Int
ern
at
i
onal
Journ
al of Ele
ctrical
an
d
Co
mput
er
En
gin
eeri
ng
(IJ
E
C
E)
Vo
l.
9
, No
.
5
,
Octo
ber
201
9
, pp.
3558
~
35
68
IS
S
N: 20
88
-
8708
,
DOI: 10
.11
591/
ijece
.
v9
i
5
.
pp3558
-
35
68
3558
Journ
al h
om
e
page
:
http:
//
ia
es
core
.c
om/
journa
ls
/i
ndex.
ph
p/IJECE
Fin
din
g a
suit
able th
re
sh
old valu
e for
an iris
-
ba
sed
authenti
cation sy
stem
Na
r
ongrit
W
angkeeree
1
, S
ir
apat
B
oonkr
ong
2
1
Facul
t
y
of
Infor
m
at
ion
T
ec
hnolo
g
y
,
King
Mongkut’s Unive
rsit
y
o
f
Technol
og
y
N
orth
Bangkok
,
T
hai
l
and
2
School
of
In
for
m
at
ion
T
ec
hnolo
g
y
,
In
sti
tut
e
of
S
oci
a
l
T
ec
hno
log
y
,
Surana
ree Uni
ver
sit
y
of
T
ec
hn
olog
y
,
Tha
i
la
nd
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Oct
5
, 2
01
8
Re
vised
A
pr
5
,
201
9
Accepte
d
Apr
17
, 201
9
Authent
i
ca
t
ion
i
s
the
first
l
ine
of
def
ense
o
f
a
n
y
informat
ion
te
chno
log
y
s
y
stems
.
One
of
m
an
y
popul
ar
m
et
hods
used
today
is
biometr
ic
,
and
ir
is
aut
hen
ti
c
at
ion
is
gai
n
ing
popu
la
r
ity
.
How
eve
r,
th
e
thr
eshold
v
al
u
e
de
emed
t
o
be
se
cur
e
and
a
ppropria
t
e
has
n
ot
b
ee
n
thorou
g
hl
y
stud
ie
d
.
Thr
eshold
is
a
val
ue
tha
t
d
efi
n
e
s
the
acce
p
ta
bl
e
amount
of
the
co
rre
ct
b
it
s
of
th
e
i
m
age
bef
ore
sec
ure
l
y
passing
the
au
the
nt
ic
a
tion
proc
ess.
Th
er
efo
re,
th
e
m
ai
n
ai
m
of
thi
s
rese
arc
h
was
to
find
a
sec
ur
e
and
sui
ta
bl
e
thr
eshold
v
al
ue
us
ed
in
iris
aut
hen
ti
c
at
ion
s
y
stem,
where
iri
s
l
oca
l
iz
a
ti
on
was
done
b
y
using
C
i
rcl
e
Hough
Tra
nsform
techn
ique
.
Iris
image
dat
ab
ase
s
v.
4
fro
m
the
Ch
ine
se
A
ca
dem
y
of
Scie
nc
es
Insti
tute
of
Autom
atic
(
CAS
IA)
were
us
ed
in
thi
s
rese
arch.
The w
a
y
to
f
ind
the
appr
o
pria
t
e
thre
shold
was
to
t
est
for
th
e
r
ight
b
al
an
ce
o
f
th
e
GA
R,
FR
R
and
FA
R
v
al
ues
when
tr
y
in
g
to
ver
if
y
the
p
erson’s
ide
n
ti
t
y
.
The
result
s
of
the
t
est
r
evea
le
d
tha
t
th
e
appr
opria
t
e
thre
shol
d
had
the
v
al
ue
of
72.
9246
per
ce
n
t
of
a
ll
th
e
av
ai
l
able
b
it
s
of
the
ir
is
image
.
Both
had
a
hig
h
GA
R
and
ver
y
low
FA
R
an
d
FR
R
v
al
ues.
I
t
c
an
be
con
cl
ud
e
d
th
at
the
ob
ta
in
e
d
thr
eshold
val
ue
was suit
ab
le
and
se
cur
e
.
Ke
yw
or
d
s
:
Au
t
hen
ti
cat
io
n
Ir
is
-
base
d
a
uthe
ntica
ti
on
Thr
e
shold
v
al
ue
Copyright
©
201
9
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed
.
Corres
pond
in
g
Aut
h
o
r:
Naro
ngrit
W
a
ngkee
ree,
Faculty
of In
form
ation
Tec
hn
ology,
King Mo
ng
ku
t
’s Un
i
ver
sit
y o
f
Tec
hnology
North
Ban
gkok,
Ba
ngkok
-
10
800,
Thail
an
d.
Em
a
il
: s56
07
0119
10021@e
m
ai
l. k
m
utn
b.a
c.th
1.
INTROD
U
CTION
To
day,
Bi
om
etr
ic
syst
em
s
are
widely
us
e
d
i
n
authe
ntica
t
ion
process
in
orde
r
t
o
i
den
ti
fy
an
ind
ivi
du
al
.
Bi
om
e
tric
can
be
div
id
ed
i
nto
tw
o
m
ai
n
m
et
ho
ds
.
T
he
fir
st
is
phys
ic
al
bio
m
et
ric
w
hich
incl
ud
es
face
,
fin
gerpr
i
nt
pal
m
and
iris
rec
ogniti
on
[
1
-
5].
The
sec
ond
i
s
kn
own
a
s
be
hav
i
or
al
bio
m
et
ric,
w
hich
in
cl
ud
es
wa
lkin
g
patte
r
n,
ty
pi
ng
patte
r
n
a
nd
hand
-
w
ritt
en
sig
natu
re
r
ecognit
ion
[
6].
Bi
om
e
tric
syst
e
m
s
can
help
e
nh
a
nce
the
sec
ur
it
y
of
ide
ntific
at
ion
and
a
uth
e
ntica
ti
on
m
echan
is
m
s.
It
is
cl
ai
m
ed
t
o
be
str
onger
tha
n
a
pas
swor
d
recog
niti
on
sys
tem
since p
ass
words ca
n b
e
f
org
otte
n,
disap
pear
e
d
a
nd stol
en [7
,
8].
On
e
of
the
m
os
t
widely
us
e
d
bio
m
et
ric
m
eth
ods
is
iris
rec
ogniti
on.
It
is
pro
ved
t
o
be
e
ff
ic
ie
nt
a
nd
com
es
with
pr
om
isi
ng
le
vel
of
secu
rity
[9
]
.
In
this
syst
e
m
,
an
iris
is
require
d
in
orde
r
to
ve
rify
the
pe
rson
’
s
sp
eci
al
cha
racteri
sti
cs
[10].
T
he
pa
rt
of
the
ir
is
that
is
us
e
d
f
or
i
den
ti
ty
ver
i
ficat
ion
is
locat
ed
be
twee
n
the
blac
k
center
par
t
of t
he
ey
e
(pu
pil)
and the
w
hite p
art of the
eye
ba
ll
(
scl
era)
.
Au
t
hen
ti
cat
io
n
by
i
ris
rec
ogni
ti
on
i
nvolv
es
e
xtracti
on
of
a
s
et
of
iris
im
ages
of
a
giv
e
n
ey
e.
T
hey
ar
e
then
u
se
d t
o ge
ner
at
e a
f
i
nal t
e
m
plate
(
iris t
e
m
pla
te
)
an
d
i
ris d
at
a,
in
b
it
s,
are
us
e
d
in
iris
te
st by c
om
par
ing al
l
po
i
nts
with
t
he
tem
plate
.
A
c
ha
nce
t
o
m
at
ch
a
ll
po
i
nts
is
le
ast
possi
ble
th
ough
a
n
i
ris
im
age
b
el
ongs
to
the
s
a
m
e
per
s
on
[
11]
.
A
s
a
res
ult,
the det
erm
inati
on
of
er
ror
rate
bet
ween
iris
te
m
plate
and
iris
te
s
t
for
aut
he
ntica
ti
on
is
need
e
d.
A
n
ac
cur
acy
rate
use
d
f
or
a
uth
e
ntica
ti
on
by
iris
re
cogniti
on
is
60
pe
rce
nt
or
us
i
ng
sta
ti
sti
cal
f
orm
ulas
for
a
c
o
m
par
ison
[12].
H
owe
ver,
i
n
t
he
sec
ur
it
y
as
pect,
th
e
m
entioned
ac
cur
acy
rate
can
no
t
be
w
orkab
l
e
as
a
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Find
i
ng a su
it
able
thres
hold v
alu
e f
or
an iri
s
-
base
d au
t
hen
t
ic
ation
syste
m
(
Naro
ngrit
W
angkee
ree
)
3559
chan
ce
for
e
rror
rate
is
qu
it
e
hi
gh.
Othe
r
e
xperim
ents
f
ocused
on
Eq
ual
Error
Ra
te
(EE
R)
value
or
t
hresh
ol
d
value
of
t
he
int
ersecti
on
of
F
AR
an
d
FRR
at
the
acce
pta
nce
le
vel
of
a
ppr
oxim
a
te
ly
50
pe
r
cent,
w
hich
is
a
good
for
us
a
bili
ty
bu
t
no
t
secu
re
[
13
]
.
I
n
t
his
pa
per,
a
thre
shol
d
value
by
int
ersecti
on
of
G
AR
a
nd
FRR
will
be
determ
ined.
