I
nd
o
ne
s
ia
n J
o
urna
l o
f
E
lect
rica
l En
g
ineering
a
nd
Co
m
pu
t
er
Science
Vo
l.
3
6
,
No
.
2
,
N
o
v
em
b
e
r
20
24
,
p
p
.
1
2
7
3
~
1
28
3
I
SS
N:
2
502
-
4
7
52
,
DOI
: 1
0
.
1
1
5
9
1
/ijee
cs
.v
3
6
.
i
2
.
pp
1
2
7
3
-
1
28
3
1273
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ee
cs
.
ia
esco
r
e.
co
m
Phishing
we
bsite
detec
tion
using
n
o
v
el i
ntegra
ti
o
n
o
f
BE
RT a
nd
XLNet
with
deep
lea
rning
sequen
ti
a
l mo
dels
K
o
ng
a
ra
Srini
v
a
s
a
Ra
o
1
,
Dines
h Va
llu
r
u
2
,
Sa
t
is
hk
um
a
r
P
a
t
na
la
3
,
Ra
v
i Ba
bu
Dev
a
re
dd
i
4
,
T
um
m
a
la
pa
lli Siv
a
Ra
ma
K
rish
na
5
,
An
da
v
a
ra
pu
Sra
v
a
n
i
6
1
D
e
p
a
r
t
me
n
t
o
f
C
S
E
,
F
a
c
u
l
t
y
o
f
S
c
i
e
n
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
,
I
C
F
A
I
F
o
u
n
d
a
t
i
o
n
f
o
r
H
i
g
h
e
r
E
d
u
c
a
t
i
o
n
(
I
F
H
E)
,
H
y
d
e
r
a
b
a
d
,
I
n
d
i
a
2
D
e
p
a
r
t
me
n
t
o
f
I
n
f
o
r
mat
i
o
n
Te
c
h
n
o
l
o
g
y,
M
L
R
I
TM
E
n
g
i
n
e
e
r
i
n
g
C
o
l
l
e
g
e
,
H
y
d
e
r
a
b
a
d
,
I
n
d
i
a
3
D
e
p
a
r
t
me
n
t
o
f
C
S
E
/
I
T,
G
M
R
I
n
st
i
t
u
t
e
o
f
Te
c
h
n
o
l
o
g
y
,
R
a
j
a
m,
I
n
d
i
a
4
D
e
p
a
r
t
me
n
t
o
f
C
S
E
,
G
I
TA
M
S
c
h
o
o
l
o
f
Te
c
h
n
o
l
o
g
y
,
G
I
TA
M
U
n
i
v
e
r
si
t
y
,
H
y
d
e
r
a
b
a
d
,
I
n
d
i
a
5
D
e
p
a
r
t
me
n
t
o
f
C
S
E
,
Jaw
a
h
a
r
l
a
l
N
e
h
r
u
T
e
c
h
n
o
l
o
g
i
c
a
l
U
n
i
v
e
r
si
t
y
K
a
k
i
n
a
d
a
,
K
a
k
i
n
a
d
a
,
I
n
d
i
a
6
D
e
p
a
r
t
me
n
t
o
f
C
S
E
,
G
I
TA
M
S
c
h
o
o
l
o
f
Te
c
h
n
o
l
o
g
y
,
G
I
TA
M
D
e
e
me
d
t
o
b
e
U
n
i
v
e
r
si
t
y
,
V
i
s
a
k
h
a
p
a
t
n
a
m
,
I
n
d
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Mar
13
,
2
0
2
4
R
ev
is
ed
Au
g
5
,
2
0
2
4
Acc
ep
ted
A
ug
11
,
2
0
2
4
P
h
ish
i
n
g
we
b
sites
p
o
se
a
sig
n
ifi
c
a
n
t
th
re
a
t
t
o
o
n
l
in
e
se
c
u
rit
y
,
n
e
c
e
ss
it
a
ti
n
g
ro
b
u
st
d
e
tec
ti
o
n
m
e
c
h
a
n
ism
s
to
s
a
fe
g
u
a
rd
u
se
rs'
se
n
siti
v
e
in
f
o
rm
a
ti
o
n
.
T
h
is
stu
d
y
e
x
p
lo
re
s
th
e
e
ffica
c
y
o
f
v
a
ri
o
u
s
d
e
e
p
lea
rn
in
g
a
rc
h
it
e
c
tu
re
s
fo
r
p
h
ish
in
g
we
b
site
d
e
tec
ti
o
n
.
I
n
i
ti
a
ll
y
,
trad
i
ti
o
n
a
l
se
q
u
e
n
ti
a
l
m
o
d
e
ls,
in
c
lu
d
i
n
g
re
c
u
rre
n
t
n
e
u
ra
l
n
e
tw
o
rk
s
(RNN
),
lo
n
g
sh
o
rt
-
term
m
e
m
o
ry
(LS
TM
),
a
n
d
g
a
ted
re
c
u
rre
n
t
u
n
i
t
(G
RU),
a
c
h
i
e
v
e
a
c
c
u
ra
c
ies
o
f
9
5
%
,
9
6
%
,
a
n
d
9
6
.
5
%
,
re
sp
e
c
ti
v
e
ly
,
o
n
a
c
u
ra
ted
d
a
ta
se
t.
Bu
il
d
in
g
u
p
o
n
t
h
e
se
re
su
lt
s,
h
y
b
rid
a
rc
h
it
e
c
tu
re
s
th
a
t
c
o
m
b
in
e
th
e
str
e
n
g
th
s
o
f
trad
it
io
n
a
l
se
q
u
e
n
ti
a
l
m
o
d
e
ls
wit
h
sta
te
-
of
-
th
e
-
a
rt
lan
g
u
a
g
e
re
p
re
se
n
tatio
n
m
o
d
e
ls,
b
id
irec
ti
o
n
a
l
e
n
c
o
d
e
r
re
p
re
se
n
tatio
n
s
fro
m
tran
sf
o
rm
e
rs
(
BERT
)
a
n
d
XLNe
t,
a
re
in
v
e
stig
a
ted
.
Co
m
b
in
a
ti
o
n
s
su
c
h
a
s
RNN
wit
h
BE
RT
,
BERT
wit
h
LS
TM
,
B
ERT
with
G
RU,
RNN
with
XLNe
t,
XLNe
t
with
L
S
TM
,
a
n
d
XLNe
t
with
G
RU
a
r
e
e
v
a
lu
a
ted
.
Th
r
o
u
g
h
e
x
p
e
rime
n
tat
io
n
,
a
c
c
u
ra
c
ies
o
f
9
4
.
5
%
,
9
6
.
5
%
,
9
6
.
1
%
,
9
5
.
7
%
,
9
7
.
4
%
,
a
n
d
9
7
%
,
re
sp
e
c
ti
v
e
ly
,
a
re
a
c
h
iev
e
d
,
d
e
m
o
n
str
a
ti
n
g
th
e
e
ffe
c
ti
v
e
n
e
ss
o
f
h
y
b
rid
d
e
e
p
lea
rn
in
g
a
rc
h
it
e
c
tu
re
s
in
e
n
h
a
n
c
i
n
g
p
h
is
h
in
g
d
e
tec
ti
o
n
p
e
rfo
rm
a
n
c
e
.
Th
e
se
fin
d
in
g
s
c
o
n
tr
ib
u
te
to
a
d
v
a
n
c
i
n
g
t
h
e
sta
te
-
of
-
th
e
-
a
rt
in
c
y
b
e
rse
c
u
rit
y
p
ra
c
ti
c
e
s
a
n
d
u
n
d
e
rsc
o
re
th
e
imp
o
rtan
c
e
o
f
lev
e
ra
g
in
g
d
i
v
e
rse
m
o
d
e
l
t
y
p
e
s t
o
c
o
m
b
a
t
o
n
li
n
e
th
re
a
ts effe
c
ti
v
e
ly
.
K
ey
w
o
r
d
s
:
B
E
R
T
GR
U
L
STM
Ph
is
h
in
g
web
s
ite
R
NN
XL
Net
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
Din
esh
Vallu
r
u
Dep
ar
tm
en
t o
f
I
n
f
o
r
m
atio
n
T
e
ch
n
o
lo
g
y
,
ML
R
I
T
M
E
n
g
in
ee
r
in
g
C
o
lleg
e
Hy
d
er
ab
ad
,
I
n
d
ia
E
m
ail:
d
in
esh
.
v
allu
r
u
1
5
@
m
lr
i
tm
.
ac
.
in
1.
I
NT
RO
D
UCT
I
O
N
T
h
e
p
r
o
life
r
atio
n
o
f
p
h
is
h
in
g
web
s
ites
p
o
s
es
a
s
ig
n
if
ica
n
t
th
r
ea
t
to
cy
b
e
r
s
ec
u
r
ity
,
en
d
a
n
g
er
in
g
th
e
p
r
iv
ac
y
an
d
f
in
a
n
cial
s
ec
u
r
ity
o
f
in
ter
n
et
u
s
er
s
wo
r
ld
wid
e.
Ph
is
h
in
g
attac
k
s
em
p
lo
y
d
ec
e
p
tiv
e
tech
n
iq
u
es
to
tr
ick
u
s
er
s
in
to
d
i
v
u
lg
in
g
s
en
s
itiv
e
in
f
o
r
m
atio
n
s
u
ch
as
p
ass
wo
r
d
s
,
cr
ed
it
ca
r
d
n
u
m
b
er
s
,
o
r
p
er
s
o
n
a
l
id
en
tific
atio
n
d
etails.
T
r
ad
itio
n
al
m
eth
o
d
s
o
f
d
etec
tin
g
p
h
is
h
in
g
web
s
ites
o
f
ten
r
ely
o
n
h
eu
r
is
tics
,
r
u
les,
o
r
m
an
u
ally
cu
r
ated
b
lack
lis
ts
,
wh
ich
m
ay
s
tr
u
g
g
le
to
k
ee
p
p
ac
e
with
th
e
ev
o
lv
i
n
g
s
o
p
h
i
s
ticatio
n
o
f
p
h
is
h
in
g
tactics.
I
n
r
ec
e
n
t
y
ea
r
s
,
d
ee
p
lear
n
in
g
(
DL
)
h
as
em
er
g
ed
a
s
a
p
r
o
m
is
in
g
ap
p
r
o
ac
h
f
o
r
e
n
h
an
cin
g
p
h
is
h
in
g
d
ete
ctio
n
b
y
lev
er
a
g
in
g
th
e
p
o
wer
o
f
n
eu
r
al
n
etwo
r
k
s
to
au
t
o
m
atica
lly
lear
n
p
atter
n
s
an
d
f
ea
tu
r
es
f
r
o
m
lar
g
e
-
s
ca
le
d
ata.
T
r
ad
itio
n
al
s
eq
u
en
tial
m
o
d
els
s
u
ch
as
r
ec
u
r
r
en
t
n
eu
r
al
n
etwo
r
k
s
(
R
NN)
,
lo
n
g
s
h
o
r
t
-
ter
m
m
em
o
r
y
(
L
STM
)
,
an
d
g
ated
r
ec
u
r
r
e
n
t
u
n
it
(
GR
U)
h
av
e
s
h
o
wn
co
n
s
id
er
ab
le
s
u
cc
ess
in
ca
p
tu
r
in
g
s
eq
u
e
n
tial
d
ep
en
d
e
n
cies w
ith
in
tex
t d
ata,
m
ak
in
g
t
h
em
s
u
itab
le
ca
n
d
id
a
tes f
o
r
p
h
is
h
in
g
d
etec
tio
n
task
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
3
6
,
No
.
2
,
No
v
em
b
er
20
24
:
1
2
7
3
-
1
28
3
1274
Ho
wev
er
,
th
e
a
d
v
en
t
o
f
tr
an
s
f
o
r
m
er
-
b
ased
lan
g
u
ag
e
r
ep
r
esen
tatio
n
m
o
d
els,
ex
em
p
lifie
d
b
y
b
id
ir
ec
tio
n
al
en
c
o
d
er
r
e
p
r
ese
n
tatio
n
s
f
r
o
m
tr
a
n
s
f
o
r
m
er
s
b
i
d
ir
ec
tio
n
al
en
co
d
er
r
e
p
r
esen
ta
tio
n
s
(
B
E
R
T
)
f
r
o
m
T
r
an
s
f
o
r
m
e
r
s
an
d
XL
Net,
h
as
r
ev
o
lu
tio
n
ized
th
e
f
ield
o
f
n
atu
r
al
lan
g
u
a
g
e
p
r
o
ce
s
s
in
g
(
NL
P)
b
y
o
f
f
er
i
n
g
u
n
p
ar
alleled
ca
p
ab
ilit
ies
in
ca
p
tu
r
in
g
co
n
tex
tu
al
r
elatio
n
s
h
i
p
s
an
d
s
em
a
n
tic
u
n
d
er
s
tan
d
i
n
g
in
tex
t
d
ata.
T
h
ese
m
o
d
els
h
av
e
d
em
o
n
s
tr
ated
r
e
m
ar
k
ab
le
p
er
f
o
r
m
a
n
ce
ac
r
o
s
s
v
ar
io
u
s
NL
P
task
s
,
p
r
o
m
p
tin
g
in
ter
est
in
th
eir
p
o
ten
tial
ap
p
licatio
n
t
o
cy
b
e
r
s
ec
u
r
ity
d
o
m
ain
s
s
u
ch
as
p
h
is
h
in
g
d
etec
tio
n
.
I
n
th
is
p
a
p
er
,
w
e
p
r
o
p
o
s
ed
a
n
o
v
el
in
teg
r
atio
n
o
f
B
E
R
T
an
d
XL
N
et
with
tr
ad
itio
n
al
s
eq
u
en
tial
m
o
d
els f
o
r
p
h
is
h
in
g
d
etec
tio
n
.
Desai
et
a
l.
[
1
]
p
r
o
p
o
s
ed
a
p
h
i
s
h
in
g
d
etec
tio
n
m
eth
o
d
th
at
f
o
cu
s
es
o
n
ac
h
iev
in
g
m
a
x
im
u
m
ac
cu
r
ac
y
wh
ile
h
av
in
g
a
s
m
all
f
ea
tu
r
e
co
llectio
n
.
A
v
ar
iety
o
f
p
er
f
o
r
m
an
ce
cr
iter
ia
wer
e
u
s
ed
to
ass
ess
th
e
ef
f
ec
tiv
en
ess
o
f
th
e
m
ac
h
in
e
lear
n
in
g
(
ML
)
m
o
d
els
th
at
wer
e
u
s
ed
.
