion of traditional health records to elec-

tronic health records (EHRs). Security and the tech-

nology used raises additional problems for healthcare

applications. The format of electronic health records

should be standardized, readable and usable. Elec-

tronic health records should be kept secure, private,

shareable with healthcare stakeholders and easily ac-

cessible in case of emergency. Smart card technology

can meet these requirements of EHR-based health-

care applications. Furthermore, smart card technol-

ogy provides secure computation facility. Smart cards

can be used for secure calculation of some basic op-

erations. Thus, some basic analyses can be done on

smart cards by using electronic health records.

3.2 Variant Call Format (VCF)

Variant Call Format (VCF) (Danecek et al., 2011) is a

text ﬁle format storing gene sequence variations such

as SNPs, insertions, deletions and structural variants.

The major advantage of VCF is that only variations

are stored with annotations and positions on the ref-

erence genome. BCF (Binary VCF) format is binary

version of VCF format. The type of information kept

in each format is essentially the same. Supporting

both BCF and VCF ﬁle is just a matter of software

update. It does not concern the main goal of our study.

Variant Call Format has three parts. Meta-

information part contains lines beginning with ”##”

and giving the description of FORMAT, INFO and

FILTER entries. There are some optional lines such

as GT and GP in meta-information part. VCF ﬁle has

one header line beginning with ”#CHROM”. In VCF

ﬁle, each variation is represented in one tab-delimited

data line. A data line is called a VCF record. The ﬁrst

nine ﬁelds of the record are used to describe variants.

• CHROM: an identiﬁer from the reference

genome.

• POS: position in the reference genome.

• ID: A unique identiﬁer for variant (rs number). If

a variation is not pre-deﬁned, the missing value is

used

• REF: reference bases (A,C,G,T,N).

• ALT: comma separated list of alternate non-

reference alleles.

• QUAL: phred-scaled quality score for the asser-

tion made in ALT.

• FILTER: PASS if this position has passed all ﬁl-

ters. Otherwise, a semicolon-separated list of

codes for ﬁlters that fail.

• INFO: additional information. As with the INFO

ﬁeld, there are several common, reserved key-

words that are standards across the community

such as GQ (conditional genotype quality, en-

coded as a phred quality) and DP (read depth at

this position for this sample).

• FORMAT: colon-separated list of data subﬁelds.

4 PERSONAL GENOMIC SMART

CARDS

In this study, we propose usage of smart cards for se-

cure storage of personal genomes. Nowadays, smart

cards are used as an ID card in many countries. Fur-

thermore, there are smart health card applications in

many countries such as France (Smart Card Alliance,

2006a), Germany (Smart Card Alliance, 2006b) and

Taiwan (Smart Card Alliance, 2005). If personal

genomic data is considered as an electronic health

record, our proposal can be applied to existing appli-

cations easily.

The human genome consists of 3 billion base

pairs. Each pair can be represented with two bits.

That means the entire human genome occupies ap-

proximately 750 Megabytes of disk space. At least

99% of the human genome sequence is the same in

all people. Therefore, there is a difference of at most

1% between any human genome and the reference hu-

man genome. Any human genome can be represented

with variations from the reference genome. Typically,

HEALTHINF2014-InternationalConferenceonHealthInformatics

312

Figure 1: Variant Call Format (Image taken from (Wikipedia, 2013)).

these variations are stored in VCF format. The size

of VCF ﬁle is usually more than 1 GB that is too big

to store in a smart card. Therefore, we need an extra

method to represent human genome in a smart card.

4.1 Creating Index File

The Single Nucleotide Polymorphism Database

(Sherry et al., 2001) (dbSNP) is a free public archive

for genetic variation within and across different

species developed and hosted by the National Center

for Biotechnology Information (NCBI) in collabora-

tion with the National Human Genome Research In-

stitute (NHGRI). In this study, we work on the dbSNP

build 137 for Homo sapiens (dbsnp 137.b37.vcf). It

contains a range of molecular variation: (1) SNPs,

(2) short deletion and insertion polymorphisms (in-

dels/DIPs), (3) microsatellite markers or short tandem

repeats (STRs), (4) multinucleotide polymorphisms

(MNPs), (5) heterozygous sequences, and (6) named

variants.

Variations which alter the protein coding ex-

onic regions have the most direct effect on pheno-

typic properties. Most of the high penetrance ge-

netic diseases are associated with exonic variants.