C
on
s
eq
ue
ntly
,
th
is
resear
ch
pai
d
at
te
ntio
n
t
o
a
thres
hold
val
ue
from
wh
ic
h
a
n
acc
ur
acy
rate
from
a
com
par
iso
n
of
iris
te
m
plate
and
i
ris
te
st
can
be
a
ccepta
ble
for
aut
he
ntica
ti
on
in
a
su
it
abl
e
and
secu
re
m
ann
e
r
.
Con
tri
bu
ti
on
of
this
stu
dy
is,
th
eref
or
e
,
t
o
fin
d
a
s
uitable
t
hr
es
ho
l
d
val
ue
f
or
a
n
i
ris
-
base
d
a
ut
hen
ti
cat
io
n
syst
e
m
.
T
he
ef
fici
ency
of
t
he
a
uth
e
nt
ic
at
ion
m
et
ho
d
us
i
ng
iris
rec
ogniti
on
ca
n
be
m
easur
ed
an
d
e
valuat
e
d
by
usi
ng
the
f
ollow
i
ngs.
Fir
stl
y,
the
False
Re
j
ect
io
n
Ra
t
e
(F
RR
)
is
th
e
pro
portio
n
of
auth
entic
or
c
orrect
iris
that
are
i
ncorr
ect
ly
de
ni
ed.
Sec
ondly,
False
A
cce
pt
at
ion
Ra
te
(
F
AR)
is
t
he
propo
rtion
of
i
m
po
stors
or
fa
ke
iris
that
are
acc
epte
d
by
the
bio
m
et
ric
syst
e
m
.
The
G
enu
i
ne
Acce
ptance
Ra
te
(GA
R)
is
def
i
ned
as
,
G
AR
= 100
–
FRR
[
14
]
.
In
o
r
de
r
to
m
e
asur
e
a
nd
ev
al
uate
the
ef
fici
ency
an
d
sec
ur
it
y
as
sai
d
ab
ove
,
a
thres
hold v
a
lue
m
us
t
be
set
.
A
to
o
l
ow
t
hr
es
hold
value
m
ay
resu
lt
in
a
ver
ific
at
io
n
of
authe
ntica
ti
on
con
ta
ini
ng
a
hi
gh
val
ue
of
Ge
nu
i
ne
Accepta
nce
Ra
te
(GAR)
a
nd
a
lo
w
value
of
F
al
se
Re
j
ect
io
n
Ra
te
(F
RR
),
but
a
high
value
of
False
A
ccept
at
io
n
Ra
te
(FAR
),
high
pe
rfor
m
ance
of
us
a
bili
ty
bu
t
it
is
no
t
sec
ur
e.
W
it
h
a
to
o
hi
gh
th
reshold
val
ue
,
it
can
res
ult
in
a
ver
ific
at
io
n
of
aut
hen
ti
cat
io
n co
ntaini
ng
a
low
value
of
G
enu
i
ne Acc
ept
ance
Ra
te
(
GAR
) a
nd
a
hi
gh
value
of
False
Re
j
ect
i
on
Ra
te
(F
RR
),
and
a
ff
ect
s
t
he
inco
rr
ect
ne
ss
of
th
e
e
rror
rate
for
False
Acce
ptati
on
Ra
te
(FAR)
to
be
l
ow
acc
ordi
ng
ly
.
T
hough
a
uth
e
ntica
ti
on
with
t
oo
high
of
t
hr
es
hold
val
ue
ca
n
res
pond
to
the
secur
it
y
as
pec
t,
the
real
a
ppli
cat
ion
can
not
be
do
ne
as
c
orrect
data
can
be
filt
ered
at
the
sam
e
ti
m
e.
Ther
e
f
or
e,
a
n
a
naly
sis
to
fin
d
an
ap
pro
pr
ia
te
thres
ho
l
d
val
ue
sh
ould
be
c
oncer
ne
d
ab
out
real
ap
plica
ti
on
with
secur
e
as
pect [
15
]
.
2.
BACKG
ROU
ND K
NOWL
EDGE
AND REL
ATED
W
ORK
This
pa
per
f
oc
us
es
on
the
fin
ding
of
a
su
it
a
ble
an
d
secu
re
thres
ho
l
d
val
u
e
for
an
iris
a
ut
hen
ti
cat
ion
syst
e
m
.
In
ord
er
to
ac
qu
i
re
a
n
un
der
sta
nd
i
ng
of
this
te
c
hniqu
e,
relat
ed
t
he
or
ie
s
a
nd
rese
arch
es
i
nclu
di
ng
i
ris
recog
niti
on
sys
tem
and
Ci
rcle
H
ough
Tra
nsf
or
m
,
an
d
m
et
h
od
s
for
fin
ding
th
reshold
valu
es
will
be
ex
pl
ai
ned
in this se
ct
i
on
2
.
1
.
Iri
s
reco
gn
iti
on sy
stem
and
ci
rcl
e ho
u
gh
t
ransfor
m
Au
thentic
at
ion
or
identific
at
ion
pr
ocess
us
ing
iris
reco
gn
it
ion
syst
em
is
con
sidered
to
be
the
m
os
t
hig
hly
secur
ed
bio
m
et
ric
te
chn
olo
gy.
The
eff
ic
ie
ncy
of
detect
ion
is
based
on
pu
pil
dilat
ion
and
im
age
acqu
isi
ti
on
to
be
us
ed
in
the
reco
gn
it
ion
pr
ocess.
Other
factor
s
include
too
low
and
too
br
igh
t
li
gh
t
that
ca
n
le
ad
to
err
or
in
detect
ion
.
Ther
efo
re,
bef
or
e
us
ing
an
ey
e
im
age
fo
r
te
st
or
reco
gn
it
ion
,
a
pr
ocess
to
red
uce
a
n
err
or
of
reco
gn
it
ion
s
ho
uld
be
do
ne.
Fo
r
exam
ple,
con
ver
ti
ng
im
age
color
s
to
gr
ay
scal
e
so
as
to
el
im
inate
a
pr
ob
le
m
of
iris
color
. An
ey
e
im
age
is
com
po
sed
of
p
up
il
, iris,
scl
era,
ey
el
ashes,
ey
ebr
ow
s
and
the
top
p
art of
ey
e.
Ho
wev
er,
a
par
t
that
can
be
us
ed
fo
r
authen
ti
cat
ion
is
the
black
center
par
t
of
the
ey
e
or
iris
wh
ic
h
is
locat
ed
between
p
up
il
an
d
the w
hite part o
f
the eye
ball (scle
ra)
[
16
,
17
]
as seen
in
F
igu
re 1
.
Figure
1. Parts
of the
hum
an
ey
e
A
par
t
of
this
research
is
to
con
sider
the
per
fo
rm
ance
of
an
iris
re
cog
niti
on
syst
em
im
plem
ented
by
Ci
rcle
Ho
ug
h
Tran
sfo
rm
te
chn
iqu
e
to
detect
an
iris
im
age.
The
iris
reco
gn
it
ion
pr
ocess
can
be
div
ided
into
f
ou
r
par
ts:
ey
e im
age acq
uisit
ion
,
iris
and
pu
pil
segm
entat
ion
,
no
isy
iris
im
age
segm
entat
ion
and
featur
e
extracti
on
and
en
cod
ing
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
20
19
:
3558
-
3568
3560
2.1.1.
E
ye
im
age
acqu
isi
ti
on
In
this
research
,
im
ages
us
ed
in
iris
reco
gn
it
ion
syst
em
are
fr
om
CASI
A
Ir
is
Im
age
Datab
ase
f
or
Bi
om
et
ric
Id
eal
Test
[1
8
].
The
iris
im
ages
wer
e
captur
ed
by
a
hig
h
reso
luti
on
cam
era
so
bo
th
du
al
-
ey
e
iris
a
nd
face
patte
rn
s
wer
e
included
in
the
im
age
wh
ic
h
m
ade
them
su
it
able
in
this
research
.
Ir
is
im
ages
of
CASI
A
we
r
e
captur
ed
with
a
sel
f
-
dev
el
op
ed
cl
os
e
-
up
iris
cam
era.
The
m
os
t
com
pelli
ng
featur
e
of
the
iris
cam
era
is
that
it
has
been
design
ed
with
a
ci
rcu
la
r
Near
-
i
nf
rare
d
(N
IR
LED)
arr
ay
,
with
su
it
able
lum
ino
us
flux
fo
r
iris
im
aging
.
Be
cause
of
this
no
vel
design
,
the
iris
cam
era
can
captur
e
ver
y
cl
ear
iris
im
ages
and
well
-
su
it
ed
fo
r
stud
yi
ng
.
The
syst
em
all
ow
s
the
us
er
to
be
anywh
ere
fr
om
1
to
3
feet
(0
.3
m
et
ers)
aw
ay
fr
om
the
cam
era
that
l
ocates
th
e
fo
cus
on
the iris
as seen
in
F
igu
re 2
.
2.1.2.
I
ri
s and
pupil se
gmen
t
at
i
on
The
first
ste
p
is
to
isolat
e
the
act
ual
iris
reg
ion
in
a
dig
it
al
ey
e
im
age.
The
iris
reg
ion
can
be
app
ro
xim
at
ed
by
two
ci
rcles,
on
e
fo
r
the
pu
pil
/i
ris
bo
un
da
ry
and
the
oth
er
on
e
fo
r
the
iris/
scl
era
bo
un
da
r
y
.