Ou
r
d
ataset
in
clu
d
ed
a
c
o
llectio
n
o
f
a
u
t
h
e
n
t
i
c
U
R
L
s
f
r
o
m
t
h
e
A
l
e
x
a
d
a
t
a
s
e
t
a
s
w
e
l
l
a
s
p
h
i
s
h
i
n
g
U
R
L
s
t
a
k
e
n
f
r
o
m
t
h
e
P
h
i
s
h
T
a
n
k
d
a
t
a
s
e
t
.
H
e
e
t
a
l
.
[
2
]
ex
tr
ac
ted
s
em
an
ti
c
an
d
lo
n
g
-
r
an
g
e
d
ep
en
d
en
cies
f
r
o
m
web
s
ite
UR
L
s
tr
in
g
s
u
s
in
g
T
in
y
-
B
E
R
T
.
T
h
ese
ch
ar
ac
ter
is
tics
wer
e
em
p
lo
y
ed
in
a
Stack
in
g
alg
o
r
ith
m
-
b
as
ed
class
if
ier
with
f
o
u
r
b
asic
lear
n
er
s
:
ca
teg
o
r
ical
b
o
o
s
tin
g
(
C
atB
o
o
s
t
)
,
ex
tr
em
e
g
r
ad
ien
t
b
o
o
s
tin
g
(
XGBo
o
s
t
)
,
lig
h
t
g
r
ad
i
en
t
b
o
o
s
tin
g
m
a
ch
in
e
(
L
ig
h
tGB
M
)
,
an
d
g
r
a
d
ien
t
b
o
o
s
ted
d
ec
is
io
n
tr
ee
s
(
GB
DT
)
.
Ph
is
h
in
g
web
s
ites
wer
e
d
etec
ted
with
o
u
t
h
u
m
a
n
f
ea
tu
r
e
ex
tr
ac
tio
n
.
T
h
r
o
u
g
h
th
e
s
tack
in
g
en
s
em
b
le,
b
asic
lear
n
er
s
h
elp
ed
ea
ch
o
th
er
r
ed
u
ce
ca
teg
o
r
izatio
n
m
is
tak
es
an
d
im
p
r
o
v
e
g
en
e
r
a
lizatio
n
,
i
m
p
r
o
v
i
n
g
ac
cu
r
ac
y
.
Usi
n
g
a
d
ee
p
r
ein
f
o
r
ce
m
en
t le
ar
n
in
g
(
D
R
L
)
-
b
ased
class
if
ie
r
an
d
c
u
ttin
g
-
ed
g
e
f
ea
t
u
r
e
s
elec
tio
n
,
[
3
]
ac
ce
le
r
ated
tr
ai
n
in
g
wh
ile
im
p
r
o
v
ed
class
if
icatio
n
p
er
f
o
r
m
an
ce
.
Data
f
r
o
m
Me
n
d
eley
was
u
s
ed
to
t
est
f
iv
e
s
tatis
tical
an
d
c
o
r
r
el
a
tio
n
-
b
ased
f
ea
tu
r
e
-
r
an
k
i
n
g
tec
h
n
iq
u
es.
Usi
n
g
th
e
Gin
i
in
d
ex
,
th
e
s
elec
tio
n
s
tr
ateg
y
d
ec
r
ea
s
ed
th
e
n
u
m
b
er
o
f
co
lu
m
n
s
in
th
e
d
ataset
b
y
2
7
%,
s
av
ed
m
o
r
e
t
h
an
1
0
%
o
f
tr
ain
in
g
tim
e,
an
d
im
p
r
o
v
e
d
class
if
icatio
n
s
co
r
es.
P
illai
et
a
l.
[
4
]
ex
p
lo
r
ed
u
n
u
s
u
al
web
s
ite
cla
s
s
if
ier
ev
asio
n
attac
k
s
an
d
d
etec
tio
n
.
T
o
ad
d
r
ess
th
ese
d
ef
icien
cies,
UR
L
d
ata
ex
tr
ac
tio
n
an
d
ML
web
s
ite
class
if
icatio
n
wer
e
r
ec
o
m
m
e
n
d
ed
.
T
o
m
ain
tain
p
h
is
h
in
g
w
eb
s
ite
f
u
n
ctio
n
ality
an
d
ap
p
e
ar
an
ce
,
a
d
v
er
s
ar
ial
s
am
p
les tar
g
eted
class
if
icatio
n
f
ea
t
u
r
es.
A
m
u
ltimo
d
al
r
ep
r
esen
tatio
n
tech
n
iq
u
e
co
m
b
in
in
g
tex
tu
al
an
d
im
a
g
e
-
b
ased
elem
en
ts
t
o
id
en
tify
f
r
au
d
u
le
n
t
web
s
ites
[
5
]
.
Featu
r
es
wer
e
ex
tr
ac
te
d
b
y
two
c
o
n
v
o
lu
ti
o
n
al
n
e
u
r
al
n
etwo
r
k
(
C
NN
)
m
o
d
els
an
d
in
teg
r
ated
f
o
r
d
ec
is
io
n
-
m
ak
i
n
g
.
T
h
is
m
o
d
el
en
h
an
ce
d
Mat
th
ews
co
r
r
elatio
n
c
o
ef
f
icien
t
(
MCC
)
p
er
f
o
r
m
an
ce
b
y
4
%
an
d
d
ec
r
ea
s
ed
f
alse
p
o
s
itiv
es
b
y
1
.
6
%.
An
ar
ch
ite
ctu
r
e
b
y
[
6
]
u
s
ed
s
p
ec
ial
c
h
a
r
ac
ter
s
to
s
ep
ar
ate
UR
L
s
in
to
f
o
u
r
p
ar
ts
:
p
r
o
to
co
l
ty
p
e,
d
o
m
ain
,
s
u
b
-
d
o
m
ai
n
,
an
d
to
p
-
lev
el
d
o
m
ain
(
T
L
D)
.
Featu
r
es
wer
e
u
tili
ze
d
to
cr
ea
te
a
v
o
ca
b
u
lar
y
d
atab
ase.
A
m
o
d
if
ied
Fas
tTe
x
t
wo
r
d
em
b
ed
d
in
g
m
eth
o
d
p
r
o
d
u
ce
d
n
u
m
er
ic
f
ea
tu
r
e
v
ec
t
o
r
s
.
T
h
ese
v
ec
to
r
s
an
d
p
r
e
-
p
r
o
ce
s
s
ed
UR
L
ex
a
m
p
les
tau
g
h
t
a
B
iLST
M
class
if
ier
.
Ph
is
h
in
g
UR
L
s
wer
e
d
etec
ted
with
ex
ce
llen
t
ac
cu
r
ac
y
an
d
litt
le
p
r
o
ce
s
s
in
g
co
s
t.
Z
o
n
ta
an
d
Sath
iy
an
a
r
ay
an
an
[
7
]
e
x
am
in
ed
ev
o
lv
in
g
th
r
ea
ts
an
d
r
is
k
-
r
e
d
u
ctio
n
s
tr
ateg
ies
u
s
in
g
m
alicio
u
s
d
etec
tio
n
to
d
ef
en
d
we
b
b
r
o
wser
s
.
I
t
s
tr
ess
ed
th
e
n
ec
ess
ity
o
f
p
r
o
ac
tiv
e
d
ete
ctio
n
in
b
u
ild
i
n
g
cy
b
er
s
ec
u
r
ity
f
r
am
ewo
r
k
s
th
at
ca
n
an
ticip
at
e
an
d
d
ef
ea
t
f
u
tu
r
e
cy
b
er
attac
k
er
s
'
m
eth
o
d
s
as
well
as
r
ea
ct
to
ex
is
tin
g
attac
k
s
.
No
wr
o
o
zi
et
a
l.
[
8
]
u
s
e
d
ML
class
if
ier
s
f
o
r
d
etec
tio
n
.
Ou
r
f
alse
n
eg
ativ
e
r
ate
was
0
.
0
0
3
7
.
K
-
m
ea
n
s
was
u
s
ed
to
clu
s
ter
d
ata
an
d
ex
am
in
e
d
ec
is
io
n
tr
ee
-
b
ased
m
o
d
els'
s
en
s
i
tiv
ity
to
lim
ited
k
n
o
wled
g
e
attac
k
s
,
in
clu
d
in
g
ze
r
o
th
o
r
d
er
o
p
tim
izatio
n
.
Me
n
o
n
a
n
d
An
an
d
h
u
[
9
]
u
s
ed
ML
-
b
ased
UR
L
d
etec
tio
n
.
A
n
ew
co
llec
tio
n
o
f
UR
L
attr
ib
u
tes
an
d
b
eh
av
io
r
s
p
lu
s
a
ML
alg
o
r
ith
m
c
o
m
p
r
is
ed
th
e
r
ec
o
m
m
en
d
e
d
d
etec
tio
n
ap
p
r
o
ac
h
.
T
h
e
test
in
g
r
esu
lts
s
h
o
we
d
th
at
th
e
p
r
o
p
o
s
ed
UR
L
attr
ib
u
tes
an
d
b
eh
a
v
io
r
m
ig
h
t
im
p
r
o
v
e
h
ar
m
f
u
l
UR
L
r
ec
o
g
n
itio
n
.
Sto
ler
iu
et
a
l.
[
1
0
]
s
u
g
g
ested
a
m
eth
o
d
th
at
m
a
k
es
u
s
e
o
f
ML
tech
n
iq
u
es
an
d
th
r
ea
t
in
tellig
en
ce
d
ata
to
id
e
n
tify
d
an
g
er
o
u
s
s
h
o
r
t
UR
L
s
.
T
h
e
m
o
s
t
ef
f
icien
t m
eth
o
d
was
r
an
d
o
m
f
o
r
est
(
R
F)
,
w
h
ich
p
r
o
d
u
ce
d
a
h
ig
h
lev
el
o
f
ac
cu
r
ac
y
.
T
h
e
tech
n
o
lo
g
y
was
p
u
t in
to
p
r
ac
tice
an
d
its
ef
f
ica
cy
was c
o
n
f
ir
m
e
d
in
p
r
ac
tical
ap
p
licatio
n
s
.
Ma
licio
u
s
UR
L
s
wer
e
f
o
u
n
d
in
[
1
1
]
in
a
n
u
m
b
er
o
f
ap
p
s
.
T
h
e
d
ataset,
wh
ich
in
clu
d
ed
m
o
r
e
th
an
6
lak
h
UR
L
s
f
o
r
th
e
im
p
lem
en
tatio
n
,
was
d
iv
id
ed
in
to
f
o
u
r
ca
teg
o
r
ies
:
p
h
is
h
in
g
,
b
en
i
g
n
,
d
ef
ac
em
en
t,
a
n
d
m
alwa
r
e.
T
h
r
ee
ML
m
o
d
els
n
am
ely
L
ig
h
tGB
M,
XGBo
o
s
t,
an
d
r
a
n
d
o
m
ize
d
f
o
r
est
wer
e
u
s
ed
to
id
en
tif
y
an
d
ca
teg
o
r
ize
d
an
g
er
o
u
s
UR
L
s
.
Ar
o
r
a
et
a
l.
[
1
2
]
co
n
d
u
cted
a
cu
r
r
en
t
ev
alu
atio
n
o
f
th
e
i
m
p
o
r
tan
t
alg
o
r
ith
m
s
u
s
ed
to
id
e
n
tify
f
r
au
d
u
len
t
U
R
L
s
,
with
RF
d
em
o
n
s
tr
atin
g
th
e
g
r
ea
test
ac
cu
r
ac
y
.
Ad
d
itio
n
ally
,
a
n
u
m
b
e
r
o
f
co
n
ce
r
n
s
with
d
ev
elo
p
in
g
alte
r
n
ativ
e
alg
o
r
ith
m
s
f
o
r
i
d
en
tify
in
g
d
a
n
g
er
o
u
s
UR
L
s
wer
e
co
v
er
ed
.
ML
an
d
DL
tech
n
iq
u
e
s
wer
e
u
s
ed
in
[
1
3
]
to
id
en
tify
d
a
n
g
er
o
u
s
UR
L
s
th
at
m
ig
h
t
lead
to
cy
b
er
attac
k
s
.
Ma
n
y
ML
an
d
DL
alg
o
r
ith
m
s
wer
e
u
s
ed
,
s
u
ch
a
s
RF
an
d
Naïv
e
B
ay
es
f
o
r
M
L
an
d
R
NN
an
d
L
STM
f
o
r
D
L
.
W
an
g
et
a
l.
[
1
4
]
in
tr
o
d
u
ce
d
Ph
is
h
B
E
R
T
,
a
p
r
etr
ain
ed
d
ee
p
t
r
an
s
f
o
r
m
e
r
n
etwo
r
k
f
o
r
p
h
is
h
in
g
UR
L
d
etec
tio
n
.
I
t
was
f
in
e
-
tu
n
e
d
o
n
b
e
n
ig
n
an
d
m
alicio
u
s
UR
L
d
ata
u
s
in
g
s
u
p
er
v
is
ed
ad
v
e
r
s
ar
ial
m
eth
o
d
s
,
o
u
tp
e
r
f
o
r
m
in
g
c
u
r
r
en
t
s
tate
-
of
-
th
e
-
ar
t
m
eth
o
d
s
in
ef
f
icien
c
y
,
r
o
b
u
s
tn
ess
,
an
d
ac
cu
r
ac
y
.
Mu
ltip
le
b
in
ar
y
class
if
ier
s
wer
e
tr
ain
ed
to
d
if
f
e
r
en
ti
ate
o
n
e
ca
teg
o
r
y
f
r
o
m
th
e
r
em
ain
d
er
[
1
5
]
.
On
e
v
s
all
class
if
ica
tio
n
ca
n
ac
cu
r
ately
p
r
e
d
ict
h
a
r
m
f
u
l
UR
L
s
ev
en
with
m
in
im
al
d
atasets
co
m
p
r
is
in
g
s
im
p
ly
d
o
m
ain
an
d
p
at
h
.
T
h
is
tec
h
n
iq
u
e
s
h
o
ws
p
o
t
en
tial
in
p
r
o
d
u
cin
g
ac
cu
r
ate
f
i
n
d
in
g
s
with
lo
w
co
m
p
u
ter
r
eso
u
r
ce
s
,
o
u
tp
er
f
o
r
m
in
g
co
n
v
en
tio
n
al
ap
p
r
o
a
ch
es.