Therefore we select the variations which are in ex-

onic regions, decreasing the amount of stored data

by 100 fold while keeping most of the genetic in-

formation. TruSeq Exome Targeted Regions BED

ﬁle (Illumina, ) is a ﬁle containing information on

the Targeted Exons in the TruSeq Exome product.

We get the intersection of dbsnp 137.b37.vcf and

TruSeq exome targeted regions.b37.bed ﬁles with

the help of the intersectBed utility of bedtools (Quin-

lan and Hall, 2010). In this way, we obtain the list

of predeﬁned variations corresponding to exonic re-

gions.

The primary goal of our study is to store maxi-

mum amount of functional information in minimum

amount of storage space. This is crucial for making it

cost effective to apply this method to common clini-

cal use. It is evident that protein coding regions con-

tain the most valuable functional information and the

weight of functional value of non-coding regions is

unclear. Moreover, whole exome sequencing meth-

ods are gaining popularity fast, so exonic variations

are becoming more accessible. We opt to keep only

exonic variations in smart cards. Still, it is possible

to add the variations in regions which have clinical

evidence for their functional value to the reference

database. This way it becomes possible to store the

most valuable information in a smart card.

4.2 Representing a Human Genome in a

Smart Card

In this study, we represent a human genome in

a smart card with variations from the reference

genome. As we mention above, we generate the

list of variations corresponding to the exonic re-

gions by intersecting the corresponding VCF ﬁle with

TruSeq exome targeted regions.b37.bed ﬁle. We sort

the list of variants according to ID ﬁeld.

We ﬁlter variations according to FILTER ﬁeld,

conditional genotype quality (GQ) ﬁeld and read

depth (DP) ﬁeld. The variation is selected, If FILTER

= PASS (meaning the variation has passed all ﬁlters)

and DP ≥ 8 and GQ ≥ 15.

In the previous subsection, we explain how we

create the index ﬁle. For each variation in the index

ﬁle, we reserve 1-bit ﬁeld from the storage of smart

card. If the VCF ﬁle that will be indexed includes the

corresponding variation, 1-bit ﬁeld is set to 1 other-

wise it is set to 0. If the number of variations in the in-

dex ﬁle is not the multiple of 8, we pad this data with

the required number of bits. We use Huffman Cod-

ing in order to decrease the size of data. We choose

Huffman Coding because a bit in any position can be

obtained without extracting all data.

HumanGenomeinaSmartCard

313

A human genome can include variations that are

not deﬁned in dbSNP. We store these variations in the

format shown in Table 1 and Table 2

Table 1: Variation Storage Format.

Type Genotype Position Information

2 bits 1 bit 32 bits

∗

Please see Table 2

Table 2: Allocation of Information Field.

Type The Number of Bits in

Information Field

SNP (0) 2 bits for reference base

Deletion (1) 16 bits for the size of

deletion

Insertion (2) (16 bits +

(SizeOfInsertion * 2 bits))

for the size of insertion

and DNA sequence

4.3 A Simple System for Disease Risk

Computation

In (Ayday et al., 2013), Ayday et al. proposed a secure

system for privacy preserving computation of the dis-

ease risk. In this system, all genomic data (SNP con-

tent) are stored in Secure Processing Unit (SPU). The

genetic test is performed by using homomorphic en-

cryption and privacy preserving integer comparison.

The effect of each SNP to a disease is expressed with

odds ratio (OR). The OR is the ratio of the proportion

of individuals in the case group having a speciﬁc ge-

netic variation to the proportion of individuals in the

control group having the same genetic variation. If

a disease is related with multiple variations, the dis-

ease risk is calculated by taking the weighted aver-

age of the OR of each related variations. Ayday et

al. used logistic regression model in Equation 1 for

disease risk computation.

ln(

1 − Pr

) = α +

∑

(1)

They calculate OR of SNP

as a OR

= exp(β

) where

is the regression coefﬁcient. The calculation over-

all genetic risk is shown in Equation 2.

ln(

1 − Pr

) = α +

∑

iεϕx

(X) (2)

In Equation 2, the effect of the SNP

to the overall ge-

netic risk is represented with p

and α is the intercept

of the model.

In this study, we propose a simple system in which

genomic smart cards can be used for privacy preserv-

ing risk computation. The proposed system is sum-

marized in Figure 2. In Step 1, an individual provides

his DNA sample to the certiﬁcated institution (CI) for

genome sequencing. CI generates a smart card con-

taining genomic data that is created with the method

explained in Section 4.1 and 4.2. In Step 2, CI sends

the genomic smart card to the individual. The individ-

ual wants to check his disease susceptibility for dis-

ease X. In Step 3, the individual provides his genomic

smart card to the medical unit (MU) for susceptibility

test. MU has access to indexed variant database (IVD)

that is generated as described in Section 4.1. IVD is

stored locally or is accessed over the remote server.