Be
fo
re
detect
ion
of
these
bo
un
dar
ie
s,
the
edg
es
of
the
ey
e
im
age m
us
t be
fo
un
d
fr
om
pix
el
intensit
y.
Fr
om
th
e
edg
e
im
age,
the
Ci
rcu
la
r
Ho
ug
h
Tran
sfo
rm
can
be
us
ed
to
detect
the
centers
and
rad
i
i
of
the
two
bo
un
da
r
i
e
s
accord
ing
to
Dau
gm
an
Algo
rithm
as seen
in
F
igu
re 3
.
Jo
hn
Dau
gm
an
pr
op
os
ed
Dau
gm
an
Algo
rithm
,
a
m
ajo
r
par
t
of
Ir
is
Re
cog
niti
on
Syst
em
,
f
or
segm
entat
ion
pr
ocess
[1
3],
[1
4
].
The
al
go
rithm
can
be
wr
it
te
n
in
a
fu
nction
fo
rm
as,
(
,
0
,
)
|
(
)
∗
∫
(
,
)
2
,
,
|
.
Fr
om
the
Fu
nction,
(
,
)
is
a
pr
oced
ure
to
find
pix
el
intensit
y
(
,
)
fr
om
ey
e
im
ages
us
ed
in
a
te
st,
den
otes
the
rad
ius
of
var
iou
s
ci
rcu
la
r
reg
ion
s
with
the
cente
r
coo
rd
inate
s
at
(
0
,
0
)
,
is
the
sta
nd
ard
dev
ia
ti
on
of
the
Gau
ssian
distribu
ti
on
,
den
otes
a
Gau
ssian
filt
er
of
scal
e
sigm
a
(
)
,
(
0
,
0
)
is
the
assum
ed
centre
coo
rd
inate
s
of
the
iris,
is
the
con
tou
r
of
the
ci
rcle
giv
en
by
the
par
am
et
ers
(
,
0
,
0
)
,
Pu
pil
and
li
m
bu
s
bo
un
dari
es
are
exp
ect
at
ion
to
m
axim
iz
e
the
con
tou
r
integra
l
der
ivati
ve,
wh
ere
the
intensit
y
value
ov
er
the
ci
rcu
la
r
bo
rd
ers
wo
uld
m
ake
a
su
dd
en
chan
ge.
(
)
is
a
sm
oo
thing
f
un
ct
ion
co
ntro
ll
ed
by
do
ne
by incr
easi
ng
the intensity
o
f
the captur
ed
im
age
.
Figure
2. Exa
m
ples o
f
iris i
m
ages u
se
d
i
n t
his
researc
h
Figure
3. Se
gme
ntati
on
Proces
s
2.1.3.
Circl
e
h
ough tra
nsfor
m
Ci
rcle Ho
ug
h
Tran
sfo
rm
(CHT)
is a featur
e
extracti
on
te
chn
iqu
e
fo
r
detect
ing
ci
rcles
su
ch a
s
ey
es by
locat
ing
ci
rcu
la
r
ob
j
ect
s
fr
om
an
inp
ut
im
age.
Altho
ug
h
there
are
a
nu
m
ber
of
al
go
rithm
s
fu
nctionin
g
li
ke
Ci
rcle
Ho
ug
h
Tran
sfo
rm
,
it
is
m
or
e
con
siderab
ly
us
ed
and
eff
ect
ive
wh
en
com
par
ed
to
oth
ers.
The
qu
al
it
y
a
nd
color
of
the
im
age
are
adj
us
te
d
bef
or
e
im
plem
ented
in
CHT
pr
ocess.
Acco
rd
ing
to
Dau
gm
an’
s
al
go
rithm
,
there
wer
e
so
m
e
research
es
rely
ing
on
the
pr
ocess
in
detect
ing
ey
e
im
ages
fo
r
authen
ti
cat
ion
[1
3,
14
]
as
see
n
in the
(1)
.
∇
≡
(
,
)
(
,
)
=
1
2
2
−
(
−
0
)
2
+
(
−
0
)
2
2
2
.
(
,
)
(1)
The
(
1)
is
a
sm
oo
thing
func
ti
on
by
a
s
uitable
siz
e
of
σ
from
edg
e
detec
ti
on
te
ch
nique
s
for
iris
rec
og
niti
on
syst
e
m
.
Edg
e
m
ap
is
a
sel
ect
ion
pr
oced
ure
to
increase
a
wo
rk
ing
eff
ic
ie
ncy
of
Cir
cl
e
Ho
ug
h
Tran
sfo
rm
i
n
or
der
to
get
m
or
e
accurate
sh
apes
by
con
sidering
the
edg
e
po
ints,
as
descr
ibed
by
the
fo
rm
ula;
(
,
)
,
=
1
,
2
,
…
.
,
,
wh
ic
h
can b
e w
ritt
en
in the
(2)
and
(
3
)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Find
i
ng a su
it
able
thres
hold v
alu
e f
or
an iri
s
-
base
d au
t
hen
t
ic
ation
syste
m
(
Naro
ngrit
W
angkee
ree
)
3561
(
,
,
)
=
∑
ℎ
(
,
n
=
0
,
,
,
)
,
(2)
W
her
e
ℎ
(
,
,
,
,
)
=
{
1
(
,
,
,
,
)
=
0
;
0
ℎ
.
(3)
An
analy
sis
of
li
m
bu
s
and
pu
pil
wh
ic
h
are
bo
th
m
od
el
ed
as
ci
rcles
and
the
par
am
et
ric
Fu
nction
c
a
n
be
def
ined
in
(4)
.
(
,
,
,
,
)
=
(
−
)
2
+
(
−
)
2
−
2
.
(4)
The
center
of
the
ci
rcle
is
(
,
)
and
Ra
diu
s
is
,
wh
en
an
edg
e
po
int
is
ou
t
of
the
ci
rcle,
the
fu
nct
i
on
value i
s
equ
al
to
0
and
the
value of
Fu
nction
is equ
al
to
1
wh
ereas
Fu
nction
ℎ
is a
basic
pr
inciple
of
Ci
rcle
Ho
ug
h
Tran
sfo
rm
te
chn
iqu
e
.
Even
tho
ug
h
there
are
oth
er
al
go
rithm
s
pr
op
ose
d
fo
r
the
sam
e
pu
rp
os
e
su
ch
as
,
bu
t
Ci
rcle
Ho
ug
h Transf
or
m
is
sti
ll
deem
ed
app
ro
pr
ia
te
since
the
al
go
rithm
has
al
so
been
u
sed
and
ap
plied
i
n
[18
-
22]
.
2.1.4.
T
he rem
oval
of n
oise fact
ors
In
this
research
, a
rem
ov
al
of
no
ise
f
act
or
s that
aff
ect
an
accuracy
of
the
iris
reco
gn
it
ion
syst
em
suc
h
as
up
per
ey
el
ashes
and
lower
ey
el
ashes
in
an
ey
e
im
age as
the
bo
th
ey
e
la
sh
es
wer
e
necessary
since
they
coul
d
cause a h
igh
n
um
ber
o
f
err
or
s in
detect
ion
[
2
3
]
as seen
in
Figu
re 4
.
2.1.5.
N
orm
aliz
at
ion
proces
s
It
was
fo
un
d
that
an
err
or
in
detect
ion
cou
ld
be
caused
by
iris
incon
sist
ence/
pu
pil
dilat
ion
and
li
ght
sh
ining
into
t
he
ey
es
du
ring
data
colle
ct
ion
as
well
as
an
un
equ
al
com
par
ison
su
ch
as
a
distance
of
im
age
captur
e,
cam
era
ro
ta
ti
on
or
cam
era
ang
le
,
head
ti
lt
and
ey
e
ro
ll
ing
.
Ey
e
no
rm
al
iz
at
ion
pr
ocess
can
increas
e
sign
ific
antly
a
diff
eren
ce in
color
level
between
the
bl
ack
and
the
wh
it
e
par
ts
of
the
ey
e
to
red
uce t
he
err
or
i
n
the
detect
[2
4
]
-
[
30]
.
The
no
rm
al
iz
at
ion
m
od
ule
us
es
ey
e
im
age
to
transf
or
m
the
iris
te
xtu
re
fr
om
cartesi
an
t
o
po
la
r
coo
rd
inate
s.
The
pr
ocess,
o
ften
cal
le
d
iris un
wr
app
ing
, yie
lds a r
ect
ang
ular
entit
y.
Figu
re
5
sh
ow
s
an
iris
im
age
with
detect
ed
pu
pill
ary
and
iris
bo
un
dari
es
and
the
no
rm
al
iz
ed
reg
i
on
.
As
seen
in
Figu
re
5(
b)
,
ey
el
id
occlusion
and
ey
el
ash
pr
esence
in
the
iris
reg
ion
can
cause
artefact
s
in
th
e
no
rm
al
iz
at
ion
im
a
ge
bef
or
e featur
e extracti
on
.
Figure
4. Re
m
ov
al
of
no
ise
(a)
(b)
Figure
5
.
O
rigi
nal
an
d n
or
m
alizat
ion
iris im
a
ge,
(a) O
rigi
nal I
ris
Im
age,
(b) N
orm
al
iz
ation
iris
i
m
age
2.1.6.
Fea
tu
re
ext
r
act
i
on
or data
enc
od
in
g
The sy
ste
m
extracte
d an
ey
e im
age
in
a center
area b
et
ween
the
two ci
rcle
con
tou
rs,
cal
le
d t
he ret
ina
,
includin
g
sm
al
l
black
sp
ots
in
the
reti
na.