L
ex
ical
UR
L
ca
teg
o
r
izatio
n
was
s
tu
d
ied
in
[
1
6
]
.
XGBo
o
s
t,
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
e
(
SVM
)
,
an
d
a
r
tific
ial
n
eu
r
al
n
etwo
r
k
(
ANN
)
d
etec
ted
m
alicio
u
s
UR
L
s
with
8
8
%,
8
7
%,
an
d
8
8
%
ac
cu
r
ac
y
.
p
h
is
h
in
g
UR
L
i
m
b
alan
ce
s
af
f
ec
ted
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
P
h
is
h
in
g
w
eb
s
ite
d
etec
tio
n
u
s
in
g
n
o
ve
l
i
n
teg
r
a
tio
n
o
f
B
E
R
T a
n
d
X
LN
et
…
(
K
o
n
g
a
r
a
S
r
in
i
va
s
a
R
a
o
)
1275
d
et
ec
tio
n
.
SVMs,
R
Fs
,
DT
s
,
KNNs,
an
d
B
ay
esian
o
p
tim
izatio
n
class
if
ied
UR
L
s
in
[
1
7
]
.
DR
L
SH,
B
PLSH,
an
d
r
a
n
d
o
m
in
s
tan
ce
s
elec
tio
n
wer
e
u
s
ed
f
o
r
e
f
f
icien
cy
.
S
VM
s
f
ar
ed
well
with
e
x
ten
d
e
d
tr
ain
in
g
d
u
r
atio
n
s
,
wh
ile
R
Fs
h
ad
g
r
e
at
ac
cu
r
ac
y
,
r
ec
all,
an
d
F1
s
co
r
es.
I
n
UR
L
ca
teg
o
r
izatio
n
,
in
s
tan
ce
s
elec
tio
n
g
r
ea
tly
af
f
ec
ted
m
o
d
el
p
er
f
o
r
m
a
n
ce
.
C
y
b
er
L
en
,
a
DL
-
b
ased
m
alicio
u
s
UR
L
d
etec
tio
n
s
y
s
tem
,
w
as
s
u
g
g
ested
in
[
1
8
]
.
Po
s
itio
n
em
b
ed
d
in
g
d
ec
r
ea
s
ed
to
k
en
am
b
ig
u
ity
,
wh
er
ea
s
f
ac
to
r
izatio
n
m
ac
h
in
e
(
FM)
lear
n
t
laten
t
lex
ical
f
ea
tu
r
e
in
ter
ac
tio
n
s
.
T
C
N
d
etec
ted
UR
L
to
k
en
lo
n
g
-
d
is
tan
ce
d
ep
e
n
d
en
cies.
C
ap
s
Net
an
d
I
n
d
R
NN
co
m
b
in
e
d
to
co
llect
m
u
lti
-
m
o
d
al
d
ata
a
n
d
m
e
r
g
e
tex
t
u
r
e
an
d
tex
t
[
1
9
]
.
T
h
e
n
etwo
r
k
f
ilter
ed
d
ee
p
ch
ar
ac
ter
is
tics
u
s
in
g
an
atten
tio
n
s
tr
ateg
y
to
im
p
r
o
v
e
h
ar
m
f
u
l
UR
L
ca
teg
o
r
izatio
n
ac
cu
r
ac
y
,
ac
co
r
d
in
g
to
tr
ials
.
L
ak
s
h
m
an
ar
ao
et
a
l.
[
2
0
]
ap
p
lied
ML
tech
n
i
q
u
es
to
id
e
n
tify
p
h
is
h
in
g
ass
au
lts
.
T
h
e
p
r
o
p
o
s
al
in
clu
d
e
d
two
p
r
io
r
ity
-
b
ased
alg
o
r
ith
m
s
.
T
h
e
f
in
al
f
u
s
io
n
class
if
ier
was
ch
o
s
en
f
r
o
m
t
h
ese
m
eth
o
d
s
.
An
in
n
o
v
ativ
e
f
u
s
io
n
c
lass
if
ier
was
u
s
ed
o
n
a
UC
I
d
ataset
an
d
y
ield
ed
ex
ce
llen
t
r
esu
lts
.
Nan
awa
r
e
[
2
1
]
u
s
ed
M
L
to
id
en
tify
p
h
is
h
in
g
UR
L
s
an
d
tex
t
p
r
o
ce
s
s
in
g
to
ass
ess
tex
t
f
o
r
p
h
is
h
in
g
attac
k
in
d
icatio
n
s
,
ad
d
r
ess
in
g
a
m
ajo
r
s
ec
u
r
ity
is
s
u
e
f
r
o
m
co
u
n
ter
f
eit
web
s
ites
an
d
UR
L
s
.
C
h
en
et
a
l.
[
2
2
]
s
u
g
g
ested
a
p
h
is
h
in
g
UR
L
d
etec
ti
o
n
ap
p
r
o
ac
h
b
ased
o
n
UR
L
co
n
te
n
t,
u
s
in
g
b
asic
f
ea
tu
r
es
f
o
r
p
r
elim
in
ar
y
s
cr
e
en
in
g
a
n
d
C
NN
f
o
r
m
alicio
u
s
UR
L
id
en
tific
atio
n
u
s
in
g
p
ag
e
c
o
n
ten
t,
with
r
ea
s
o
n
ab
le
ac
c
u
r
ac
y
.
T
h
e
h
y
b
r
id
m
eth
o
d
o
lo
g
y
[
2
3
]
u
s
ed
cla
s
s
ical,
ML
,
an
d
DL
to
id
en
tif
y
d
an
g
er
o
u
s
UR
L
s
.
T
o
id
en
tify
n
ew
an
d
ex
is
tin
g
d
an
g
er
o
u
s
UR
L
s
,
it u
s
ed
s
h
allo
w
an
d
DL
co
u
p
led
with
s
ig
n
at
u
r
es.
Nad
k
ar
n
i a
n
d
B
o
r
k
ar
[
2
4
]
u
s
ed
RF
with
h
ash
v
ec
to
r
izatio
n
to
id
en
tif
y
f
r
a
u
d
u
len
t
UR
L
s
with
9
7
.
5
%
ac
c
u
r
ac
y
.
T
h
ey
cr
ea
te
d
a
Flas
k
-
b
ased
r
ea
l
-
tim
e
UR
L
ca
teg
o
r
izatio
n
web
a
p
p
.
2.
M
E
T
H
O
D
T
h
e
p
r
o
p
o
s
ed
m
eth
o
d
is
s
h
o
wn
in
Fig
u
r
e
1
.
I
n
t
h
is
wo
r
k
,
a
h
y
b
r
id
m
o
d
el
was
d
e
v
elo
p
ed
f
o
r
lu
n
g
ca
n
ce
r
d
etec
tio
n
,
em
p
lo
y
in
g
a
co
m
b
in
atio
n
o
f
c
o
n
v
e
n
tio
n
al
ML
alg
o
r
ith
m
s
an
d
au
to
e
n
co
d
er
s
f
o
r
f
ea
tu
r
e
ex
tr
ac
tio
n
.
T
h
e
d
ataset
co
n
s
i
s
ts
o
f
C
T
s
ca
n
im
ag
es
o
b
tain
ed
f
r
o
m
Kag
g
le,
ca
teg
o
r
ized
in
to
th
r
ee
class
es:
b
en
ig
n
,
m
alig
n
an
t,
an
d
n
o
r
m
al
,
r
ep
r
esen
tin
g
v
ar
io
u
s
lu
n
g
ca
n
ce
r
s
tag
es.
Fig
u
r
e
1
.
Pro
p
o
s
ed
m
o
d
el
f
o
r
p
h
is
h
in
g
web
s
ite
d
etec
tio
n
I
n
th
is
p
ap
er
,
a
co
m
p
r
eh
en
s
iv
e
m
eth
o
d
o
lo
g
y
was
p
r
o
p
o
s
ed
f
o
r
p
h
is
h
in
g
web
s
ite
d
etec
tio
n
,
lev
er
ag
in
g
a
c
o
m
b
in
atio
n
o
f
tr
ad
itio
n
al
s
eq
u
en
tial m
o
d
els
an
d
s
tate
-
of
-
th
e
-
ar
t
lan
g
u
ag
e
r
ep
r
esen
tatio
n
m
o
d
els
s
u
ch
as
B
E
R
T
an
d
XL
Net.
T
h
e
m
eth
o
d
o
lo
g
y
c
o
m
p
r
is
ed
s
ev
er
al
in
ter
co
n
n
ec
ted
s
tep
s
aim
e
d
at
s
y
s
tem
atica
lly
in
v
esti
g
atin
g
th
e
ef
f
icac
y
o
f
d
if
f
er
en
t
DL
ar
ch
itectu
r
es
a
n
d
in
teg
r
atio
n
s
tr
ateg
ies
in
e
n
h
an
cin
g
p
h
is
h
in
g
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
3
6
,
No
.
2
,
No
v
em
b
er
20
24
:
1
2
7
3
-
1
28
3
1276
d
etec
tio
n
p
e
r
f
o
r
m
an
ce
.
A
d
ataset
f
r
o
m
Kag
g
le
was
u
s
ed
f
o
r
ex
p
er
im
en
ts
.
Firstl
y
,
th
e
c
h
allen
g
e
o
f
im
b
alan
c
ed
d
ata
was
ad
d
r
ess
ed
b
y
em
p
lo
y
in
g
tech
n
iq
u
es
s
u
ch
as
s
y
n
th
etic
m
in
o
r
ity
o
v
er
-
s
am
p
lin
g
tech
n
iq
u
e
(
SMOT
E
)
to
en
s
u
r
e
a
b
alan
ce
d
r
e
p
r
esen
t
atio
n
o
f
p
h
is
h
in
g
an
d
leg
itima
te
UR
L
s
in
th
e
d
ataset.
T
h
is
s
tep
was
cr
u
cial
f
o
r
m
itig
atin
g
b
iases
an
d
im
p
r
o
v
in
g
th
e
r
o
b
u
s
tn
ess
o
f
th
e
m
o
d
els.
Su
b
s
eq
u
en
tly
,
th
e
ef
f
ec
tiv
en
ess
o
f
tr
a
d
itio
n
al
s
eq
u
en
tial
m
o
d
els,
i
n
clu
d
in
g
R
NN,
L
STM
,
an
d
GR
U,
in
ca
p
tu
r
in
g
s
eq
u
en
tial
d
ep
en
d
en
ci
es
with
in
UR
L
d
ata
was
ex
p
lo
r
ed
.
T
h
ese
m
o
d
els
wer
e
tr
ain
ed
o
n
th
e
p
r
ep
r
o
ce
s
s
ed
an
d
b
alan
ce
d
d
ataset,
an
d
th
eir
p
er
f
o
r
m
an
ce
was e
v
alu
ated
u
s
in
g
s
tan
d
ar
d
m
etr
ics to
estab
lis
h
b
aselin
e
p
er
f
o
r
m
a
n
ce
.
Nex
t,
th
e
in
teg
r
atio
n
o
f
B
E
R
T
an
d
XL
Net,
two
t
r
an
s
f
o
r
m
er
-
b
ased
lan
g
u
ag
e
r
ep
r
esen
tat
io
n
m
o
d
els
r
en
o
wn
ed
f
o
r
th
ei
r
ab
ilit
y
to
ca
p
tu
r
e
co
n
te
x
tu
al
r
elatio
n
s
h
i
p
s
an
d
s
em
an
tic
u
n
d
e
r
s
tan
d
in
g
in
tex
t
d
ata,
was
in
v
esti
g
ated
.
Pre
-
tr
ain
ed
B
E
R
T
an
d
XL
Net
m
o
d
els
wer
e
f
in
e
-
tu
n
ed
o
n
a
lar
g
e
co
r
p
u
s
o
f
tex
t
d
ata,
an
d
v
ar
io
u
s
in
teg
r
atio
n
ap
p
r
o
ac
h
es
with
th
e
tr
ad
itio
n
al
s
eq
u
en
tial
m
o
d
els
wer
e
ex
p
lo
r
ed
,
s
u
ch
as
co
m
b
in
in
g
o
u
tp
u
t
r
e
p
r
esen
ta
tio
n
s
an
d
em
b
ed
d
in
g
s
.
T
h
is
h
y
b
r
i
d
izatio
n
aim
ed
to
lev
er
a
g
e
b
o
th
th
e
co
n
tex
tu
a
l
u
n
d
er
s
tan
d
i
n
g
o
f
lan
g
u
ag
e
m
o
d
els an
d
th
e
s
eq
u
e
n
tial d
ep
en
d
en
cies c
ap
tu
r
ed
b
y
tr
ad
itio
n
al
m
o
d
els.
Fo
llo
win
g
m
o
d
el
in
teg
r
atio
n
,
th
e
h
y
b
r
i
d
m
o
d
els
wer
e
tr
a
in
ed
an
d
e
v
alu
ated
o
n
a
s
p
lit
d
ataset,
co
m
p
r
is
in
g
tr
ain
in
g
,
v
alid
atio
n
,
an
d
test
s
et
s
.
E
ac
h
m
o
d
el
u
n
d
er
wen
t
tr
ain
in
g
u
s
in
g
a
p
p
r
o
p
r
iate
o
p
tim
izatio
n
alg
o
r
ith
m
s
an
d
h
y
p
e
r
p
ar
am
e
ter
s
,
with
tr
ain
in
g
p
r
o
g
r
ess
m
o
n
ito
r
ed
u
s
in
g
v
alid
atio
n
d
ata
t
o
p
r
ev
en
t
o
v
er
f
itti
n
g
.
T
h
e
tr
ain
ed
m
o
d
e
ls
wer
e
th
en
e
v
alu
ated
o
n
th
e
test
s
et
u
s
in
g
s
tan
d
ar
d
ev
al
u
atio
n
m
etr
ics,
a
n
d
s
tatis
t
ical
s
ig
n
if
ican
ce
test
s
wer
e
co
n
d
u
cted
to
co
m
p
ar
e
th
eir
p
e
r
f
o
r
m
an
ce
ag
ain
s
t
b
a
s
elin
e
m
o
d
els
an
d
s
tan
d
alo
n
e
im
p
lem
e
n
tatio
n
s
o
f
B
E
R
T
an
d
XL
Net.