In Step 4, MU queries IVD for variants that poten-

tially cause the disease X. In Step 5, MU gets the in-

dexes of queried variants. In this point, MU sends in-

dexes of corresponding variants and contribution co-

efﬁcients of corresponding variants to the smart card.

The smart card checks whether the individual has the

corresponding variants and computes the disease risk.

Risk calculations can be done with the method used

in (Ayday et al., 2013).

Figure 2: Usage of Genomic Smart Cards.

In this system, we do not need homomorphic en-

cryption and privacy preserving integer comparison.

We know smart cards can be equipped with secure

crypto processors. These microprocessors are tamper-

resistant so they can be used to store and process pri-

vate or sensitive information. The information on

smart card is not accessible through external means

and can be accessed only by the embedded software

which should contain the appropriate security mea-

sures. In our system, an attacker or MU can not obtain

the contents of variants because disease risk compu-

tation is performed in secure crypto processor of the

smart card. Therefore, our method preserves the pri-

HEALTHINF2014-InternationalConferenceonHealthInformatics

314

vacy of genomic data relying on the security strength

of secure crypto processor.

As smart cards are tamper-proof devices, after

writing the genomic data, it will be marked as secret

and will not be readable from out-side the card. That

means, like secret and private keys, genomic data will

not be disclosed, at any time of the card’s lifetime,

by the card once it is written. Only some functions

which will be executed by card’s microprocessor, will

be available for computing disease risks. These func-

tions will take SNP IDs and their risk factors as pa-

rameters and will return a disease risk e.g. as a per-

centage or score. To prevent queries without consent

of the card holder, at every call to these functions,

card will ask its owner to enter his/her PIN. By this

method, even two consecutive queries will require the

card holder to enter his/her PIN twice, therefore a

function call can not be made without holder’s per-

mission.

5 EXPERIMENTAL RESULTS

We implemented our method in C language. In our

experiment, we use six whole genomes VCF ﬁles.

These VCF ﬁles are obtained from the whole genome

sequencing that are done by Advanced Genomics

and Bioinformatics Research Group (

IGBAM) of In-

formatics and Information Security Research Center

ITAK B

ILGEM).

Table 3: Data Sizes on a Smart Card.

Sample VCF

File

(KByte)

Compressed

Data

(Byte)

Undeﬁned

Varia-

tions

(Byte)

Total

(Byte)

1 1.188.364 73.080 9.487 82.567

2 1.151.748 72.484 8.833 81.317

3 1.146.093 71.978 7.955 79.933

4 1.174.939 72.563 8.747 81.310

5 1.304.739 70.587 3.626 74.213

6 1.237.843 71.144 12.129 83.273

Table 3 shows the results for six different whole

genomes. A typical smart card holds 256 kilobytes

of data and new high capacity cards hold 4 to 256

megabytes without compromising security. Our ex-

periment shows that our method requires maximum

100 kilobytes space. That means it is feasible for low

capacity cards and it can be easily implemented for a

whole human genome.

6 CONCLUSIONS

In this paper, we present a method for storing hu-

man genome in a smart card. Any human genome

can be represented with variations from the reference

genome. Our method selects the variations corre-

sponding to the exonic regions and ﬁlters them ac-

cording to their variant quality and genotype qual-

ity. It indexes the previously known variations by us-

ing dbSNP database. Furthermore, our method deals

with the variations that are speciﬁc to the person. We

show that our method can reduce a single whole hu-

man genome to the size small enough to be stored in a

smart card. We also present a privacy preserving sys-

tem in which genomic smart cards are used for disease

risk computation. There are two important advantages

of the proposed system: no need for infrastructure and

no need to operate on encrypted data.

As a future work, we will implement our method

on a Java Card smart cards. Thus we can compare the

performance of our disease risk computation method

with those of previously proposed methods.

REFERENCES

23andme (2013). https://www.23andme.com/welcome/.

[Online; accessed 16-July-2013].

Adida, B. and Kohane, I. (2006). Geneping: secure, scal-

able management of personal genomic data. BMC Ge-

nomics, 7(1):1–10.

Ayday, E., Raisaro, J. L., and Hubaux, J.-P. (2012). Privacy-

Enhancing Technologies for Medical Tests Using Ge-

nomic Data. Technical report.