The
data
was
extracte
d
and
transf
or
m
ed
into
the
bin
ary
iris
by
con
vo
lving
enco
din
g,
i.e.,
bit
0
a
nd
bit
1
as
seen
in
Figu
re
6
[3
1
].
The
al
go
rithm
fo
r
extrac
ti
ng
[3
2
]
a
nd
gen
erati
ng
the iris tem
plate
is as f
ollow
s.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
20
19
:
3558
-
3568
3562
Figure
6
.
Feat
ure
e
xtracti
on fo
r
iris
Feat
ure Ext
ract
ion A
lgo
ri
th
m
[32]
le
ng
th = size (p
olar_
arr
ay
, 2
)*
2*
ns
cal
es
te
m
plate
= zero
s (
siz
e (p
olar_
arr
ay
, 1
),
leng
th)
le
ng
th2
= size (p
olar_
a
rr
ay
, 2
)
h
= 1
: si
ze (p
olar_
arr
ay
, 1
)
m
ask
= zero
s (
siz
e (tem
plate
))
For
k=1
to
ns
cal
es
Then
E1
= E0
{k
}
H1
= r
eal
(
E1)
> 0
H2
= im
age (
E1)
> 0
H3
= ab
s (
E1)
< 0
.0
00
1
For
i = 0
to (
le
ng
th2
-
1)
Then
j
a = d
ou
ble (
2*
ns
cal
es*(
i))
te
m
pl
at
e
(h
, j
a + (
2*
k)
-
1)
= H
1
(h
, i+1)
te
m
plate
(h
, j
a + (
2*
k)
)
= H2
(
h,
i+1)
m
ask
(h
, j
a + (
2*
k)
-
1)
= n
oise_ar
ray (h
, i+1)
| H
3(
h,
i+1)
m
ask
(h
, j
a + (
2*
k)
)
= n
oise_ar
ray (h
, i+1)
| H
3(
h,
i+1)
End Fo
r
End Fo
r
End A
lgo
ri
th
m
2.2.
Thre
sh
ol
d
value
Thr
esho
ld
value
is
a
ran
gin
g
com
par
ison
value
of
data
fr
om
iris
te
m
plate
and
iris
te
st.
The
com
par
ison
is
do
ne
by
bit
diff
eren
ce
at
each
po
int
and
po
sit
ion
.
The
Gen
uin
e
Acceptance
Ra
te
(G
A
R
)
,
False
Re
j
ect
ion
Ra
te
(F
RR
)
and
False
Acceptat
ion
Ra
te
(F
AR)
valu
es
are
al
so
evaluated.
Thr
esho
ld
Value
ha
s
to
be
in
su
it
able
and
safe
ran
ge.
If
a
ran
ge
of
thresh
old
value
is
too
hig
h,
it
can
aff
ect
the
eff
ic
ie
ncy
of
authen
ti
cat
ion
in
te
rm
s
of
data
filt
er
or
data
colli
sion
of
iris
data
wh
il
e
a
cor
rect
data
m
igh
t
be
filt
ered
ou
t
at
the
sam
e ti
m
e.
Ho
wev
er,
if
a
ran
ge
of
thresh
old
value
is
too
low,
it
can
resu
lt
in
hig
h
eff
ic
ie
ncy
of
reco
gn
it
i
on
of
co
rr
ect
iris d
at
a b
ut incr
easi
ng
d
at
a colli
sion
o
f
iris data acc
or
din
gly.
Equ
al
err
or
rate
(EER)
is
a
bio
m
et
ric
secur
it
y
syst
em
al
go
rithm
us
ed
to
pr
edeterm
ine
the
thresho
l
d
value
fo
r
it
s
False
Acceptat
ion
Ra
te
(F
AR)
and
it
s
False
Re
j
ect
ion
Ra
te
(F
RR
).
W
hen
the
rates
are
equ
a
l
,
the
com
m
on
value
is
ref
err
ed
to
as
the
equ
al
err
or
rate.
The
value
ind
ic
at
es
that
the
pr
op
ort
ion
of
fals
e
acce
ptances
is
equ
al
to
the
pr
op
ort
ion
of
false
rej
ect
ion
s.
The
lower
the
equ
al
err
or
rate
value,
the
hig
her
th
e
accuracy of
the b
iom
et
ric syst
em
[
33
]
as seen
in
the
F
igu
re 7
.
In
theor
y,
the
cor
rect
iris
sh
ou
ld
al
ways
value
hig
her
than
the
im
po
stors
iris
.
A
sing
le
thresh
old
coul
d
then
be
us
ed
to
separ
at
e
the
cor
rect
i
ris
fr
om
the
im
po
stor’
s
iris
[3
4
-
36
].
Figu
re
7
sh
ow
s
EER
value/Thr
esh
ol
d
value
of
the
intersect
ion
of
FA
R
and
FRR
at
the
acce
ptance
le
vel
of
app
ro
xim
at
el
y
50
per
cent,
wh
ic
h
is
a
go
od
fo
r
us
a
bili
ty
bu
t
no
t
secur
e
.
In
this
pap
er,
a
thresh
old
value
by
intersect
ion
of
GA
R
and
FRR
will
be
determ
ined.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Find
i
ng a su
it
able
thres
hold v
alu
e f
or
an iri
s
-
base
d au
t
hen
t
ic
ation
syste
m
(
Naro
ngrit
W
angkee
ree
)
3563
Figure
7. Eq
ua
l err
or r
at
e
(EE
R) for FRR
a
nd F
AR
[33]
3.
METHO
DOL
OGY
This
research
us
ed
a
data
set
of
iris
im
age
database
fr
om
the
Chinese
Acad
em
y
of
Scie
nces
In
sti
tut
e
of
Au
tom
at
ion
(CAS
IA
)
wh
ic
h
had
a
total
of
22
,5
00
iris
im
ages
fr
om
1,
65
0
vo
lun
te
ers.
All
iris
im
ages
we
r
e
8
-
bit
gr
ay
-
le
vel
.JPEG
file
s
captur
ed
with
a
ci
rcu
la
r
Near
-
infr
ar
ed
cam
era
(N
IR
LED)
.
The
iris
local
iz
at
io
n
was
im
plem
ented
us
i
ng
Ci
rcle
Ho
ug
h
Tran
sfo
rm
te
chn
iqu
e
pr
ior
to
find
ing
a
su
it
able
and
secur
e
thresho
l
d
value.
The
exp
erim
ental
te
st
was
div
ided
into
two
par
ts.
The
first
was
to
us
e
a
set
of
iris
im
ages
fo
r
identify
i
ng
a
thresh
old
value.
The
secon
d
was
to
us
e
the
rest
of
the
iris
im
ages
to
te
st
the
validit
y
of
the
ob
ta
ined
thresho
l
d
value.
3.1.
Data
set
use
d
to f
ind
t
h
resh
old
va
lue
Data
set
us
ed
to
find
a
thresh
old
value
con
ta
ined
ey
es
im
ages
fr
om
the
CASI
A
V.
4
database
f
or
Bi
om
et
ric
te
sti
ng
.
W
it
hin
that,
there
was
a
gr
ou
p
of
CASI
A
-
Iri
s
-
G
roup1
con
ta
ined
4
,
000
iris
im
ages
fr
om
20
0
per
so
ns
.
3.2.
Data
set
use
d
in th
res
ho
ld
value
test
in
g
The
data
set
that
wo
uld
be
us
ed
to
te
st
the
thresh
old
value
fo
r
the
app
ro
pri
at
eness
con
ta
ined
tw
o
gr
ou
ps
al
l
of
wh
ic
h
wer
e
fr
om
the
CASI
A
V.
4
database.
The
first
gr
ou
p
of
the
te
st
data
was
fr
om
th
e
CASI
A
-
Ir
is
-
Gro
up3
database
wh
ic
h
con
ta
ined
iris
im
ages
create
d
fr
om
an
al
go
rithm
to
im
it
at
e
real
ey
es
.
The
secon
d
gr
ou
p
was
fr
om
the
CASI
A
-
Ir
is
-
Group
2
database.
The
iris
im
ages
in
this
database
con
ta
ined
90
0
iris
im
ages
and
wer
e
gather
ed
fr
om
45
0
vo
lun
te
ers
wh
o
too
k
their
ow
n
im
ages
fr
om
m
ob
il
e
ph
on
e
.
Ther
e
we
r
e
2,
00
0
iris
im
ages
in this
database.
This
m
eans
that
the im
age
qu
al
it
y
was no
t
in
per
fect
con
diti
on
.
Ho
wev
er,
it
was
decided
that
this
gr
ou
p
of
im
ages w
ou
ld
ref
le
ct
real
-
wo
rld
app
li
cat
ion
m
or
e.
That
was
the
reaso
n
that t
hi
s
database
was
included
in
the
te
st
dataset
.
These
iris
im
ages
wer
e
us
ed
to
te
st
the
ob
ta
ined
thresh
old
value
i
n
te
rm
s
of
accuracy
and
secur
it
y.
Since
an
err
or
in
authen
ti
cat
ion
fr
om
iris
reco
gn
it
ion
par
tl
y
cam
e
fr
om
iris
im
ages
us
ed
in
a
te
st
with r
egar
ds
to
br
igh
tness,
hig
h reso
luti
on
, a
nd
a
distance of
b
ei
ng
away
fr
om
a
cam
era,
these factor
s w
ere already taken
into accou
nt w
hen
carr
yi
ng
o
ut the test pr
ocess.