Fin
ally
,
th
e
r
esu
lts
wer
e
an
aly
ze
d
an
d
in
ter
p
r
ete
d
to
id
en
tify
th
e
m
o
s
t
ef
f
ec
tiv
e
m
o
d
els
an
d
i
n
teg
r
atio
n
s
tr
ateg
ies
f
o
r
p
h
is
h
in
g
d
etec
tio
n
.
I
n
s
ig
h
ts
in
to
th
e
lear
n
ed
r
ep
r
esen
tatio
n
s
an
d
d
ec
is
io
n
-
m
ak
in
g
p
r
o
ce
s
s
es
o
f
th
e
h
y
b
r
id
m
o
d
els
wer
e
g
ain
ed
.
An
y
lim
itatio
n
s
o
r
ch
allen
g
es
en
co
u
n
ter
e
d
d
u
r
i
n
g
th
e
ex
p
er
im
en
tatio
n
p
r
o
ce
s
s
wer
e
d
is
cu
s
s
ed
,
an
d
av
en
u
es
f
o
r
f
u
tu
r
e
r
esear
ch
wer
e
p
r
o
p
o
s
ed
to
a
d
d
r
ess
th
e
m
.
2
.
1
.
P
his
hin
g
URL co
llect
io
n
T
h
e
d
ataset
[
2
5
]
co
n
tain
s
5
4
9
,
3
4
6
s
am
p
les
with
two
co
lu
m
n
s
.
On
e
co
lu
m
n
s
er
v
es
as
a
t
wo
-
ca
teg
o
r
y
p
r
ed
ictio
n
lab
el,
d
is
tin
g
u
is
h
in
g
b
etwe
en
"Go
o
d
"
s
ites
,
wh
ich
ar
e
n
o
t
p
h
is
h
in
g
an
d
c
o
n
tain
n
o
h
ar
m
f
u
l
UR
L
s
,
an
d
"Bad
"
s
ites
,
wh
ich
ar
e
p
h
i
s
h
in
g
s
ites
with
h
ar
m
f
u
l
UR
L
s
.
Ou
t
o
f
th
e
to
tal,
1
5
6
,
4
2
0
U
R
L
s
wer
e
class
if
ied
as b
ad
(
p
h
is
h
in
g
)
,
wh
ile
3
9
2
,
9
2
0
wer
e
ca
teg
o
r
ized
as g
o
o
d
.
2
.
2
.
Cla
s
s
im
ba
la
nce
wit
h SM
O
T
E
C
las
s
im
b
alan
ce
r
ef
e
r
s
to
t
h
e
s
ce
n
ar
io
wh
e
r
e
o
n
e
class
(
o
f
ten
th
e
m
in
o
r
ity
class
)
is
s
ig
n
if
ican
tly
u
n
d
er
r
ep
r
esen
ted
co
m
p
ar
ed
t
o
th
e
o
th
er
class
(
es)
in
a
cl
ass
if
icatio
n
p
r
o
b
lem
.
I
n
th
e
co
n
tex
t
o
f
p
h
is
h
in
g
web
s
ite
d
etec
tio
n
,
class
im
b
al
an
ce
o
cc
u
r
s
wh
en
th
e
n
u
m
b
er
o
f
p
h
is
h
in
g
UR
L
s
is
m
u
ch
lo
wer
th
an
th
e
n
u
m
b
er
o
f
leg
itima
te
UR
L
s
in
th
e
d
ataset.
Ad
d
r
ess
in
g
class
im
b
alan
ce
is
cr
u
cial
b
ec
au
s
e
ML
m
o
d
els
tr
ain
ed
o
n
im
b
alan
ce
d
d
atasets
ten
d
to
ex
h
ib
it
b
iases
to
war
d
s
th
e
m
ajo
r
ity
cla
s
s
,
lead
in
g
to
p
o
o
r
p
er
f
o
r
m
a
n
ce
in
d
etec
tin
g
in
s
tan
ce
s
o
f
th
e
m
in
o
r
ity
class
.
On
e
c
o
m
m
o
n
tech
n
iq
u
e
u
s
ed
to
allev
iate
class
im
b
alan
ce
is
SMOT
E
.
2.
3
.
Dee
p
lea
rning
s
equentia
l a
lg
o
rit
hm
s
I
n
th
is
s
tu
d
y
,
we
lev
er
a
g
e
th
r
e
e
f
o
u
n
d
atio
n
al
DL
alg
o
r
ith
m
s
n
am
ely
R
NN,
L
ST
M
,
an
d
GR
U
f
o
r
th
e
task
o
f
p
h
is
h
in
g
we
b
s
ite
d
etec
tio
n
.
R
NNs
ca
p
tu
r
e
tem
p
o
r
a
l
d
ep
en
d
e
n
cies
in
s
eq
u
en
tial
d
ata
lik
e
UR
L
s
to
an
aly
ze
it.
W
e
u
s
e
d
R
NNs
as
a
b
aselin
e
m
o
d
el
f
o
r
p
h
is
h
in
g
d
etec
tio
n
p
e
r
f
o
r
m
an
ce
m
ea
s
u
r
em
en
ts
.
L
STM
s
ar
e
ad
ep
t
at
m
o
d
elin
g
s
eq
u
e
n
ce
s
with
lo
n
g
-
r
a
n
g
e
d
ep
en
d
e
n
cies,
m
ak
in
g
th
em
well
-
s
u
it
ed
f
o
r
th
e
n
u
an
ce
d
p
atter
n
s
o
f
te
n
f
o
u
n
d
i
n
p
h
is
h
i
n
g
UR
L
s
.
W
e
also
e
x
am
in
e
d
GR
U
n
etwo
r
k
s
,
a
s
im
p
ler
L
STM
,
f
o
r
p
h
is
h
in
g
web
s
ite
id
en
ti
f
icatio
n
.
GR
Us
ca
n
ca
p
tu
r
e
tem
p
o
r
al
r
elati
o
n
s
h
ip
s
wh
ile
b
ein
g
co
m
p
u
t
atio
n
ally
ef
f
icien
t,
o
f
f
er
in
g
an
alter
n
ativ
e
to
L
ST
M
-
b
ased
m
o
d
els.
2.
4
.
P
re
t
ra
ined
m
o
dels
f
o
r
p
his
hi
ng
websi
t
e
det
ec
t
io
n
I
n
th
is
p
ap
er
,
we
d
elv
ed
i
n
to
th
e
u
tili
za
tio
n
o
f
p
r
et
r
a
in
ed
lan
g
u
ag
e
r
e
p
r
esen
tatio
n
m
o
d
els,
s
p
ec
if
ically
B
E
R
T
an
d
XL
Net,
f
o
r
th
e
task
o
f
p
h
is
h
in
g
web
s
ite
d
etec
tio
n
.
A
tr
an
s
f
o
r
m
er
-
b
ased
p
a
r
ad
ig
m
ca
lled
B
E
R
T
ca
p
tu
r
ed
b
id
ir
e
ctio
n
al
co
n
tex
tu
al
lin
k
ag
es
i
n
tex
t
d
ata,
r
e
v
o
lu
tio
n
izin
g
NL
P
.
Pre
tr
ain
ed
o
n
m
ass
iv
e
q
u
an
titi
es
o
f
tex
t
in
p
u
t,
B
E
R
T
en
co
d
ed
lef
t
an
d
r
ig
h
t
co
n
tex
ts
to
b
u
ild
co
m
p
lex
tex
t
s
eq
u
en
ce
r
ep
r
esen
tatio
n
s
.
On
a
d
ataset
o
f
p
h
is
h
in
g
a
n
d
au
th
e
n
tic
UR
L
s
,
we
f
in
e
-
tu
n
ed
p
r
etr
ai
n
ed
B
E
R
T
m
o
d
els
f
o
r
p
h
is
h
in
g
d
etec
tio
n
u
s
i
n
g
co
n
t
ex
tu
al
an
d
s
em
an
tic
k
n
o
wled
g
e.
XL
Net,
an
o
th
er
tr
an
s
f
o
r
m
er
-
b
ased
lan
g
u
ag
e
r
ep
r
esen
tatio
n
a
r
ch
itectu
r
e,
e
n
h
an
ce
d
B
E
R
T
.
XL
Net
u
s
ed
a
p
e
r
m
u
tatio
n
-
b
ased
tr
ain
i
n
g
aim
to
ca
p
tu
r
e
b
id
ir
ec
tio
n
al
co
n
tex
t a
n
d
u
s
e
a
u
to
r
eg
r
ess
iv
e
m
o
d
el
b
en
e
f
its
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
P
h
is
h
in
g
w
eb
s
ite
d
etec
tio
n
u
s
in
g
n
o
ve
l
i
n
teg
r
a
tio
n
o
f
B
E
R
T a
n
d
X
LN
et
…
(
K
o
n
g
a
r
a
S
r
in
i
va
s
a
R
a
o
)
1277
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
3
.
1
.
Appl
y
ing
SM
O
T
E
w
it
h
ph
is
hin
g
web
s
it
e
da
t
a
s
et
T
o
r
ec
tify
th
e
class
im
b
alan
ce
with
in
th
e
d
ataset,
we
em
p
l
o
y
ed
th
e
SMOT
E
.
T
h
is
p
r
o
ce
s
s
in
v
o
lv
ed
g
en
er
atin
g
s
y
n
t
h
etic
in
s
tan
ce
s
f
o
r
th
e
m
in
o
r
it
y
class
,
p
h
is
h
in
g
UR
L
s
,
to
eq
u
alize
th
e
r
ep
r
esen
tatio
n
o
f
b
o
th
class
es
in
th
e
d
ataset.
Af
ter
th
e
ap
p
licatio
n
o
f
SMOT
E
,
th
e
n
u
m
b
er
o
f
b
ad
UR
L
s
am
p
les
ar
e
r
aised
to
m
o
r
e
th
an
3
,
0
0
,
0
0
0
.
I
n
t
h
e
f
in
al
d
ataset
we
co
n
s
id
er
e
d
3
,
0
0
0
0
0
s
am
p
les
b
ad
UR
L
s
an
d
3
,
0
0
0
0
0
s
am
p
les
o
f
g
o
o
d
UR
L
s
.
T
h
is
r
eb
alan
cin
g
o
f
th
e
d
ataset
e
n
ab
les
o
u
r
ML
m
o
d
els
to
lear
n
f
r
o
m
a
m
o
r
e
r
e
p
r
esen
tativ
e
s
et
o
f
ex
am
p
les,
en
h
a
n
cin
g
th
ei
r
ab
il
ity
to
ac
cu
r
ately
d
etec
t p
h
is
h
in
g
UR
L
s
.
3
.2
.
Appl
y
ing
ha
s
hin
g
v
ec
t
o
rize
r
I
n
th
is
p
r
ep
r
o
ce
s
s
in
g
s
tep
,
we
ap
p
lied
th
e
h
ash
in
g
v
ec
to
r
izer
tech
n
iq
u
e
to
co
n
v
er
t
th
e
tex
tu
al
f
ea
tu
r
es,
s
u
ch
as
UR
L
s
o
r
web
p
ag
e
co
n
ten
t,
in
to
n
u
m
er
ical
r
ep
r
esen
tatio
n
s
.
Hash
in
g
v
ec
to
r
izer
was
em
p
lo
y
ed
to
tr
an
s
f
o
r
m
th
e
v
ar
iab
le
-
len
g
t
h
tex
t
d
ata
in
t
o
f
ix
e
d
-
s
ize
n
u
m
e
r
ical
v
ec
t
o
r
s
,
f
ac
ilit
atin
g
th
e
s
u
b
s
eq
u
en
t
ap
p
licatio
n
o
f
DL
m
o
d
els.
B
y
u
tili
zin
g
th
is
m
e
th
o
d
,
we
aim
ed
t
o
p
r
e
p
ar
e
t
h
e
in
p
u
t
d
ata
f
o
r
th
e
u
p
co
m
in
g
DL
m
o
d
els,
n
am
ely
R
NN,
L
STM
,
an
d
GR
U
.
3.
3
.
Appl
y
ing
t
ra
ditio
na
l de
ep
lea
rning
s
eque
ntia
l a
lg
o
rit
hm
s
W
e
ap
p
lied
R
NN,
L
STM
an
d
GR
U
to
th
e
b
alan
ce
d
d
ataset
f
o
r
p
h
is
h
in
g
w
eb
s
ite
d
etec
tio
n
.
Fo
r
th
e
R
NN
m
o
d
el,
we
e
m
p
lo
y
e
d
a
s
im
p
le
ar
ch
itectu
r
e
co
n
s
is
tin
g
o
f
an
em
b
ed
d
in
g
lay
e
r
,
f
o
llo
w
ed
b
y
a
s
in
g
le
R
NN
lay
er
an
d
a
d
en
s
e
o
u
tp
u
t
lay
e
r
with
s
ig
m
o
id
ac
tiv
atio
n
.
T
r
ai
n
in
g
was
co
n
d
u
cte
d
o
n
th
e
b
al
an
ce
d
d
ataset,
with
p
h
is
h
in
g
an
d
leg
i
tim
ate
UR
L
s
eq
u
ally
r
ep
r
esen
ted
.
T
h
e
m
o
d
el
u
tili
ze
d
Ad
am
o
p
tim
izer
an
d
b
in
ar
y
cr
o
s
s
-
en
tr
o
p
y
lo
s
s
d
u
r
in
g
tr
ain
in
g
.
Up
o
n
e
v
alu
atio
n
o
n
a
s
ep
a
r
ate
test
s
et,
th
e
R
NN
m
o
d
el
ac
h
i
ev
ed
a
n
ac
cu
r
ac
y
o
f
9
4
.
5
%.
Nex
t,
we
u
tili
ze
d
a
s
tan
d
ar
d
L
STM
ar
ch
itectu
r
e
c
o
m
p
r
is
in
g
an
em
b
ed
d
i
n
g
lay
er
,
a
s
in
g
le
L
STM
lay
er
,
an
d
a
d
e
n
s
e
o
u
tp
u
t
lay
e
r
with
s
ig
m
o
id
ac
tiv
atio
n
.
Similar
to
th
e
R
NN
m
o
d
el,
tr
ain
in
g
was
p
er
f
o
r
m
e
d
o
n
th
e
b
alan
ce
d
d
ataset
u
s
in
g
Ad
am
o
p
tim
izer
an
d
b
in
ar
y
cr
o
s
s
-
en
tr
o
p
y
lo
s
s
.
T
h
e
L
STM
m
o
d
el
ex
h
ib
ited
im
p
r
o
v
e
d
p
e
r
f
o
r
m
an
ce
,
a
ch
ie
v
in
g
a
n
ac
cu
r
ac
y
o
f
9
6
.