Ayday, E., Raisaro, J. L., Laren, M., Jack, P., Fellay, J., and

Hubaux, J.-P. (2013). Privacy-Preserving Computa-

tion of Disease Risk by Using Genomic, Clinical, and

Environmental Data. In Proceedings of USENIX Se-

curity Workshop on Health Information Technologies

(HealthTech ’13).

Baldi, P., Baronio, R., De Cristofaro, E., Gasti, P., and

Tsudik, G. (2011). Countering gattaca: efﬁcient and

secure testing of fully-sequenced human genomes. In

Proceedings of the 18th ACM conference on Com-

puter and communications security, CCS ’11, pages

691–702, New York, NY, USA. ACM.

Blanton, M. and Aliasgari, M. (2010). Secure outsourcing

of dna searching via ﬁnite automata. In Proceedings

of the 24th annual IFIP WG 11.3 working conference

on Data and applications security and privacy, DB-

Sec’10, pages 49–64, Berlin, Heidelberg. Springer-

Verlag.

Bruekers, F., Katzenbeisser, S., Kursawe, K., and Tuyls,

P. (2008). Privacy-preserving matching of dna pro-

ﬁles. Cryptology ePrint Archive, Report 2008/203.

http://eprint.iacr.org/.

HumanGenomeinaSmartCard

315

Canim, M., Kantarcioglu, M., and Malin, B. (2012). Secure

management of biomedical data with cryptographic

hardware. Trans. Info. Tech. Biomed., 16(1):166–175.

Counsyl (2013). https://www.counsyl.com. [Online; ac-

cessed 16-July-2013].

Danecek, P., Auton, A., Abecasis, G., Albers, C. A., Banks,

E., DePristo, M. A., Handsaker, R. E., Lunter, G.,

Marth, G. T., Sherry, S. T., McVean, G., Durbin, R.,

and Group, . G. P. A. (2011). The variant call format

and vcftools. Bioinformatics, 27(15):2156–2158.

Illumina. Truseq exome targeted regions bed ﬁle.

http://support.illumina.com/downloads/truseq exome

targeted regions bed ﬁle.ilmn. [Online; accessed

19-December-2013].

Jha, S., Kruger, L., and Shmatikov, V. (2008). Towards

practical privacy for genomic computation. In Secu-

rity and Privacy, 2008. SP 2008. IEEE Symposium on,

pages 216–230.

Quinlan, A. R. and Hall, I. M. (2010). Bedtools: a ﬂexi-

ble suite of utilities for comparing genomic features.

Bioinformatics, 26(6):841–842.

Reber, William, L. and Perttunen, Cary, D. (1997). Personal

human genome card and methods and systems for pro-

ducing same. Patent Application. WO 1997/031327

A1.

Riva, A., Mandl, K. D., Oh, D. H., Nigrin, D. J., Butte, A. J.,

Szolovits, P., and Kohane, I. S. (2001). The personal

internetworked notary and guardian. I. J. Medical In-

formatics, 62(1):27–40.

Sherry, S. T., Ward, M., Kholodov, M., Baker, J., Phan, L.,

Smigielski, E. M., and Sirotkin, K. (2001). dbsnp:

the ncbi database of genetic variation. Nucleic Acids

Research, 29(1):308–311.

Simons, W. W., Mandl, K. D., and Kohane, I. S. (2005).

Model formulation: The ping personally controlled

electronic medical record system: Technical architec-

ture. JAMIA, 12(1):47–54.

Smart Card Alliance (2005). Taiwan

health care smart card project. http://

www.smartcardalliance.org/resources/pdf/Taiwan

Health Card Proﬁle.pdf”. Accessed: 2013-11-25.

Smart Card Alliance (2006a). French

sesam vitale health card. http://

www.smartcardalliance.org/resources/pdf/Sesam

Vitale.pdf. Accessed: 2013-11-25.

Smart Card Alliance (2006b). German health card. http://

www.smartcardalliance.org/resources/pdf/German

Health Card.pdf. Accessed: 2013-11-25.

Troncoso-Pastoriza, J. R., Katzenbeisser, S., and Celik, M.

(2007). Privacy preserving error resilient dna search-

ing through oblivious automata. In Proceedings of the

14th ACM conference on Computer and communica-

tions security, CCS ’07, pages 519–528, New York,

NY, USA. ACM.

Wikipedia (2013). Variant call format

— Wikipedia, the free encyclopedia.

https://en.wikipedia.org/wiki/Variant Call Format.

[Online; accessed 8-July-2013].

HEALTHINF2014-InternationalConferenceonHealthInformatics

316