3.3.
Thre
sho
l
d
val
ue anal
ys
is
The
analy
sis
thresh
old
value
is
pr
op
ose
d
an
ov
erv
ie
w
are
div
ided
into
two
par
ts:
find
ing
a
su
it
abl
e
determ
inati
on
an
d
secur
e thr
esho
ld v
al
ue,
an
d
te
sti
ng
the thr
esho
ld v
al
ue.
3.3.1.
E
ye
im
ages d
atase
t
The
ey
e
im
age
dataset
s
us
ed
in
iris
r
ecog
niti
on
syst
em
wer
e
fr
om
CASI
A
Ir
is
Im
age
Database
V.
4
f
or
Bi
om
et
ric
Id
eal
Test
.
The
iris
im
ages
in
the
CASI
A
Ir
is
Im
age
Database
V.
4
fo
r
Bi
om
et
ric
Id
eal
Test
we
r
e
detect
ed
or
locat
ed
by
the
Ci
rcle
Ho
ug
h
Tran
sfo
rm
m
et
ho
d.
The
data
was
then
enco
ded
and
t
ran
sfo
rm
ed
t
o
create
the b
inary v
al
ue
of
the iris b
y using
the algo
rithm
stat
ed
in secti
on
2
.1
.6
.
3.3.2.
Findi
n
g app
r
op
ri
at
e
thre
sho
l
d
va
lu
e
Diff
eren
t
ran
ges
and
nu
m
ber
s
fo
r
the
thresh
old
values
wer
e
exam
ined i
n
or
der
to
find
the
app
ro
pr
i
a
t
e
and
secur
e
va
lue.
The
evaluati
on
fo
r
the
secur
e
thresh
old
value
was
do
ne
by
com
par
ing
bin
ary
bits
betwe
e
n
the
iris
te
m
plate
and
the
iris
te
st.
In
oth
er
wo
rd
s,
the
bin
ary
bits
of
the
iris
te
m
plate
and
iris
te
st
wer
e
com
par
e
d
and
te
ste
d,
in bo
th
values a
nd
po
sit
ion
s.
Dif
fer
ent
thresh
old
values
wer
e
set
f
or
the an
al
ysi
s.
They
wer
e
>=5
0,
>=55
,
>=60
,
>=65
,
>=70
,
>=75
and
>=
80
per
cent
of
al
l
the
bin
ary
bits.
Each
thresh
old
value
was
evaluat
e
d
us
ing
three
crit
eria,
nam
el
y
Gen
uin
e
Acceptance
Ra
te
(G
AR),
False
Re
j
ect
ion
Ra
te
(
FRR
)
and
Fals
e
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
20
19
:
3558
-
3568
3564
Acceptat
ion
Ra
te
(F
AR).
This
was
do
ne
to
determ
ine
a
su
it
able
and
secur
e
thresh
old
value.
The
pr
ocess
is
dep
ic
te
d
in
Figu
re
8.
The
req
uired
thresh
old
value
sh
ou
ld
hav
e
a
hig
h
GA
R
value,
low
FRR
va
lue
and
th
e
lowest FA
R value [
35
].
Algo
rithm
fo
r Fi
nd
ing
the Su
it
able and
Secur
e
Thr
esho
ld
Value Th
is
sect
ion
ex
plains
ho
w
a
su
it
abl
e
and
secur
e
thresh
old
value
is
determ
ined.
The
al
go
rithm
beg
ins
with
the
com
par
ison
between
the
iris
te
m
plate
and
the
iris
te
st.
The
GA
R,
FRR
and
FA
R
are
al
so
determ
ine
d
us
ing
the
al
go
rithm
below
.
No
te
that
th
e
thresh
old
v
al
ue
in the algo
rithm
is the v
al
ue
set
as ex
plained
in the p
rev
iou
s p
arag
rap
h.
The
com
par
ison
is
do
ne
bit
by
bit
fr
om
the
first
to
the
nth
bit
at
each
po
int
to
find
the
m
at
ching
per
centage.
The
GA
R
valu
e
is
the
per
centage
of
the
cor
rect
iris
acce
pted
by
the
thresh
old
value.
The
FRR
valu
e
is
a
per
centage
of
the
cor
rect
iris
rej
ect
ed
by
the
thresh
old
value.
The
FA
R
value
is
am
ou
nt
of
im
po
stor’
s
irise
s
acce
pted
by
the
thresh
old
value.
Hen
ce,
it
is
necessary
that
this
per
centage
is
as
sm
al
l value
as
po
ssible
so
that
the im
po
stor’
s irises are n
ot accepted b
y t
he
thresh
old
v
al
ue.
Th
e thr
ee values
can b
e com
pu
te
d
accord
ing
ly
.
Figure
8
.
Fin
di
ng a
s
uitable
a
nd sec
ur
e t
he
t
hr
es
hold
value
Bi
t
Co
mpa
ri
so
n A
lgo
r
ithm
Be
gin
Re
ad
Ir
isTe
am
pate
Re
ad
Ir
isTe
st
For
i=
1
to n
Then
IF
Ir
isTe
am
pate (i)
== Ir
isTe
st
(i)
Then
Ma
tc
hed
Bi
ts = M
at
ched
Bi
ts + 1
El
se
Un
m
at
ched
Bi
ts = Un
m
at
ched
Bi
ts + 1
End IF
End Fo
r
Corr
ect
= Mat
ched
Bi
ts*1
00
/n
IF
Com
par
e
>=Thresh
ol
d
Value
Then
Re
su
lt
= Pass
El
se
Re
su
lt
= N
o
Pass
End IF
End C
omp
are
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Find
i
ng a su
it
able
thres
hold v
alu
e f
or
an iri
s
-
base
d au
t
hen
t
ic
ation
syste
m
(
Naro
ngrit
W
angkee
ree
)
3565
4.
RESU
LT
S
AND DI
SCUS
S
ION
4.1.
Thre
sho
ld
val
ue de
termina
t
ion
The
thresh
old
value
deem
ed
app
ro
pr
ia
te
and
secur
e
was
fo
un
d
us
ing
the
m
et
ho
d
exp
la
ined
in
th
e
pr
eviou
s
sect
ion
.
In
oth
er
wo
rd
s,
the
Ci
rcle
Ho
ug
h
Tran
sfo
rm
te
chn
iqu
e
was a
pp
li
ed
to
bo
th
the
iris
te
m
plate
s
and
the
iris
te
st
im
ages
fo
r
local
iz
at
ion
pu
rp
os
es.
The
com
par
ison
s
between
the
te
m
plate
s
and
the
te
st
im
ages
wer
e
carried
ou
t,
hav
ing
set
the
values
of
the
thresh
old
.
Th
e
resu
lt
s
that
wer
e
loo
ked
fo
r
wer
e
the
nu
m
ber
of
the
iris
te
st
im
ages
that
passed
the
sp
eci
fied
thresh
old
.
The
values
of
GA
R,
FRR
and
FA
R
wer
e
ob
ta
ined
as
sh
ow
n
Table 1.
Table
1
.
Re
s
ults from
co
m
par
ing t
he
i
ris tem
plate
s and iris
test
i
m
ages ag
ai
ns
t
pr
e
-
sp
eci
fied
th
res
ho
l
d values
Perf
o
r
m
an
ce
Thresh
o
ld
Value
>=5
0
>=5
5
>=6
0
>=6
5
>=7
0
>=7
5
>=8
0
GAR
9
2
.26
3
5
8
0
.64
3
3
7
2
.08
2
5
6
4
.18
2
5
5
5
.46
7
1
4
4
.53
2
9
4
0
.85
9
1
FRR
7
.73
6
5
1
9
.35
6
7
2
7
.91
7
5
3
5
.81
7
5
4
4
.53
2
9
5
5
.46
7
1
5
9
.14
0
9
FAR
7
0
.84
7
1
3
9
.26
3
4
2
4
.93
6
7
1
7
.03
9
3
1
2
.67
47
9
.13
6
0
6
.13
5
4
Table
1
s
h
o
w
s
t
h
e
c
o
m
p
a
r
i
s
o
n
r
e
s
u
l
t
s
i
n
p
e
r
c
e
n
t
a
g
e
w
h
e
n
e
v
a
l
u
a
t
i
n
g
c
l
a
s
s
i
f
i
e
r
p
e
r
f
o
r
m
a
n
c
e
o
f
G
A
R
,
F
R
R
,
a
n
d
F
A
R
.
T
h
e
a
i
m
o
f
t
h
i
s
p
a
p
e
r
w
a
s
t
o
f
i
n
d
a
s
e
c
u
r
e
t
h
r
e
s
h
o
l
d
v
a
l
u
e
f
o
r
a
n
i
r
i
s
a
u
t
h
e
n
t
i
c
a
t
i
o
n
s
y
s
t
e
m
.
T
h
a
t
i
s
,
f
o
r
s
e
c
u
r
i
t
y
p
u
r
p
o
s
e
s
,
t
h
e
p
r
o
c
e
s
s
w
a
s
c
a
r
r
i
e
d
o
u
t
t
o
f
i
n
d
t
h
e
v
a
l
u
e
t
h
a
t
g
a
v
e
a
l
o
w
F
A
R
v
a
l
u
e
.