5
%
u
p
o
n
e
v
alu
atio
n
o
n
th
e
test
s
et.
Similar
ly
,
f
o
r
th
e
GR
U
m
o
d
el,
we
em
p
lo
y
ed
a
co
m
p
ar
ab
le
ar
ch
itectu
r
e
to
th
e
L
STM
m
o
d
el,
in
clu
d
in
g
an
em
b
ed
d
in
g
lay
er
,
a
s
in
g
le
GR
U
lay
er
,
an
d
a
d
e
n
s
e
o
u
tp
u
t
lay
er
with
s
ig
m
o
id
ac
tiv
atio
n
.
T
r
ain
in
g
o
f
th
e
GR
U
m
o
d
el
was
co
n
d
u
cte
d
o
n
th
e
b
alan
ce
d
d
at
aset
u
s
in
g
Ad
am
o
p
tim
izer
an
d
b
in
ar
y
cr
o
s
s
-
en
tr
o
p
y
lo
s
s
.
Up
o
n
ev
alu
atio
n
,
th
e
GR
U
m
o
d
el
ac
h
iev
ed
an
ac
c
u
r
ac
y
o
f
9
6
.
1
%
o
n
th
e
test
s
et.
T
ab
le
1
an
d
Fig
u
r
e
2
s
h
o
ws
F1
s
co
r
e
an
d
ac
cu
r
ac
y
o
f
DL
s
eq
u
e
n
tial m
o
d
els.
T
ab
le
1
.
R
esu
lt with
tr
ad
itio
n
a
l seq
u
en
tial m
o
d
les
M
o
d
e
l
A
c
c
u
r
a
c
y
F1
-
sc
o
r
e
R
NN
9
4
.
5
%
9
3
%
L
STM
9
6
.
5
%
9
5
%
GR
U
9
6
.
1
%
9
4
.
8
%
Fig
u
r
e
2
.
P
er
f
o
r
m
an
c
e
o
f
DL
s
eq
u
en
tial m
o
d
els
3.
4
.
I
nitia
lizing
t
he
B
E
RT
m
o
del
Firstl
y
,
a
to
k
e
n
izer
is
in
itiali
ze
d
u
s
in
g
th
e
B
E
R
T
m
o
d
el'
s
to
k
en
ize
r
.
T
h
is
to
k
e
n
izer
is
cr
u
cial
f
o
r
co
n
v
er
tin
g
tex
tu
al
f
ea
tu
r
es
o
f
UR
L
s
o
r
web
p
ag
e
co
n
ten
t
i
n
to
a
f
o
r
m
at
th
at
th
e
B
E
R
T
m
o
d
el
ca
n
co
m
p
r
eh
en
d
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
3
6
,
No
.
2
,
No
v
em
b
er
20
24
:
1
2
7
3
-
1
28
3
1278
T
h
is
m
o
d
el,
p
r
e
-
t
r
ain
ed
o
n
ex
ten
s
iv
e
tex
t
d
ata,
is
ad
ep
t
at
class
if
y
in
g
s
eq
u
en
ce
s
in
to
two
ca
teg
o
r
ies.
I
n
th
e
p
h
is
h
in
g
web
s
ite
d
etec
tio
n
co
n
tex
t,
it
d
is
ce
r
n
s
wh
eth
er
UR
L
s
o
r
web
p
ag
e
co
n
ten
t
ar
e
p
h
is
h
in
g
attem
p
ts
o
r
leg
itima
te.
3.
5
.
I
nitia
lizing
t
he
XL
Net
m
o
del
XL
Net
m
o
d
el
in
itializatio
n
is
s
im
ilar
to
B
E
R
T
,
in
v
o
lv
in
g
lo
ad
in
g
p
r
e
-
tr
ain
e
d
weig
h
ts
an
d
s
ettin
g
u
p
th
e
m
o
d
el
a
r
ch
itectu
r
e.
Pre
-
tr
ain
ed
weig
h
ts
f
o
r
XL
Net
ca
n
b
e
o
b
tain
e
d
f
r
o
m
a
p
u
b
licly
a
v
ailab
le
p
r
e
-
tr
ai
n
ed
XL
Net
m
o
d
el.
T
h
e
ar
ch
itect
u
r
e
o
f
XL
Net
also
co
n
s
is
ts
o
f
m
u
ltip
le
tr
an
s
f
o
r
m
er
la
y
er
s
with
s
el
f
-
atten
tio
n
m
ec
h
an
is
m
s
,
b
u
t
with
a
p
er
m
u
tatio
n
lan
g
u
a
g
e
m
o
d
elin
g
o
b
jectiv
e.
Du
r
in
g
in
itializatio
n
,
ad
ju
s
tm
en
ts
d
o
n
e
m
ad
e
to
f
in
e
-
t
u
n
e
th
e
m
o
d
el
f
o
r
th
e
tar
g
et
task
,
s
u
ch
as
m
o
d
if
y
in
g
h
y
p
er
p
ar
a
m
eter
s
o
r
ad
d
in
g
task
-
s
p
ec
if
ic
lay
er
s
f
o
r
class
if
icatio
n
o
r
r
eg
r
ess
io
n
.
3.
6
.
I
m
ple
m
ent
ing
hy
brid a
rc
hite
ct
ure
wit
h
B
E
RT
I
n
th
is
s
tep
,
we
e
x
p
lo
r
e
d
h
y
b
r
id
a
r
ch
itectu
r
es
th
at
i
n
teg
r
ate
th
e
B
E
R
T
m
o
d
el
with
tr
ad
itio
n
al
s
eq
u
en
tial m
o
d
els.
B
y
co
m
b
in
in
g
th
e
co
n
te
x
tu
al
u
n
d
er
s
tan
d
i
n
g
o
f
B
E
R
T
with
th
e
ab
ilit
y
o
f
s
eq
u
en
tial m
o
d
els
to
ca
p
tu
r
e
tem
p
o
r
al
d
e
p
en
d
e
n
cies,
we
aim
ed
to
en
h
an
ce
th
e
p
er
f
o
r
m
an
ce
o
f
p
h
is
h
in
g
web
s
ite
d
etec
tio
n
.
W
e
in
v
esti
g
ate
d
th
r
ee
h
y
b
r
id
ar
ch
i
tectu
r
es
.
Fig
u
r
e
3
s
h
o
ws th
e
r
e
s
u
lts
o
f
B
E
R
T
with
all
DL
m
o
d
els.
Fig
u
r
e
3
.
Per
f
o
r
m
an
c
e
o
f
B
E
R
T
+D
L
s
eq
u
en
tial m
o
d
els
3.
6
.
1
.
I
m
plem
ent
ing
hy
brid
B
E
R
T
a
nd
RN
N
I
n
th
is
p
h
ase,
we
co
m
b
in
e
d
th
e
o
u
tp
u
t
r
ep
r
esen
tatio
n
s
f
r
o
m
th
e
B
E
R
T
m
o
d
el
with
th
e
h
id
d
en
s
tate
s
o
f
a
R
NN.
T
h
is
h
y
b
r
id
m
o
d
el
lev
er
ag
es
b
o
th
B
E
R
T
'
s
co
n
te
x
tu
al
u
n
d
e
r
s
tan
d
in
g
an
d
R
NN'
s
ab
il
ity
to
ca
p
tu
r
e
s
eq
u
en
tial
p
atter
n
s
i
n
th
e
d
a
ta
.
T
h
e
m
o
d
el
B
E
R
T
+RNN
ac
h
iev
ed
a
n
ac
c
u
r
ac
y
o
f
9
5
%
an
d
a
n
F1
-
s
co
r
e
o
f
9
3
%.
3.
6
.
2
.
I
m
plem
ent
ing
hy
brid
B
E
R
T
a
nd
L
S
T
M
Si
m
ilar
to
B
E
R
T
+
R
NN,
we
co
m
b
in
e
d
t
h
e
o
u
t
p
u
t
r
ep
r
esen
tatio
n
s
f
r
o
m
th
e
B
E
R
T
m
o
d
e
l
with
th
e
h
id
d
en
s
tates
o
f
a
L
STM
n
etwo
r
k
.
T
h
e
L
STM
n
etwo
r
k
en
h
an
ce
s
th
e
B
E
R
T
m
o
d
el
b
y
c
ap
tu
r
in
g
l
o
n
g
-
ter
m
d
ep
en
d
e
n
cies in
th
e
d
ata.
T
h
e
m
o
d
el
B
E
R
T
+L
STM
attain
ed
an
ac
cu
r
ac
y
o
f
9
8
% a
n
d
an
F1
-
s
co
r
e
o
f
9
6
%.
3.
6
.
3
.
I
m
plem
ent
ing
hy
brid
B
E
R
T
a
nd
G
RU
I
n
th
is
h
y
b
r
id
a
r
ch
itectu
r
e,
we
co
m
b
in
e
d
th
e
o
u
tp
u
t
r
ep
r
ese
n
tatio
n
s
f
r
o
m
th
e
B
E
R
T
m
o
d
el
with
th
e
h
id
d
en
s
tates
o
f
a
GR
U
.
GR
U
o
f
f
er
s
a
s
im
p
lifie
d
ar
c
h
itectu
r
e
co
m
p
a
r
ed
to
L
STM
wh
ile
s
till
ca
p
tu
r
in
g
s
eq
u
en
tial
in
f
o
r
m
atio
n
e
f
f
ec
ti
v
ely
.
T
h
e
B
E
R
T
+G
R
U
m
o
d
el
ac
h
iev
ed
an
ac
c
u
r
ac
y
o
f
9
6
.
5
%
an
d
an
F1
-
s
co
r
e
o
f
9
5
%.
3.
7
.
I
m
ple
m
ent
ing
hy
brid a
rc
hite
ct
ure
wit
h
XL
Net
I
n
th
is
s
tep
,
we
ex
p
lo
r
e
d
h
y
b
r
id
ar
ch
itectu
r
es
th
at
in
teg
r
ate
th
e
XL
Net
m
o
d
el
with
t
r
ad
itio
n
al
s
eq
u
en
tial
m
o
d
els.
T
h
ese
ar
ch
itectu
r
es
co
m
b
in
e
XL
Net'
s
p
e
r
m
u
tatio
n
lan
g
u
ag
e
m
o
d
elin
g
o
b
jectiv
e
with
th
e
s
eq
u
en
tial
m
o
d
elin
g
ca
p
ab
ilit
ies
o
f
R
NN,
L
STM
,
an
d
GR
U.
Fig
u
r
e
4
s
h
o
ws
th
e
r
esu
lts
o
f
B
E
R
T
wi
th
all
DL
m
o
d
els.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
P
h
is
h
in
g
w
eb
s
ite
d
etec
tio
n
u
s
in
g
n
o
ve
l
i
n
teg
r
a
tio
n
o
f
B
E
R
T a
n
d
X
LN
et
…
(
K
o
n
g
a
r
a
S
r
in
i
va
s
a
R
a
o
)
1279
3.
7
.
1.
I
m
plem
ent
ing
hy
brid
XL
Net
a
nd
RNN
I
n
th
is
s
tep
,
we
co
m
b
in
e
d
th
e
o
u
tp
u
t r
e
p
r
esen
tatio
n
s
f
r
o
m
t
h
e
XL
Net
m
o
d
el
with
th
e
h
id
d
e
n
s
tates o
f
a
R
NN.
T
h
is
h
y
b
r
id
m
o
d
el
ca
p
italizes
o
n
XL
Net'
s
lan
g
u
ag
e
m
o
d
elin
g
ab
ilit
ies
an
d
R
NN's
s
eq
u
en
tial
m
o
d
elin
g
ca
p
ab
ilit
ies.
T
h
e
m
o
d
el,
XL
Net+
R
NN,
ac
h
iev
ed
a
n
ac
cu
r
ac
y
o
f
9
5
.
6
% a
n
d
an
F1
-
s
co
r
e
o
f
9
5
%
.
Fig
u
r
e
4
.
Per
f
o
r
m
an
c
e
o
f
XL
Net+
DL
s
eq
u
en
tial m
o
d
els
3.
7
.
2
.
I
m
plem
ent
ing
hy
br
id
XL
Net
a
nd
L
ST
M
Similar
to
XL
Net+
R
NN,
we
co
m
b
in
d
e
th
e
o
u
tp
u
t
r
ep
r
esen
tatio
n
s
f
r
o
m
th
e
XL
Net
m
o
d
e
l
with
th
e
h
id
d
en
s
tates
o
f
a
L
STM
n
et
wo
r
k
.
L
STM
'
s
ab
ilit
y
to
ca
p
tu
r
e
lo
n
g
-
ter
m
d
ep
en
d
en
cies
co
m
p
lem
en
ts
XL
Net'
s
p
er
m
u
tatio
n
la
n
g
u
a
g
e
m
o
d
elin
g
o
b
jectiv
e.
T
h
e
XL
Net+
L
STM
m
o
d
el
d
em
o
n
s
tr
ated
an
ac
c
u
r
ac
y
o
f
9
8
.
5
% a
n
d
an
F1
-
s
co
r
e
o
f
9
5
%
.
3
.
7
.
3
.
I
m
plem
ent
ing
hy
b
rid X
L
Net
a
nd
G
RU
I
n
th
is
h
y
b
r
id
a
r
ch
itectu
r
e,
we
co
m
b
i
n
ed
th
e
o
u
tp
u
t
r
ep
r
esen
tatio
n
s
f
r
o
m
th
e
XL
Net
m
o
d
el
with
th
e
h
id
d
en
s
tates
o
f
a
GR
U
n
etwo
r
k
.
GR
U
o
f
f
er
s
a
s
im
p
ler
ar
c
h
itectu
r
e
co
m
p
ar
e
d
to
L
STM
w
h
ile
s
till
ca
p
tu
r
in
g
s
eq
u
en
tial
in
f
o
r
m
a
tio
n
ef
f
ec
t
iv
ely
,
en
h
an
ci
n
g
th
e
ca
p
ab
ili
ties
o
f
XL
Net
f
o
r
p
h
is
h
in
g
web
s
ite
d
etec
tio
n
.
XL
Net+
GR
U
m
o
d
el
ac
h
iev
ed
an
ac
cu
r
ac
y
o
f
9
7
.
5
% a
n
d
an
F1
-
s
co
r
e
o
f
9
7
%.
3
.
8
.