M
o
r
e
o
v
e
r
,
i
n
t
e
r
m
s
o
f
c
o
r
r
e
c
t
n
e
s
s
,
t
h
e
A
c
c
e
p
t
a
n
c
e
R
a
t
e
(
G
A
R
)
v
a
l
u
e
n
e
e
d
e
d
t
o
b
e
g
r
e
a
t
e
r
t
h
a
n
F
a
l
s
e
R
e
j
e
c
t
i
o
n
R
a
t
e
(
F
R
R
)
v
a
l
u
e
.
F
r
o
m
T
a
b
l
e
1
,
i
t
c
a
n
b
e
s
e
e
n
t
h
a
t
t
h
e
t
h
r
e
s
h
o
l
d
v
a
l
u
e
s
t
h
a
t
s
a
t
i
s
f
i
e
d
t
h
e
a
b
o
v
e
c
r
i
t
e
r
i
a
a
r
e
t
h
e
v
a
l
u
e
s
o
f
>
=
7
0
a
n
d
>
=
7
5
.
T
h
e
t
h
r
e
s
h
o
l
d
v
a
l
u
e
o
f
>
=
7
0
h
a
d
t
h
e
v
a
l
u
e
s
o
f
G
A
R
=
5
5
.
4
6
7
1
,
F
R
R
=
4
4
.
5
3
2
9
a
n
d
F
A
R
=
1
2
.
6
7
4
7
.
T
h
e
t
h
r
e
s
h
o
l
d
v
a
l
u
e
o
f
>
=
7
5
h
a
d
t
h
e
v
a
l
u
e
s
o
f
G
A
R
=
4
4
.
5
3
2
9
,
F
R
R
=
5
5
.
4
6
7
1
a
n
d
F
A
R
=
9
.
1
3
6
0
.
T
h
e
r
e
s
u
l
t
s
f
r
o
m
T
a
b
l
e
1
w
e
r
e
t
h
e
n
p
l
o
t
t
e
d
i
n
F
i
g
u
r
e
9
i
n
o
r
d
e
r
t
o
f
i
n
d
a
s
u
i
t
a
b
l
e
a
n
d
s
e
c
u
r
e
t
h
r
e
s
h
o
l
d
v
a
l
u
e
.
Figu
re
9
disp
la
ys
three
li
nes
of
gr
aph
.
They
are
the
GA
R,
FRR
and
FA
R
values
fo
r
each
of
th
e
sp
eci
fied
thresh
old
values.
The
gr
aph
sh
ow
ed
that
the
values
of
GA
R
and
FA
R
te
nd
ed
to
decr
ease
as
th
e
thresh
old
value
increased.
Ho
wev
er,
the
values
of
FRR
wen
t
in
the
op
po
sit
e
directi
on
.
In
oth
er
wo
rd
s,
the
FRR
values
increased
as
the
thresh
old
value
increased.
In
reali
ty
,
a
su
it
able
thresh
old
value
wo
ul
d
be
the
on
e
that
ho
lds
a
hig
h
value
of
GA
R
and
a
low
value
of
FA
R.
Ho
wev
er,
it
wo
uld
be
diff
ic
ult
to
determ
ine
an
exact
valu
e
fr
om
Table
1.
It
was,
theref
or
e,
necessary
to
include
the
FRR
li
ne
into
the
gr
aph
to
assist
in
the
determ
ining
of
the thr
esho
ld v
al
ue.
Figure
9
.
Th
res
ho
l
d
value
a
nal
ysi
s
The
analy
sis
of
thresh
old
ran
ged
with
the
two
-
po
int
equ
at
ion
was
im
plem
ented.
In
this
pap
e
r
,
a
thresh
old
value
by
intersect
ion
of
GA
R
and
FRR
will
be
determ
ined.
Figu
re
9
sh
ow
s
that
the
intercepti
on
s
occu
rr
ed
at
two
po
sit
ion
s.
The
first
was
wh
ere
the
FA
R
li
ne
cro
ssed
with
the
FRR
li
ne.
The
secon
d
was
wh
e
r
e
the
GA
R
li
ne c
ro
ssed
with
the
FRR
li
ne.
The
po
int
of
interest
fo
r
the
sake
of
this
pap
er
was
the
intersect
ion
of
the
GA
R
and
FRR
li
nes.
This
po
int
sp
eci
fied
the
su
it
able
and
secur
e
thresh
old
value
becau
se
the
GA
R
valu
e
was
hig
her
than
FRR
value.
At
the
sam
e
ti
m
e,
the
FA
R
value
or
the
false
acce
ptance
rate
was
app
ro
xi
m
at
el
y
10
per
cent
of
al
l
the
im
ages,
wh
ic
h
cou
ld
be
co
ns
idered
su
it
able
and
secur
e
[1
7
].
Fu
rt
her
m
or
e,
cal
culat
ed
by
the
two
-
po
in
t
equ
at
ion
,
the
intersect
ion
of
interest
occu
rr
ed
at
the
thresh
old
value
of
app
ro
xim
at
el
y
72
.92
46
.
Ou
r
ob
ta
ined
thresh
old
value
is
hig
her
that
tho
se
cl
ai
m
ed
by
[
13
]
and
[
12
]
wh
os
e
values
are
50
and
60
,
resp
ect
ively
.
Ou
r
v
al
ue
wo
uld
b
e test
ed
fu
rther
in
the n
ext sect
ion
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
20
19
:
3558
-
3568
3566
4.
2.
Te
sting
th
e o
bt
ained t
hresho
ld va
lue
The
ob
ta
ined
thresh
old
value
of
72
.9
24
6
per
cent
of
al
l
the
avail
able
bits
was
te
ste
d
fo
r
cor
rectn
e
s
s
us
ing
oth
er
data
set
s
as
each
data
set
con
ta
ined
diff
eren
t
pr
op
erti
es
accord
ing
to
a
con
diti
on
of
data
colle
ct
io
n.
The
data
set
s
inv
olv
ed
in
the
te
st
wer
e
fr
om
two
gr
ou
ps
of
the
CASI
A
V.
4
database
-
CASI
A
-
Ir
is
-
G
r
oup2
a
nd
CASI
A
-
Ir
is
-
Group3.
Table 2
Sh
ow
s
Data set
o
f
te
st t
hr
esho
ld v
al
ue
72
. 9
24
6
.
Table
2
.
Data s
et
of test
thr
e
shold
value
72. 9
246
Ir
is
Test
Perf
o
r
m
an
ce
GAR
FRR
FAR
Ir
is
-
CAS
IA
-
Grou
p
2
7
8
.00
2
2
.00
0
.00
Ir
is
-
CAS
IA
-
Grou
p
3
7
7
.00
2
3
.00
2
.00
Av
erage
7
7
.50
2
2
.50
1
.00
Fr
om
Table
2,
the
ob
ta
ined
thresh
old
value
of
72
.9
246
was
te
ste
d
against
the
CASI
A
-
Iri
s
-
Group2
database.
It
was
fo
un
d
that
78
.
00
p
er
cent
of
the
iris
te
st
im
ages
resu
lt
ed
in
acce
ptance
rate
(G
AR),
22
.0
0
pe
r
cent
resu
lt
ed
in
false
rej
ect
ion
rate
(F
RR
),
and
no
ne
fell
in
the
false
acce
ptance
rate
(F
AR)
cat
ego
ry
.
This,
theref
or
e, can
b
e con
side
red
a su
it
able and
saf
e thr
esho
l
ss
d
value.
Fo
r
the
CASI
A
-
Iri
s
-
G
rou
p3
database,
it
was
fo
un
d
that
77
.0
0
per
cent
of
al
l
the
im
ages
resu
lt
ed
i
n
GA
R,
23
.0
0
per
cent
of
the
im
ages
resu
lt
ed
in
FRR
and
on
ly
2.
00
per
cent
of
al
l
the
iris
im
ages
resu
lt
ed
in
FA
R
.
Ther
efo
re,
it
can b
e cla
im
ed
that
the thr
esho
ld
value of
72
.9
246
is
secur
e since there
was
no
per
m
issi
ble
er
r
or
detect
ed
at
all
.
Fr
om
the
te
sts
on
the
two
databases,
it
was
fo
un
d
that
at
the
thresh
old
value
of
72
.9
246,
the
aver
a
ge
GA
R
value
was
77
.5
0
per
cent,
the
aver
age
FRR
value
was
22
.5
0
per
cen
t
and
the
aver
age
FA
R
value
was
1
per
cent.
It
can
be
seen
that
the
values
rep
resen
t
a
ver
y
sm
al
l
err
or
wh
en
com
par
ed
with
oth
er
research
es
suc
h
as that of
K
han
et al
. [
37
]
w
ho
se f
al
se accepta
nce r
at
e o
r
FA
R value w
a
s ap
pr
ox
im
at
el
y 23
p
ercent.
5.
CONCL
US
I
O
N
An
iris
authen
ti
cat
ion
syst
em
need
s
a
thresh
old
value
to
analy
ze
an
accuracy
or
rej
ect
ion
of
iris
im
ages.
If
the
determ
ined
thresh
old
value
is
too
hig
h,
the
err
or
rate
of
rej
ect
ing
the
gen
uin
e
iris
im
age
ca
n
occu
r.
On
the
oth
er
han
d,
if
the
determ
ined
thresh
old
value
was
too
low,
it
will
resu
lt
in
the
err
or
rate
of
accuracy
of
incor
rect
iris
im
ages.
This
research
was
con
du
ct
ed
based
on
the
interest
of
find
ing
a
su
it
able
a
nd
secur
e
thresh
old
value
on
an
iris
authen
ti
cat
ion
syst
em
,
with
Ci
rcle
Ho
ug
h
Tran
sfo
rm
te
chn
iqu
e
us
ed
fo
r
th
e
local
iz
at
ion
of
the
iris.
The e
xp
erim
ental
te
st of
thresh
old
ran
ge
m
od
el
ing
fr
om
the
data
set
of
CASI
A
V.
4
iris
im
age
database
in
a
gr
ou
p
of
CASI
A
-
Iri
s
-
Gro
up1
rev
eal
ed
that
the
su
it
able
thre
sh
old
value
was
hav
ing
72
.
92
46
per
cent
of
the
cor
rect
bits
wh
en
com
par
ed
the
iris
te
m
plate
with
the
te
st
iris
im
age.
This
value
of
thresho
l
d
was
cl
ai
m
ed
to
be
su
it
able
and
secur
e
becau
se
it
pr
ov
ided
a
hig
her
value
of
GA
R
than
FRR
,
wh
il
e
the
FAR
value
was
low.
W
hen
the o
btained t
hr
esho
ld
value
was t
est
ed
with
oth
er
data set
su
ch a
s CASI
A i
ris im
age
databa
s
e
ver
sion
4.
0
in
a
gr
ou
p
of
CASI
A
-
Ir
is
-
Group
2,
a
total
iris
im
ages
fr
om
1,
00
0
per
so
ns
,
the
resu
lt
rev
eal
ed
that
GA
R
value
was
78
.0
0
per
cent,
FRR
value
w
as
22
.0
0
per
cent,
and
FA
R
value
was
0.
00
per
cent.
W
hen
te
ste
d
with
the
data
set
of
CASI
A
iris
im
age
database
ver
sion
4.
0
in
a
gr
ou
p
of
CASI
A
-
Ir
is
-
G
rou
p3,
a
total
of
iris
im
ages
fr
om
45
0
per
so
ns
,
it
was
fo
un
d
that
GA
R
value
was
77
.0
0
per
cent,
FRR
value
w
as
23
.0
0
per
cent,
a
nd
FA
R
value
was
2.
00
per
cent.
Ther
efo
re,
the
thresh
old
value
at
72
.9
24
6
is
con
sidered
to
be
the
su
it
able
a
nd
secur
e ran
ge
to b
e app
li
ed
fo
r
a h
igh
secur
it
y l
evel o
f
authen
ti
cat
ion
m
et
ho
d.
REFERE
NCE
S
[1]
W
u
G.,
Zh
ao
M.
,
Han
L
.
and
Li
S.
"
A
finge
rpr
int
f
ea
tur
e
ext
ra
ct
ion
al
gor
it
hm
base
d
on
opti
m
al
de
ci
si
on
for
t
ext
co
p
y
det
e
ct
ion
,"
Int
ernati
onal Journal
of
S
ec
urit
y
and
Its
Applications
,
vol.
10
(
11
)
,
pp
.
67
-
78
,
2016
.
[2]
M.
M.
Mahm
oud
Mus
le
h,
I
.
I.
Ba,
K.
M.
A.
N
ofa
l,
et
al
.
,
"
Im
proving
inform
at
i
on
sec
ur
ity
in
e
-
banki
ng
b
y
usin
g
biometri
c
fing
er
print
:
a ca
se
of m
aj
or
bank
in
M
al
a
y
si
a
,"
In
t. J. Com
put.
Sc
i. Inf.
Sec
ur
,
vol
.
10
(
3
)
,
pp
.
1
–
4
,
2012
.
[3]
W
.
Abdul,
A.
Al
za
m
il
,
H
.
Masri,
et
al
.
,
"
Finge
rpri
nt
and
ir
is
te
m
pl
at
e
pro
tecti
on
fo
r
hea
l
th
informat
ion
s
y
stem
a
ccess
and
sec
ur
ity
,"
J
.
Me
d.
Imaging
H
eal
th
Inform
,
vo
l
.
7(6)
,
pp
.
1302
–
1308
,
2017
[4]
Ye C.,
Xiong
Z.,
Ding
Y.,
Zh
ang
X.,
W
ang G.,
Xu
F.,
and
Zha
ng
K
.
"
Para
l
le
l
joi
nt
f
inge
rprin
ti
ng
an
d
enc
r
y
pt
ion for
socia
l
m
ultim
edi
a
sharing
b
ase
d
on
game
of
li
f
e
,"
Inte
rnat
io
nal
Journal
of
Sec
urit
y
and
Its
Appl
i
cat
ions
,
vol.
10(7)
,
pp
.
93
-
102
,
2016
,
[5]
Nanda
kum
ar
K.,
Jain
A.
K.,
Pan
kant
i
S.
,
"
Fing
er
print
-
base
d
fuz
z
y
v
aul
t
:
Im
ple
m
ent
a
ti
on
and
pe
r
form
anc
e
,
"
I
EEE
transacti
ons on i
nformation
foren
sics
and
sec
urity
,
vol
.
2(4)
,
pp
.
7
44
-
757
,
2007
.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Find
i
ng a su
it
able
thres
hold v
alu
e f
or
an iri
s
-
base
d au
t
hen
t
ic
ation
syste
m
(
Naro
ngrit
W
angkee
ree
)
3567
[6]
Jorgensen
Z.
an
d
Yu
T.
,
"
On
m
ouse
d
y
n
amics
a
s
a
beha
vior
al
bi
om
et
ric
for
au
th
ent
i
ca
t
ion
,"
In
P
roce
edi
ngs
of
th
e
6th
ACM
Sympo
sium on
Information, Compute
r a
nd
Comm
unic
ations
Sec
urit
y
,
ACM
,
Mar
2011
,
pp
.
476
-
482
.
[7]
Le
ng
L
.
and
Zhang
J.
"
Dual
-
ke
y
-
bindi
ng
c
ancel
ab
le
pa
lmprint
cr
y
p
tos
y
stem
for
p
alm
print
prote
c
ti
o
n
and
informat
io
n
sec
urity
,"
Journ
al
of
Net
work
an
d
Computer
App
li
cations
,
vo
l.
34
(6),
pp
.
1979
-
198
9
,
2011
.
[8]
Garg
S.,
Kum
a
r
A.
and
Hanm
andl
u
M.
,
"
Bio
m
et
ric
aut
hen
ti
c
at
ion
using
fi
n
ger
na
il
surfa
ce
,"
In
2012
12
t
h
Inte
rnational
Co
nfe
renc
e
on
Intel
li
gent Sy
st
ems Design
and
Appl
i
c
ati
ons (
ISDA)
,
pp.
497
-
502
,
Nov
2012
.
[9]
Zhu
W
.
,
Zha
o
Z
.
,
and
Wu
Y.
,
"
An
al
gor
it
hm
of
e
y
e
la
sh
es
de
tec
ti
on
for
iri
s
re
co
gnit
ion
,"
In
te
rna
ti
onal
Journal
o
f
Sec
uri
t
y
and
Its
Appl
ic
a
ti
ons
,
vo
l.
10(7)
,
pp
.
195
-
202
,
2016
.
[10]
Rat
hgeb
C.
,
Iris
biomet
rics
:
Tem
plat
e
prote
ct
ion
and
advan
ce
d
co
mpar
ators
,
Salzburg
:
Unive
rsit
y
of
Salz
bur
g,
201
1.
[11]
Mi
y
a
za
wa
K.,
It
o
K.,
Aoki
T
.
,
Koba
y
ashi
K.
a
nd
Naka
j
ima,
H
.
,
"
An
eff
ective
a
pproa
ch
f
or
ir
is
rec
ogni
ti
on
usin
g
phase
-
base
d
image
m
atching
,"
IEE
E
transacti
o
ns
on
patt
ern
analy
sis
and
machine
int
e
ll
ig
e
nce
,
vol.
30(10
),
pp.
1741
-
1756
,
2008.
[12]
Daugm
an
J.
,
"
H
ow
iri
s
re
cogni
t
ion
works
,
"
In
The
essenti
al
gu
ide
to
image
pr
oce
ss
ing
,
Ac
adem
ic
Press
,
2
009
,
pp.
715
-
739
.
[13]
Daugm
an
J.,
an
d
Dow
ning
C.
,
"
Eff
ec
t
of
sev
ere
image
compress
ion
on
iri
s
rec
ogni
ti
on
per
form
anc
e
,"
IE
E
E
Tr
ansacti
ons on information Forensic
s and
S
ec
ur
it
y
,
vol
.
3(1
),
pp.
52
-
61
,
2008.
[14]
Lee
E.
C.
,
and
Son
S.
H.
,
"
Anti
-
s
poofing
m
et
hod
for
ir
is
r
ec
ogni
tion
b
y
combinin
g
the
op
ti
c
al
and
te
xtur
al
feature
s
of
hum
an
e
y
e
,"
KSII
Tr
ansacti
ons on
Inte
rn
et
&
Information
S
yste
ms
,
vol
.
6(9
)
,
2
012.
[15]
Mali
k
J
.
,
Sain
ar
a
y
ana
n
G.
,
and
Dahi
y
a
R
.