Co
m
pa
riso
n o
f
t
ra
ditio
na
l D
L
m
o
dels
wit
h hy
bri
d m
o
dels
Fig
u
r
e
5
s
h
o
ws
th
e
ac
cu
r
ac
y
an
d
F
1
s
co
r
e
c
o
m
p
a
r
is
o
n
o
f
tr
ad
itio
n
al
DL
m
o
d
els
with
h
y
b
r
id
B
E
R
T
+D
L
an
d
XL
Net+
DL
m
o
d
els.
Fro
m
F
ig
u
r
e
5
,
it
is
o
b
s
er
v
ed
th
at
th
e
ef
f
icac
y
o
f
v
ar
io
u
s
m
o
d
els
in
p
h
is
h
in
g
web
s
ite
d
etec
tio
n
.
T
r
ad
itio
n
al
s
eq
u
e
n
tial
m
o
d
els,
in
clu
d
in
g
R
NN,
L
STM
,
an
d
GR
U,
d
em
o
n
s
tr
ated
r
esp
e
ctab
le
p
er
f
o
r
m
an
ce
s
,
wi
th
L
STM
lead
in
g
t
h
e
p
ac
k
with
a
r
em
ar
k
a
b
le
ac
cu
r
ac
y
o
f
9
6
.
5
0
%
an
d
an
F1
-
s
co
r
e
o
f
9
5
%.
Fig
u
r
e
5
.
Acc
u
r
ac
y
,
F
1
s
co
r
e
co
m
p
ar
is
o
n
o
f
DL
s
eq
u
e
n
tial m
o
d
els with
h
y
b
r
id
m
o
d
els
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
3
6
,
No
.
2
,
No
v
em
b
er
20
24
:
1
2
7
3
-
1
28
3
1280
Ho
wev
er
,
th
e
in
tr
o
d
u
ctio
n
o
f
h
y
b
r
id
ar
ch
itectu
r
es,
co
m
b
in
in
g
th
e
p
o
wer
o
f
p
r
e
-
tr
ain
e
d
lan
g
u
ag
e
m
o
d
els
with
t
h
e
s
eq
u
en
tial
m
o
d
elin
g
ca
p
ab
ilit
ies
o
f
tr
a
d
itio
n
al
m
o
d
els,
s
ig
n
if
ica
n
tly
ele
v
ated
th
e
d
etec
tio
n
ac
cu
r
ac
y
.
Am
o
n
g
t
h
e
h
y
b
r
id
m
o
d
els,
B
E
R
T
+L
STM
em
er
g
ed
as
th
e
to
p
p
e
r
f
o
r
m
er
,
ac
h
iev
in
g
an
im
p
r
ess
iv
e
ac
cu
r
ac
y
o
f
9
8
.
0
0
%
an
d
a
n
F1
-
s
co
r
e
o
f
9
6
%.
T
h
is
u
n
d
er
s
co
r
es
th
e
ad
v
a
n
tag
e
o
f
in
teg
r
atin
g
B
E
R
T
's
co
n
tex
tu
al
u
n
d
er
s
tan
d
i
n
g
with
th
e
lo
n
g
-
ter
m
d
e
p
en
d
e
n
cies c
ap
tu
r
ed
b
y
L
STM
.
Similar
ly
,
XL
Net+
L
STM
s
h
o
wca
s
ed
r
em
ar
k
ab
le
r
esu
lts
,
s
u
r
p
ass
in
g
o
th
er
m
o
d
els
with
th
e
h
ig
h
est
ac
cu
r
ac
y
o
f
9
8
.
5
0
%
an
d
a
c
o
m
m
en
d
a
b
le
F1
-
s
co
r
e
o
f
9
5
%.
T
h
e
h
ig
h
est
ac
cu
r
ac
y
ac
h
iev
ed
am
o
n
g
all
th
e
m
o
d
els
is
9
8
.
5
%,
ac
h
iev
ed
b
y
th
e
XL
Net+
L
STM
m
o
d
el.
T
h
e
h
ig
h
est
F1
-
s
co
r
e
ac
h
iev
ed
is
9
7
%,
o
b
tain
ed
b
y
th
e
u
s
er
m
o
d
el
XL
Net+
GR
U.
T
h
e
p
r
o
p
s
ed
m
o
d
el
p
er
f
o
r
m
a
n
ce
co
m
p
air
o
n
with
e
x
is
tin
g
wo
r
k
s
is
s
h
o
wn
in
T
a
b
le
2
an
d
F
ig
u
r
e
6
.
I
n
co
m
p
ar
is
o
n
to
ex
is
tin
g
m
o
d
els,
th
e
p
r
o
p
o
s
ed
m
eth
o
d
ac
h
iev
es
a
h
ig
h
er
ac
cu
r
ac
y
o
f
9
7
.
4
%,
s
u
r
p
ass
in
g
o
th
er
ap
p
r
o
ac
h
es
s
u
ch
as
A
NN
(
8
8
%),
ML
f
u
s
io
n
(
9
6
%
)
,
an
d
tr
a
d
itio
n
al
ML
(
9
7
%).
T
h
is
en
h
an
ce
m
e
n
t
u
n
d
er
s
co
r
es th
e
ef
f
ec
tiv
en
ess
o
f
h
y
b
r
id
B
E
R
T
an
d
XL
Net
w
ith
tr
ad
itio
n
al
DL
ap
p
r
o
ch
es.
T
ab
le
2
.
C
o
m
p
a
r
is
o
n
with
ex
i
s
tin
g
m
o
d
els
M
o
d
e
l
A
c
c
u
r
a
c
y
A
N
N
[
1
6
]
8
8
%
M
L
f
u
si
o
n
[
2
0
]
9
6
%
Tr
a
d
i
t
i
o
n
a
l
M
L
[
1
0
]
9
7
%
P
r
o
p
o
se
d
m
e
t
h
o
d
9
7
.
4
%
Fig
u
r
e
6
.
Acc
u
r
ac
y
c
o
m
p
ar
is
o
n
with
ex
is
tin
g
m
eth
o
d
s
T
h
e
tr
ain
ed
m
o
d
el
is
also
cr
ea
ted
as
web
ap
p
f
o
r
f
u
r
th
er
u
s
ag
e
o
f
p
r
o
p
o
s
ed
m
o
d
e
in
r
ea
l
ti
m
e.
Her
e,
th
e
u
s
er
ca
n
en
ter
th
e
web
s
ite
ad
d
r
ess
an
d
ca
n
ab
le
ch
ec
k
it
is
p
h
is
h
in
g
s
ite
o
r
n
o
t.
Fig
u
r
e
7
s
h
o
ws
s
cr
ee
n
s
h
o
ts
o
f
p
h
is
h
in
g
web
s
ite
d
etec
tio
n
to
o
l
u
s
ag
e.
I
n
F
ig
u
r
e
7
(
a
)
s
h
o
ws
th
e
to
o
l
wh
er
e
u
s
er
ca
n
ab
le
to
en
ter
web
s
ite
ad
d
r
ess
.
I
n
F
ig
u
r
e
7
(
b
)
,
u
s
er
alr
ea
d
y
en
ter
ed
a
p
h
is
h
in
g
UR
L
an
d
it
is
s
h
o
wn
a
s
p
h
is
h
in
g
UR
L
.
I
n
F
ig
u
r
e
7
(
c
)
,
u
s
er
en
te
r
ed
a
g
o
o
d
UR
L
an
d
it is
s
h
o
wn
as n
o
t
a
p
h
is
h
in
g
web
s
ite.
(
a)
(
b
)
(
c)
Fig
u
r
e
7
.
Ph
is
h
in
g
web
s
ite
d
etec
tio
n
web
ap
p
:
(
a
)
p
h
is
h
in
g
web
s
ite
d
etec
tio
n
web
ap
p
;
(
b
)
u
s
er
en
ter
ed
p
h
is
h
in
g
s
ite
an
d
web
ap
p
o
u
tp
u
t it
as p
h
is
h
in
g
s
ite
; a
n
d
(
c
)
u
s
er
en
ter
ed
g
o
o
d
UR
L
an
d
we
b
ap
p
o
u
tp
u
t it
as
n
o
t a
p
h
is
h
in
g
s
ite
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
P
h
is
h
in
g
w
eb
s
ite
d
etec
tio
n
u
s
in
g
n
o
ve
l
i
n
teg
r
a
tio
n
o
f
B
E
R
T a
n
d
X
LN
et
…
(
K
o
n
g
a
r
a
S
r
in
i
va
s
a
R
a
o
)
1281
4.
CO
NCLU
SI
O
N
I
n
co
n
cl
u
s
io
n
,
o
u
r
s
tu
d
y
d
elv
e
d
in
to
th
e
cr
itical
task
o
f
p
h
is
h
in
g
web
s
ite
d
etec
tio
n
,
aim
in
g
to
d
etec
t
m
alicio
u
s
UR
L
s
.
W
e
in
itiate
d
o
u
r
ex
p
l
o
r
atio
n
b
y
s
cr
u
tin
izin
g
th
e
p
e
r
f
o
r
m
an
ce
o
f
tr
a
d
itio
n
al
s
eq
u
en
tial
m
o
d
els,
in
clu
d
in
g
R
NN,
L
STM
,
an
d
GR
U,
wh
ich
y
ield
e
d
co
m
m
en
d
a
b
le
ac
cu
r
ac
ies
r
an
g
in
g
f
r
o
m
9
5
%
to
9
6
.
5
%.
B
u
ild
in
g
u
p
o
n
th
ese
f
o
u
n
d
atio
n
al
r
esu
lts
,
we
v
en
tu
r
ed
in
to
th
e
r
ea
lm
o
f
h
y
b
r
i
d
ar
ch
itectu
r
es,
m
er
g
i
n
g
th
e
r
o
b
u
s
tn
ess
o
f
tr
a
d
itio
n
al
m
o
d
els
with
th
e
co
n
te
x
tu
al
u
n
d
er
s
tan
d
in
g
o
f
f
er
ed
b
y
s
tate
-
of
-
th
e
-
ar
t
lan
g
u
ag
e
r
ep
r
esen
tatio
n
m
o
d
els,
B
E
R
T
an
d
XL
Net.
T
h
e
in
co
r
p
o
r
atio
n
o
f
th
ese
m
eth
o
d
o
lo
g
ies
r
esu
lted
in
a
s
ig
n
if
ican
t
en
h
an
ce
m
e
n
t
in
d
etec
tio
n
p
er
f
o
r
m
an
ce
,
with
ac
cu
r
a
cies
r
ea
ch
in
g
u
p
to
9
7
.
4
%.
No
tab
ly
,
th
e
h
y
b
r
id
m
o
d
els,
p
ar
ticu
lar
ly
th
o
s
e
in
co
r
p
o
r
ati
n
g
B
E
R
T
an
d
XL
Net,
s
h
o
w
ca
s
ed
r
em
ar
k
a
b
le
ac
cu
r
ac
ies,
u
n
d
e
r
s
co
r
in
g
th
eir
ef
f
icac
y
in
p
h
is
h
in
g
web
s
ite
d
etec
tio
n
task
s
.
T
h
e
f
in
d
in
g
s
em
p
h
asize
th
e
p
ar
am
o
u
n
t
im
p
o
r
tan
ce
o
f
em
b
r
ac
in
g
h
y
b
r
id
DL
ar
ch
itectu
r
es to
f
o
r
t
if
y
cy
b
e
r
s
ec
u
r
ity
p
r
ac
tices e
f
f
ec
tiv
ely
.
RE
F
E
R
E
NC
E
S
[
1
]
S
.
D
e
sai
,
S
.
S
a
l
u
n
k
h
e
,
R
.
D
e
sh
mu
k
h
,
a
n
d
S
.
Z
a
l
t
e
,
“
S
t
u
d
y
a
n
d
a
n
a
l
y
s
i
s
o
f
mac
h
i
n
e
l
e
a
r
n
i
n
g
m
o
d
e
l
s
f
o
r
d
e
t
e
c
t
i
o
n
o
f
p
h
i
s
h
i
n
g
U
R
Ls,
”
M
a
c
h
i
n
e
L
e
a
r
n
i
n
g
A
p
p
l
i
c
a
t
i
o
n
s
,
p
p
.
8
5
-
9
5
,
2
0
2
3
,
d
o
i
:
1
0
.
1
0
0
2
/
9
7
8
1
3
9
4
1
7
3
3
5
8
.
c
h
6
.
[
2
]
D
.
H
e
,
X
.
L
v
,
S
.
Z
h
u
,
S
.
C
h
a
n
a
n
d
K
.
-
K
.
R
.
C
h
o
o
,
"
A
me
t
h
o
d
f
o
r
d
e
t
e
c
t
i
n
g
p
h
i
s
h
i
n
g
w
e
b
si
t
e
s
b
a
s
e
d
o
n
t
i
n
y
-
b
e
r
t
s
t
a
c
k
i
n
g
,
"
i
n
I
EEE
I
n
t
e
r
n
e
t
o
f
T
h
i
n
g
s
J
o
u
r
n
a
l
,
v
o
l
.
1
1
,
n
o
.
2
,
p
p
.
2
2
3
6
-
2
2
4
3
,
1
5
Ja
n
.
1
5
,
2
0
2
4
,
d
o
i
:
1
0
.
1
1
0
9
/
JI
O
T.
2
0
2
3
.
3
2
9
2
1
7
1
.
[
3
]
A
.
M
a
c
i
,
N
.
T
a
mm
a
a
n
d
A
.
C
o
s
c
i
a
,
"
D
e
e
p
r
e
i
n
f
o
r
c
e
me
n
t
l
e
a
r
n
i
n
g
-
b
a
se
d
mal
i
c
i
o
u
s
U
R
L
d
e
t
e
c
t
i
o
n
w
i
t
h
f
e
a
t
u
r
e
se
l
e
c
t
i
o
n
,
"
2
0
2
4
I
EEE
3
r
d
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
AI
i
n
C
y
b
e
rse
c
u
ri
t
y
(
I
C
AI
C
)
,
H
o
u
st
o
n
,
TX
,
U
S
A
,
2
0
2
4
,
p
p
.
1
-
7
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
A
I
C
6
0
2
6
5
.
2
0
2
4
.
1
0
4
3
3
8
2
7
.
[
4
]
M
.
J.
P
i
l
l
a
i
,
S
.
R
e
m
y
a
,
V
.
D
e
v
i
k
a
,
S
.
R
a
mas
u
b
b
a
r
e
d
d
y
a
n
d
Y
.
C
h
o
,
"
E
v
a
s
i
o
n
a
t
t
a
c
k
s
a
n
d
d
e
f
e
n
se
me
c
h
a
n
i
sms
f
o
r
ma
c
h
i
n
e
l
e
a
r
n
i
n
g
-
b
a
s
e
d
w
e
b
p
h
i
s
h
i
n
g
c
l
a
ss
i
f
i
e
r
s
,
"
I
EEE
Ac
c
e
ss
,
v
o
l
.
1
2
,
p
p
.
1
9
3
7
5
-
1
9
3
8
7
,
2
0
2
4
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
3
.
3
3
4
2
8
4
0
.
[
5
]
M
.
A
l
s
a
e
d
i
,
F
.
A
.
G
h
a
l
e
b
,
F
.
S
a
e
e
d
,
J.
A
h
ma
d
a
n
d
M
.
A
l
a
sl
i
,
"
M
u
l
t
i
-
mo
d
a
l
f
e
a
t
u
r
e
s
r
e
p
r
e
se
n
t
a
t
i
o
n
-
b
a
s
e
d
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
m
o
d
e
l
f
o
r
m
a
l
i
c
i
o
u
s
w
e
b
s
i
t
e
d
e
t
e
c
t
i
o
n
,
"
I
E
EE
A
c
c
e
ss
,
v
o
l
.
1
2
,
p
p
.
7
2
7
1
-
7
2
8
4
,
2
0
2
4
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
3
.
3
3
4
8
0
7
1
[
6
]
K
.
M
a
n
g
a
l
a
m
a
n
d
B
.
S
u
b
b
a
,
"
P
h
i
s
h
D
e
t
e
c
t
:
a
B
i
LST
M
b
a
s
e
d
p
h
i
s
h
i
n
g
U
R
L
d
e
t
e
c
t
i
o
n
f
r
a
mew
o
r
k
u
si
n
g
F
a
st
T
e
x
t
e
m
b
e
d
d
i
n
g
s
,
"
2
0
2
4
1
6
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
C
O
Mmu
n
i
c
a
t
i
o
n
S
y
s
t
e
m
s
&
N
ETw
o
rk
S
(
C
O
M
S
N
ETS
)
,
B
e
n
g
a
l
u
r
u
,
I
n
d
i
a
,
2
0
2
4
,
p
p
.
6
3
7
-
6
4
1
,
d
o
i
:
1
0
.
1
1
0
9
/
C
O
M
S
N
E
TS5
9
3
5
1
.
2
0
2
4
.
1
0
4
2
7
0
6
7
.
[
7
]
R
.
A
.
K
,
T.
Zo
n
t
a
a
n
d
M
.
S
a
t
h
i
y
a
n
a
r
a
y
a
n
a
n
,
"
A
h
o
l
i
s
t
i
c
r
e
v
i
e
w
o
n
d
e
t
e
c
t
i
o
n
o
f
mal
i
c
i
o
u
s
b
r
o
w
ser
e
x
t
e
n
s
i
o
n
s
a
n
d
l
i
n
k
s u
s
i
n
g
d
e
e
p
l
e
a
r
n
i
n
g
,
"
2
0
2
4
I
EE
E
3
r
d
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
AI
i
n
C
y
b
e
r
sec
u
r
i
t
y
(
I
C
AI
C
)
,
H
o
u
s
t
o
n
,
TX
,
U
S
A
,
2
0
2
4
,
p
p
.
1
-
6
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
A
I
C
6
0
2
6
5
.
2
0
2
4
.
1
0
4
3
3
8
4
2
.
[
8
]
E.
N
o
w
r
o
o
z
i
,
A
b
h
i
s
h
e
k
,
M
.
M
o
h
a
mm
a
d
i
a
n
d
M
.
C
o
n
t
i
,
"
A
n
a
d
v
e
r
s
a
r
i
a
l
a
t
t
a
c
k
a
n
a
l
y
si
s
o
n
ma
l
i
c
i
o
u
s
a
d
v
e
r
t
i
sem
e
n
t
U
R
L
d
e
t
e
c
t
i
o
n
f
r
a
mew
o
r
k
,
"
I
EEE
T
r
a
n
sa
c
t
i
o
n
s
o
n
N
e
t
w
o
rk
a
n
d
S
e
r
v
i
c
e
Ma
n
a
g
e
m
e
n
t
,
v
o
l
.
2
0
,
n
o
.
2
,
p
p
.
1
3
3
2
-
1
3
4
4
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
TN
S
M
.
2
0
2
2
.
3
2
2
5
2
1
7
.
[
9
]
R
.
R
.
K
.
M
e
n
o
n
a
n
d
V
.
A
n
a
n
d
h
u
,
"
M
a
c
h
i
n
e
l
e
a
r
n
i
n
g
s
u
p
p
o
r
t
e
d
ma
l
i
c
i
o
u
s
U
R
L
d
e
t
e
c
t
i
o
n
,
"
2
0
2
3
4
t
h
I
EEE
G
l
o
b
a
l
C
o
n
f
e
re
n
c
e
f
o
r
Ad
v
a
n
c
e
m
e
n
t
i
n
T
e
c
h
n
o
l
o
g
y
(
G
C
AT)
,
B
a
n
g
a
l
o
r
e
,
I
n
d
i
a
,
2
0
2
3
,
p
p
.
1
-
5
,
d
o
i
:
1
0
.
1
1
0
9
/
G
C
A
T5
9
9
7
0
.
2
0
2
3
.
1
0
3
5
3
4
0
2
.
[
1
0
]
R
.
S
t
o
l
e
r
i
u
,
C
.
N
e
g
r
u
,
B
.
-
C
.
M
o
c
a
n
u
a
n
d
F
.
P
o
p
,
"
M
a
l
i
c
i
o
u
s
s
h
o
r
t
U
R
Ls
d
e
t
e
c
t
i
o
n
t
e
c
h
n
i
q
u
e
,
"
2
0
2
3
2
2
n
d
R
o
Ed
u
N
e
t
C
o
n
f
e
r
e
n
c
e
:
N
e
t
w
o
rk
i
n
g
i
n
E
d
u
c
a
t
i
o
n
a
n
d
R
e
se
a
rc
h
(
Ro
E
d
u
N
e
t
)
,
C
r
a
i
o
v
a
,
R
o
m
a
n
i
a
,
2
0
2
3
,
p
p
.
1
-
6
,
d
o
i
:
1
0
.
1
1
0
9
/
R
o
E
d
u
N
e
t
6
0
1
6
2
.
2
0
2
3
.
1
0
2
7
4
9
1
3
.
[
1
1
]
U
.
S
.
D
.
R
,
A
.
P
a
t
i
l
a
n
d
M
o
h
a
n
a
,
"
M
a
l
i
c
i
o
u
s
U
R
L
d
e
t
e
c
t
i
o
n
a
n
d
c
l
a
ssi
f
i
c
a
t
i
o
n
a
n
a
l
y
s
i
s
u
si
n
g
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
mo
d
e
l
s
,
"
2
0
2
3
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
I
n
t
e
l
l
i
g
e
n
t
D
a
t
a
C
o
m
m
u
n
i
c
a
t
i
o
n
T
e
c
h
n
o
l
o
g
i
e
s
a
n
d
I
n
t
e
r
n
e
t
o
f
T
h
i
n
g
s
(
I
D
C
I
o
T
)
,
B
e
n
g
a
l
u
r
u
,
I
n
d
i
a
,
2
0
2
3
,
p
p
.
4
7
0
-
4
7
6
,
d
o
i
:
1
0
.
1
1
0
9
/
I
D
C
I
o
T
5
6
7
9
3
.
2
0
2
3
.
1
0
0
5
3
4
2
2
.
[
1
2
]
R
.
A
r
o
r
a
,
R
.
G
u
p
t
a
a
n
d
P
.
Y
a
d
a
v
,
"
M
i
s
c
h
i
e
v
o
u
s
U
R
L
p
r
e
d
i
c
t
i
o
n
t
h
r
o
u
g
h
su
p
e
r
v
i
se
d
ma
c
h
i
n
e
l
e
a
r
n
i
n
g
a
l
g
o
r
i
t
h
ms
,
"
2
0
2
3
I
EEE
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
C
o
n
t
e
m
p
o
r
a
ry
C
o
m
p
u
t
i
n
g
a
n
d
C
o
m
m
u
n
i
c
a
t
i
o
n
s
(
I
n
C
4
)
,
B
a
n
g
a
l
o
r
e
,
I
n
d
i
a
,
2
0
2
3
,
p
p
.
1
-
5
,
d
o
i
:
1
0
.
1
1
0
9
/
I
n
C
4
5
7
7
3
0
.
2
0
2
3
.
1
0
2
6
2
8
8
1
.
[
1
3
]
P
.
M
a
l
a
v
i
y
a
,
S
.
B
h
a
d
j
a
,
V
.
G
a
j
j
a
r
a
n
d
Y
.
K
u
m
a
r
,
"
C
y
b
e
r
s
e
c
u
r
i
t
y
a
n
d
d
a
t
a
mi
n
i
n
g
:
d
e
t
e
c
t
i
n
g
ma
l
i
c
i
o
u
s
U
R
L
u
si
n
g
d
a
t
a
mi
n
i
n
g
t
e
c
h
n
i
q
u
e
s
,
"
2
0
2
3
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
C
o
m
m
u
n
i
c
a
t
i
o
n
,
S
e
c
u
ri
t
y
a
n
d
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
(
I
C
C
S
AI
)
,
G
r
e
a
t
e
r
N
o
i
d
a
,
I
n
d
i
a
,
2
0
2
3
,
p
p
.
5
2
7
-
5
3
4
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
C
S
A
I
5
9
7
9
3
.
2
0
2
3
.
1
0
4
2
1
3
4
8
.
[
1
4
]
Y
.
W
a
n
g
,
W
.
Z
h
u
,
H
.
X
u
,
Z.
Q
i
n
,
K
.
R
e
n
a
n
d
W
.
M
a
,
"
A
l
a
r
g
e
-
s
c
a
l
e
p
r
e
t
r
a
i
n
e
d
d
e
e
p
m
o
d
e
l
f
o
r
p
h
i
s
h
i
n
g
U
R
L
d
e
t
e
c
t
i
o
n
,
"
I
C
A
S
S
P
2
0
2
3
-
2
0
2
3
I
E
EE
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
A
c
o
u
st
i
c
s
,
S
p
e
e
c
h
a
n
d
S
i
g
n
a
l
P
ro
c
e
ss
i
n
g
(
I
C
A
S
S
P)
,
R
h
o
d
e
s
I
sl
a
n
d
,
G
r
e
e
c
e
,
2
0
2
3
,
p
p
.
1
-
5
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
A
S
S
P
4
9
3
5
7
.
2
0
2
3
.
1
0
0
9
5
7
1
9
.
[
1
5
]
K
.
S
a
r
a
n
y
a
,
K
.
V
a
r
u
n
v
i
k
a
s
h
,
K
.
K
a
r
t
h
i
k
e
y
a
n
a
n
d
S
.
R
o
s
a
n
a
k
t
h
a
r
,
"
D
e
t
e
c
t
i
n
g
m
a
l
i
c
i
o
u
s
U
R
Ls
u
s
i
n
g
d
a
t
a
a
n
a
l
y
t
i
c
s
,
"
2
0
2
3
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
S
u
st
a
i
n
a
b
l
e
C
o
m
p
u
t
i
n
g
a
n
d
S
m
a
rt
S
y
st
e
m
s
(
I
C
S
C
S
S
)
,
C
o
i
mb
a
t
o
r
e
,
I
n
d
i
a
,
2
0
2
3
,
p
p
.
1
-
6
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
S
C
S
S
5
7
6
5
0
.
2
0
2
3
.
1
0
1
6
9
4
4
7
.
[
1
6
]
C
.
G
.
V
u
n
g
a
n
d
Y
.
Y
.
W
i
n
,
"
U
R
L
c
l
a
ss
i
f
i
c
a
t
i
o
n
b
a
se
d
o
n
l
e
x
i
c
a
l
f
e
a
t
u
r
e
s
b
y
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
,
"
2
0
2
3
I
EE
E
C
o
n
f
e
r
e
n
c
e
o
n
C
o
m
p
u
t
e
r
A
p
p
l
i
c
a
t
i
o
n
s (I
C
C
A)
,
Y
a
n
g
o
n
,
M
y
a
n
mar,
2
0
2
3
,
p
p
.
3
4
5
-
3
5
0
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
C
A
5
1
7
2
3
.
2
0
2
3
.
1
0
1
8
1
5
1
4
.
[
1
7
]
S
.
A
b
a
d
,
H
.
G
h
o
l
a
my
,
a
n
d
M
.
A
s
l
a
n
i
,
“
C
l
a
ssi
f
i
c
a
t
i
o
n
o
f
m
a
l
i
c
i
o
u
s
U
R
Ls
u
s
i
n
g
mac
h
i
n
e
l
e
a
r
n
i
n
g
,
”
S
e
n
s
o
rs
,
v
o
l
.
2
3
,
n
o
.
1
8
.
M
D
P
I
A
G
,
p
.
7
7
6
0
,
S
e
p
.
0
8
,
2
0
2
3
.
d
o
i
:
1
0
.
3
3
9
0
/
s2
3
1
8
7
7
6
0
.
[
1
8
]
Y
.
Li
a
n
g
,
Q
.
W
a
n
g
,
K
.
X
i
o
n
g
,
X
.
Z
h
e
n
g
,
Z
.
Y
u
a
n
d
D
.
Z
e
n
g
,
"
R
o
b
u
st
d
e
t
e
c
t
i
o
n
o
f
ma
l
i
c
i
o
u
s
U
R
Ls
w
i
t
h
s
e
l
f
-
p
a
c
e
d
w
i
d
e
&
d
e
e
p
l
e
a
r
n
i
n
g
,
"
I
EE
E
T
ra
n
s
a
c
t
i
o
n
s
o
n
D
e
p
e
n
d
a
b
l
e
a
n
d
S
e
c
u
re
C
o
m
p
u
t
i
n
g
,
v
o
l
.
1
9
,
n
o
.
2
,
p
p
.
7
1
7
-
7
3
0
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
TD
S
C
.
2
0
2
1
.
3
1
2
1
3
8
8
.
[
1
9
]
J.
Y
u
a
n
,
G
.
C
h
e
n
,
S
.
Ti
a
n
a
n
d
X
.