,
"
Mi
n
Max
Thre
shol
d
Rang
e
(MM
T
R)
base
d
app
ro
ac
h
in
pal
m
prin
t
aut
hen
t
icati
on
b
y
ph
ase
congr
u
e
nc
y
fe
at
ure
s
,"
In
2010
Inte
rnatio
nal
Confe
ren
ce
on
Signal
and
I
mage
Proce
ss
in
g
,
pp.
388
-
393
,
De
c
2010.
[16]
Jan
F.,
Us
m
an
I.
,
and
Agha
S.
,
"
Reliable
ir
i
s
loc
alization
u
sing
Hough
trans
form
,
histogra
m
-
bisec
ti
on
an
d
ec
c
ent
ri
ci
t
y
,"
S
i
gnal
P
roc
essing,
vol.
93(1),
pp.
2
30
-
241
,
2013
.
[17]
Jan
F.,
Us
m
an
I
.
,
Khan
S
.
A.
a
n
d
Mali
k
S.
A.
,
"
Iris
local
iz
a
ti
on
base
d
on
th
e
Ho
ugh
tra
nsform
,
a
rad
i
al
-
gr
adient
oper
at
or
,
and the gray
-
l
evel
in
te
n
sit
y
."
Opti
k
,
vol
.
124(23),
pp.
59
76
-
5985
,
2013
.
[18]
Chine
se
Ac
adem
y
of
Scie
n
ce
s
Instit
u
te
of
Au
tomati
on.
Casi
a
iri
s
image
data
base
Version
4.
0.
Oc
t
23
,
201
6.
htt
p://ww
w.sinobi
om
et
ric
s.
com
[19]
Verm
a
P.,
Dube
y
M.
,
Basu
S.
an
d
Verm
a
P.
,
"
Hough
tra
nsform
me
thod
for
ir
is
recognit
ion
-
A
biometric
appr
oa
ch
,"
Inte
rnational
Jo
urnal
of Eng
inee
ring a
nd
Innov
a
t
iv
e
Technol
og
y
(
IJE
IT)
,
vol.
1(6),
pp.
43
-
48
,
2012
.
[20]
Pall
av
P.
K.
a
nd
Granork
ar
S.
R
.
,
"
Inve
st
i
gat
ion
and
anal
y
sis
of
Hough
-
DCT
-
Ham
m
ing
dist
ance
base
d
m
et
hod
of
i
ris
re
cogni
ti
on
,"
In
te
r
nati
onal
Journal
of
Inno
vat
i
ve
Technol
og
y
and
Ex
plo
ring
Engi
n
ee
ring
,
vol
.
3(1)
,
pp.
181
-
185
,
20
13.
[21]
Mukhopadh
y
a
y
P.
and
Chaudh
uri
B.
B.
,
"
A
surve
y
of
Hough
Tra
nsform
,"
Pat
te
rn
Re
cogn
i
ti
on
,
vo
l.
48(3)
,
pp.
993
-
1010
,
2
015.
[22]
Daugm
an
J.
,
"
N
ew
m
et
hods
in
i
ris
re
cogni
t
ion
,"
IE
EE
Tr
ansactions
on
S
yste
ms
,
Man,
and
Cyb
e
rneti
cs,
Part
B
(
Cybe
rneti
cs)
,
v
ol.
37(5)
,
pp
.
11
67
-
1175
,
2007
.
[23]
Min
T
.
H.
and
P
ark
R
.
H.
,
"
E
y
elid
and
e
y
e
la
sh
d
et
e
ct
ion
m
et
hod
in
the
norm
al
i
zed
ir
is
image
usi
ng
th
e
par
abol
i
c
Hough m
odel
an
d
Otsu’s
thre
sho
ldi
ng
m
et
hod
,"
Pat
te
rn
rec
ogni
t
ion
l
et
t
ers
,
vo
l
.
3
0(12),
pp
.
1138
-
1143
,
2009
.
[24]
Sulae
m
an
D.,
N
ugroho
A.
S.
and
Gali
nium
M.
,
"
I
ris
Segm
ent
at
ion
using
Gradi
ent
Magnit
ude
and
Fourier
Descri
p
to
r
for
Multi
m
oda
l
Biom
et
ri
c
Authent
i
ca
t
ion
S
y
stem
,"
Jo
urnal
of
ICT
Re
search
a
nd
Appl
i
cat
ion
s,
vol.
10(3)
,
pp
.
2
09
-
22
7
,
2016
.
[25]
Sheel
a
S.
V.
and
Vijay
a
P.
A.
,
"
I
ris
rec
ogn
it
ion
m
et
hods
-
surve
y
,"
Inte
rnationa
l
J
ournal
of
Comp
ute
r
App
li
ca
ti
on
s,
vol.
3(5)
,
pp
.
19
-
25
,
2010
.
[26]
Proenc
a
H.
,
"
Iris
rec
ogni
ti
on:
On
the
segm
ent
a
ti
o
n
of
degr
ad
ed
i
m
age
s
ac
quir
ed
in
the
v
isibl
e
wa
vel
eng
th
,
"
IE
EE
Tr
ansacti
ons on Patt
ern
Analysis and
Mac
h
ine In
te
lligen
ce,
vol
.
3
2(8),
pp
.
1502
-
1
516
,
2010
.
[27]
E.
Moham
m
adi
Arvac
heh
,
"
A
st
ud
y
of
segm
enta
ti
on
and
norm
ali
za
t
ion
for
iri
s
re
cogni
ti
on
s
y
s
te
m
s
,
"
2006.
[28]
Arvac
heh
,
E
.
M
.
,
"
A stud
y
of
seg
m
ent
at
ion
and
n
orm
al
iz
a
ti
on
for iris r
e
cogni
t
ion
s
y
stems
,"
2006
.
[29]
Proenc
a
H.
and
Alexa
ndre
L
.
A.
,
"
Iris
r
ec
ogn
it
ion
:
An
anal
y
s
is
of
t
he
al
i
asing
prob
l
em
in
the
ir
is
nor
m
al
iz
a
ti
on
stage
,"
In
2006
Inte
rn
ati
onal
Conf
ere
nce
on
Compu
tat
ional
In
te
l
li
g
enc
e
and
S
ec
u
rity
,
v
o
l.
2
,
p
p.
177
1
-
1774.
Nov 2006
.
[30]
Nith
y
ana
nd
am
S.,
Ga
y
at
h
ri
K
.
S.
and
Pri
y
ad
arshin
i
P.
L.
K.
,
"
A
n
e
w
iri
s
norm
aliz
ation
proc
ess
for
r
ec
ogni
ti
on
s
y
ste
m
with
cr
y
p
togra
p
hic
te
chn
ique
s
,"
arXiv
preprin
t
ar
Xiv:1111.
5135,
2011.
[31]
Bel
ch
er
C.
,
and
Du
Y.
"
A
sele
c
tive
fe
at
ure
infor
m
at
ion
appr
o
ac
h
for
iri
s
image
-
q
ual
ity
m
e
asure
,"
IEE
E
Tr
ansacti
o
ns
on
Information Forensic
s and
S
ec
urit
y
,
vol.
3(3)
,
pp
.
572
-
577
,
2
008.
[32]
Daugm
an
J.
,
"
H
ow
iri
s
re
cogni
t
ion
works
,"
In
The
essenti
al
gu
ide
to
image
pr
oce
ss
ing
,
Ac
adem
ic
Press
,
2009
,
pp.
715
-
739
.
[33]
C
heng
J.
M.
an
d
W
ang
H.
C.
,
"
A
m
et
hod
of
esti
m
at
ing
th
e
eq
ual
err
or
rate
fo
r
aut
om
at
i
c
spea
ker
ver
if
icati
on
,"
In
2004
Int
ernat
ional
S
ymposium on
Chine
se
Sp
oke
n
Language
Proce
ss
ing
,
pp
.
285
-
288
,
De
c
20
04.
[34]
Chen
Z
.
,
Xi
a
X
.
,
and
Lu
an
F.
,
"
Autom
at
ic
onl
ine
sign
at
ur
e
v
e
rifi
c
at
ion
base
d
on
d
y
n
amic
fu
nct
ion
fe
at
ure
s
,"
In
2016
7th
I
EEE
Inte
rnat
ional
Confe
renc
e
on
Soft
ware
Eng
ineering
and
Serv
i
ce
S
ci
en
ce
(
ICSESS)
,
Aug
2016
,
pp.
964
-
968
.
[35]
Janra
tc
h
akool
W
.
,
Boonkr
ong
S.
and
Sm
anchat
S.
,
"
Findin
g
t
he
opt
imal
val
ue
for
thre
s
hold
cr
y
ptogr
ap
h
y
on
cl
oud
comp
uti
ng
,
"
In
te
rnat
ional
Journal
of
E
le
c
tric
al
a
nd
Computer
Engi
ne
ering
(
IJ
ECE
)
,
vol
.
6(6
),
pp.
2979
-
2988
,
2016.
[36]
Desm
edt
Y.
and
Franke
l
Y
.
,
"
Thr
eshold
cr
y
ptos
y
s
te
m
s
,"
In
Confe
r
enc
e
on
th
e
Theo
ry
and
A
ppl
ic
at
i
on
of
Crypto
logy
,
Springer, Ne
w
York,
NY
,
Aug
1989
,
pp
.
307
-
3
15
.
Evaluation Warning : The document was created with Spire.PDF for Python.