P
e
i
,
"
M
a
l
i
c
i
o
u
s
U
R
L
d
e
t
e
c
t
i
o
n
b
a
s
e
d
o
n
a
p
a
r
a
l
l
e
l
n
e
u
r
a
l
j
o
i
n
t
m
o
d
e
l
,
"
I
EE
E
Ac
c
e
ss
,
v
o
l
.
9
,
p
p
.
9
4
6
4
-
9
4
7
2
,
2
0
2
1
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
1
.
3
0
4
9
6
2
5
.
[
2
0
]
A
.
La
k
sh
m
a
n
a
r
a
o
,
P
.
S
.
P
.
R
a
o
a
n
d
M
.
M
.
B
.
K
r
i
s
h
n
a
,
"
P
h
i
s
h
i
n
g
w
e
b
si
t
e
d
e
t
e
c
t
i
o
n
u
si
n
g
n
o
v
e
l
mac
h
i
n
e
l
e
a
r
n
i
n
g
f
u
s
i
o
n
a
p
p
r
o
a
c
h
,
"
2
0
2
1
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
Ar
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
a
n
d
S
m
a
r
t
S
y
st
e
m
s
(
I
C
AI
S
)
,
C
o
i
m
b
a
t
o
r
e
,
I
n
d
i
a
,
2
0
2
1
,
p
p
.
1
1
6
4
-
1
1
6
9
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
A
I
S
5
0
9
3
0
.
2
0
2
1
.
9
3
9
5
8
1
0
.
[
2
1
]
T
.
P.
N
a
n
a
w
a
r
e
,
“
M
a
l
i
c
i
o
u
s
U
R
L
d
e
t
e
c
t
i
o
n
u
s
i
n
g
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
I
n
t
e
g
ra
t
e
d
S
c
i
e
n
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
,
v
o
l
.
2
,
n
o
.
1
.
p
p
.
2
9
–
3
6
,
2
0
2
4
.
d
o
i
:
1
0
.
5
9
8
9
0
/
i
j
i
st
.
v
2
i
1
.
1
2
8
9
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
3
6
,
No
.
2
,
No
v
em
b
er
20
24
:
1
2
7
3
-
1
28
3
1282
[
2
2
]
Y
.
C
h
e
n
,
Y
.
Zh
o
u
,
Q
.
D
o
n
g
a
n
d
Q
.
Li
,
"
A
m
a
l
i
c
i
o
u
s
U
R
L
d
e
t
e
c
t
i
o
n
met
h
o
d
b
a
se
d
o
n
C
N
N
,
"
2
0
2
0
I
EE
E
C
o
n
f
e
re
n
c
e
o
n
T
e
l
e
c
o
m
m
u
n
i
c
a
t
i
o
n
s
,
O
p
t
i
c
s
a
n
d
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
(
T
O
C
S
)
,
S
h
e
n
y
a
n
g
,
C
h
i
n
a
,
2
0
2
0
,
p
p
.
2
3
-
2
8
,
d
o
i
:
1
0
.
1
1
0
9
/
TO
C
S
5
0
8
5
8
.
2
0
2
0
.
9
3
3
9
7
6
1
.
[
2
3
]
B
.
G
o
g
o
i
,
T.
A
h
m
e
d
a
n
d
A
.
D
u
t
t
a
,
"
A
H
y
b
r
i
d
a
p
p
r
o
a
c
h
c
o
mb
i
n
i
n
g
b
l
o
c
k
l
i
st
s,
ma
c
h
i
n
e
l
e
a
r
n
i
n
g
a
n
d
d
e
e
p
l
e
a
r
n
i
n
g
f
o
r
d
e
t
e
c
t
i
o
n
o
f
mal
i
c
i
o
u
s
U
R
Ls,
"
2
0
2
2
I
EEE
I
n
d
i
a
C
o
u
n
c
i
l
I
n
t
e
r
n
a
t
i
o
n
a
l
S
u
b
s
e
c
t
i
o
n
s
C
o
n
f
e
re
n
c
e
(
I
N
D
I
S
C
O
N
)
,
B
h
u
b
a
n
e
s
w
a
r
,
I
n
d
i
a
,
2
0
2
2
,
p
p
.
1
-
6
,
d
o
i
:
1
0
.
1
1
0
9
/
I
N
D
I
S
C
O
N
5
4
6
0
5
.
2
0
2
2
.
9
8
6
2
9
0
9
.
[
2
4
]
N
.
S
.
N
a
d
k
a
r
n
i
a
n
d
S
.
B
o
r
k
a
r
,
"
D
e
t
e
c
t
i
o
n
o
f
l
u
n
g
c
a
n
c
e
r
i
n
C
T
i
ma
g
e
s
u
s
i
n
g
i
ma
g
e
p
r
o
c
e
s
si
n
g
,
"
2
0
1
9
3
rd
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
T
re
n
d
s
i
n
El
e
c
t
r
o
n
i
c
s
a
n
d
I
n
f
o
rm
a
t
i
c
s (I
C
O
EI
)
,
2
0
1
9
,
p
p
.
8
6
3
-
8
6
6
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
O
EI
.
2
0
1
9
.
8
8
6
2
5
7
7
.
[
2
5
]
T.
Ti
w
a
r
i
,
“
P
h
i
sh
i
n
g
-
s
ite
-
U
R
L
s
d
a
t
a
set
.
”
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
s:
/
/
w
w
w
.
k
a
g
g
l
e
.
c
o
m/
d
a
t
a
s
e
t
s
/
t
a
r
u
n
t
i
w
a
r
i
h
p
/
p
h
i
s
h
i
n
g
-
s
i
t
e
-
u
r
l
s
(
A
c
c
e
sse
d
:
D
e
c
e
m
b
e
r
1
2
,
2
0
2
3
)
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
Dr
.
K
o
n
g
a
r
a
S
r
i
n
iv
a
sa
Ra
o
is
h
a
v
in
g
1
5
+
y
e
a
rs
o
f
tea
c
h
in
g
e
x
p
e
rien
c
e
,
h
e
re
c
e
iv
e
d
h
is P
h
.
D
.
in
CS
E
fro
m
J
NTUK
Ka
k
in
a
d
a
,
2
0
2
1
,
M
.
Tec
h
in
CS
E
fr
o
m
S
RM
Un
i
v
e
rsity
,
Ch
e
n
n
a
i,
2
0
0
6
a
n
d
B.
Tec
h
i
n
C
S
E
fr
o
m
JN
TU
Hy
d
e
ra
b
a
d
,
2
0
0
4
.
He
h
a
d
p
u
b
l
ish
e
d
p
a
p
e
rs
i
n
re
p
u
ted
n
a
ti
o
n
a
l
a
n
d
i
n
tern
a
ti
o
n
a
l
jo
u
rn
a
ls
.
He
h
a
d
a
tt
e
n
d
e
d
m
a
n
y
wo
r
k
sh
o
p
s,
c
o
n
fe
re
n
c
e
s
a
n
d
p
re
se
n
ted
v
a
ri
o
u
s
re
se
a
rc
h
p
a
p
e
r
s
a
t
n
a
ti
o
n
a
l
a
n
d
i
n
tern
a
ti
o
n
a
l
c
o
n
fe
re
n
c
e
s.
His
tea
c
h
in
g
a
re
a
s
a
re
u
n
ix
sh
e
ll
sc
rip
t
in
g
,
u
n
i
x
p
r
o
g
ra
m
m
in
g
,
n
e
two
r
k
p
ro
g
ra
m
m
in
g
,
c
o
m
p
u
ter n
e
tw
o
rk
s,
d
a
ta b
a
s
e
m
a
n
a
g
e
m
e
n
t
sy
ste
m
s,
o
p
e
ra
ti
n
g
sy
ste
m
s,
d
a
ta
wa
re
h
o
u
sin
g
a
n
d
d
a
ta
m
in
in
g
,
d
a
ta
e
n
g
i
n
e
e
rin
g
,
c
o
m
p
u
ter
p
ro
g
ra
m
m
in
g
.
His
re
se
a
rc
h
in
tere
sts
a
re
i
n
fo
rm
a
ti
o
n
re
tri
e
v
a
l
sy
ste
m
s,
d
a
tab
a
se
m
a
n
a
g
e
m
e
n
t
sy
ste
m
s,
n
a
tu
ra
l
lan
g
u
a
g
e
p
ro
c
e
ss
in
g
,
m
a
c
h
in
e
lea
rn
i
n
g
,
d
a
ta
m
in
i
n
g
,
a
n
d
c
o
m
p
u
te
r
n
e
two
rk
s
.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
d
rk
sra
o
@ifh
e
i
n
d
ia.
o
rg
.
Dr
.
Din
e
sh
V
a
ll
u
r
u
is
a
n
A
ss
o
c
iate
P
ro
fe
ss
o
r,
a
t
th
e
De
p
a
rt
m
e
n
t
o
f
I
n
fo
rm
a
ti
o
n
Tec
h
n
o
l
o
g
y
a
t
M
LRIT
M
E
n
g
in
e
e
rin
g
Co
l
leg
e
,
Tela
n
g
a
n
a
,
In
d
ia.
He
re
c
e
iv
e
d
h
is
P
h
.
D.
fro
m
An
n
a
Un
i
v
e
rsity
,
Ch
e
n
n
a
i
i
n
Co
m
p
u
ter
v
isio
n
.
He
h
a
s
m
o
r
e
th
a
n
1
2
y
e
a
rs
o
f
tea
c
h
in
g
e
x
p
e
rien
c
e
.
He
h
a
s
p
u
b
li
sh
e
d
p
a
p
e
rs
in
re
p
u
ted
n
a
ti
o
n
a
l
a
n
d
i
n
tern
a
ti
o
n
a
l
jo
u
rn
a
ls
.
He
h
a
d
a
tt
e
n
d
e
d
m
a
n
y
wo
r
k
sh
o
p
s,
a
n
d
c
o
n
fe
re
n
c
e
s,
a
n
d
p
re
se
n
ted
v
a
ri
o
u
s
re
se
a
rc
h
p
a
p
e
rs
a
t
n
a
ti
o
n
a
l
a
n
d
in
ter
n
a
ti
o
n
a
l
c
o
n
fe
re
n
c
e
s.
His
a
re
a
s
o
f
in
tere
st
in
c
lu
d
e
a
rti
ficia
l
in
telli
g
e
n
c
e
,
c
o
m
p
u
ter
v
isio
n
,
a
n
d
I
o
T.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
d
i
n
e
sh
.
v
a
ll
u
r
u
1
5
@m
lri
tm.a
c
.
in
.
S
a
tish
k
u
m
a
r
P
a
tn
a
la
re
c
e
iv
e
d
h
is
P
h
.
D
.
fr
o
m
A
n
d
h
ra
Un
i
v
e
rsity
i
n
t
h
e
a
re
a
o
f
M
AN
ET
s.
He
h
a
s
m
o
re
th
a
n
1
0
y
e
a
rs
o
f
tea
c
h
i
n
g
e
x
p
e
rien
c
e
.
He
h
a
d
p
u
b
li
s
h
e
d
p
a
p
e
rs
in
re
p
u
ted
n
a
ti
o
n
a
l
a
n
d
i
n
tern
a
ti
o
n
a
l
jo
u
rn
a
ls
.
He
h
a
d
a
tt
e
n
d
e
d
m
a
n
y
wo
r
k
sh
o
p
s,
c
o
n
fe
re
n
c
e
s
a
n
d
p
re
se
n
ted
v
a
rio
u
s
re
se
a
rc
h
p
a
p
e
rs
a
t
n
a
ti
o
n
a
l
a
n
d
in
tern
a
ti
o
n
a
l
c
o
n
fe
re
n
c
e
s
.
His
a
re
a
s
o
f
in
tere
s
t
in
c
lu
d
e
c
o
m
p
u
ter
n
e
tw
o
rk
s,
a
rti
ficia
l
in
telli
g
e
n
c
e
,
m
a
c
h
in
e
lea
rn
in
g
.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
srtsa
ti
sh
srt@g
m
a
il
.
c
o
m
.
Dr
.
Ra
v
i
Ba
b
u
De
v
a
r
e
d
d
i
is
h
a
v
in
g
1
9
+
y
e
a
rs
tea
c
h
in
g
e
x
p
e
ri
e
n
c
e
a
n
d
p
re
se
n
tl
y
wo
rk
i
n
g
a
s
As
sista
n
t
P
r
o
fe
ss
o
r,
De
p
a
rtme
n
t
o
f
Co
m
p
u
ter
sc
ien
c
e
a
n
d
En
g
i
n
e
e
rin
g
in
G
ITAM
S
c
h
o
o
l
o
f
Tec
h
n
o
l
o
g
y
,
G
ITAM
Un
iv
e
rsity
,
Hy
d
e
ra
b
a
d
,
5
0
2
3
2
9
,
In
d
ia.
He
o
b
tain
e
d
h
is
P
h
.
D
.
(Co
m
p
u
ter
S
c
ien
c
e
a
n
d
E
n
g
in
e
e
rin
g
)
fro
m
Ac
h
a
ry
a
Na
g
a
rju
n
a
U
n
iv
e
rsity
,
G
u
n
tu
r
,
re
c
e
iv
e
d
h
i
s
o
b
tai
n
e
d
M
.
Tec
h
(C
o
m
p
u
ter
S
c
ien
c
e
a
n
d
Tec
h
n
o
l
o
g
y
)
De
g
re
e
fro
m
De
p
a
rtme
n
t
o
f
Co
m
p
u
te
r
S
c
ien
c
e
a
n
d
En
g
i
n
e
e
rin
g
,
An
d
h
ra
Un
iv
e
rsity
i
n
2
0
0
9
,
B.
Te
c
h
(Co
m
p
u
ter
S
c
ien
c
e
a
n
d
En
g
i
n
e
e
rin
g
)
De
g
re
e
in
2
0
0
5
fr
o
m
An
d
h
ra
Un
i
v
e
rsity
.
His
c
u
r
re
n
t
re
se
a
rc
h
in
tere
sts
a
re
in
in
tern
e
t
o
f
t
h
in
g
s,
ima
g
e
p
r
o
c
e
ss
in
g
,
c
o
m
p
u
ter
v
isi
o
n
,
a
n
d
a
rti
f
icia
l
in
telli
g
e
n
c
e
.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
ra
v
i
b
a
b
u
.
d
e
v
a
re
d
d
i@g
m
a
il
.
c
o
m
.
Evaluation Warning : The document was created with Spire.PDF for Python.