I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
pu
t
er
E
ng
ineering
(
I
J
E
CE
)
Vo
l.
15
,
No
.
6
,
Decem
b
er
20
25
,
p
p
.
5
8
6
3
~
5
8
7
8
I
SS
N:
2088
-
8
7
0
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijece.
v
15
i
6
.
pp
5
8
6
3
-
5
8
7
8
5863
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m
Exploring
f
ea
ture
engineering
and
ex
pla
ina
ble
AI fo
r phishing
website
det
e
ction:
a sy
stema
tic
li
ter
a
ture
revi
e
w
No
ra
h Alsu
qa
y
h,
Abd
ulra
h
m
a
n M
irza
,
Are
ej
Alho
g
a
il
I
n
f
o
r
mat
i
o
n
S
y
st
e
ms D
e
p
a
r
t
me
n
t
,
C
o
l
l
e
g
e
o
f
C
o
m
p
u
t
e
r
a
n
d
I
n
f
o
r
ma
t
i
o
n
S
c
i
e
n
c
e
,
K
i
n
g
S
a
u
d
U
n
i
v
e
r
si
t
y
,
R
i
y
a
d
h
,
S
a
u
d
i
A
r
a
b
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Ap
r
1
4
,
2
0
2
5
R
ev
is
ed
Au
g
5
,
2
0
2
5
Acc
ep
ted
Sep
1
4
,
2
0
2
5
De
tec
ti
n
g
p
h
is
h
i
n
g
we
b
si
tes
is
a
ra
p
i
d
ly
e
v
o
l
v
i
n
g
f
iel
d
a
i
m
e
d
a
t
i
d
e
n
ti
fy
i
n
g
a
n
d
m
it
i
g
a
t
i
n
g
c
y
b
e
ra
t
tac
k
s
t
a
rg
e
t
i
n
g
i
n
d
i
v
i
d
u
a
ls
,
o
r
g
a
n
i
z
a
t
i
o
n
s
,
a
n
d
g
o
v
e
r
n
m
e
n
ts
.
O
n
g
o
i
n
g
p
r
o
g
re
ss
i
n
a
rt
if
ic
ial
i
n
tel
li
g
e
n
c
e
(
AI
)
h
a
s t
h
e
p
o
te
n
t
ia
l
to
re
v
o
l
u
ti
o
n
ize
p
h
i
s
h
i
n
g
d
e
tec
ti
o
n
b
y
e
n
h
a
n
c
i
n
g
m
o
d
e
l
a
c
c
u
ra
c
y
a
n
d
im
p
r
o
v
i
n
g
t
ra
n
s
p
a
re
n
c
y
t
h
r
o
u
g
h
e
x
p
lai
n
a
b
le
AI
(X
AI)
.
H
o
we
v
e
r,
si
g
n
i
fica
n
t
c
h
a
ll
e
n
g
e
s
r
e
m
a
i
n
,
p
a
rt
ic
u
l
a
r
l
y
i
n
in
te
g
ra
t
i
n
g
f
e
a
t
u
re
e
n
g
i
n
e
e
r
i
n
g
w
i
th
XA
I
t
o
a
d
d
r
e
ss
s
o
p
h
i
st
ica
te
d
p
h
is
h
i
n
g
s
tra
te
g
ie
s
i
n
c
l
u
d
i
n
g
z
e
r
o
-
d
a
y
a
tt
a
c
k
s
,
th
a
t
e
v
a
d
e
t
ra
d
i
ti
o
n
a
l
d
e
tec
t
io
n
m
e
c
h
a
n
ism
s
.
T
o
o
v
e
rc
o
m
e
t
h
e
se
c
h
a
ll
e
n
g
e
s
,
t
h
is
e
x
a
m
i
n
e
s
t
h
e
im
p
a
c
t
o
f
fe
a
t
u
re
e
n
g
i
n
e
e
r
i
n
g
a
n
d
XA
I
i
n
p
h
is
h
i
n
g
d
e
tec
t
i
o
n
,
e
m
p
h
a
s
iz
i
n
g
t
h
e
ir
a
b
il
it
y
t
o
e
n
h
a
n
c
e
a
c
c
u
ra
c
y
w
h
il
e
p
r
o
v
i
d
i
n
g
in
te
r
p
re
ta
b
i
li
t
y
.
B
y
i
n
te
g
r
a
t
i
n
g
f
e
a
t
u
re
e
x
t
ra
c
ti
o
n
wi
t
h
i
n
te
r
p
re
ta
b
l
e
m
o
d
e
ls
,
th
e
se
tec
h
n
i
q
u
e
s
im
p
ro
v
e
d
e
c
is
io
n
-
m
a
k
i
n
g
t
ra
n
s
p
a
re
n
c
y
a
n
d
s
y
st
e
m
ro
b
u
s
t
n
e
ss
.
T
h
is
p
a
p
e
r
p
re
se
n
ts
t
h
e
f
irs
t
s
y
s
tem
a
ti
c
li
ter
a
t
u
r
e
re
v
i
e
w
(S
LR
)
fo
c
u
si
n
g
o
n
t
h
e
im
p
a
c
t
o
f
fe
a
t
u
r
e
e
n
g
i
n
e
e
ri
n
g
a
n
d
XA
I
o
n
s
tate
-
of
-
t
h
e
-
a
r
t
p
h
i
s
h
i
n
g
d
e
tec
t
i
o
n
a
p
p
r
o
a
c
h
e
s.
Ad
d
it
i
o
n
a
l
l
y
,
i
t
i
d
e
n
ti
f
ies
c
r
i
ti
c
a
l
re
s
e
a
rc
h
g
a
p
s
a
n
d
c
h
a
l
le
n
g
e
s
,
i
n
c
l
u
d
i
n
g
s
c
a
la
b
i
l
it
y
iss
u
e
s
,
t
h
e
e
v
o
l
u
ti
o
n
o
f
p
h
is
h
i
n
g
tec
h
n
i
q
u
e
s
,
a
n
d
b
a
la
n
c
i
n
g
c
o
m
p
l
e
x
i
t
y
w
it
h
i
n
te
r
p
re
ta
b
il
it
y
.
T
h
e
fi
n
d
i
n
g
s
p
r
o
v
i
d
e
v
a
lu
a
b
le
a
c
a
d
e
m
i
c
i
n
s
i
g
h
t
s
w
h
i
le
o
f
fe
r
i
n
g
p
ra
c
t
ica
l
re
c
o
m
m
e
n
d
a
ti
o
n
s
fo
r
d
e
v
e
l
o
p
in
g
a
c
c
u
ra
te
a
n
d
i
n
te
r
p
re
ta
b
le
p
h
is
h
i
n
g
d
e
tec
ti
o
n
s
y
s
te
m
s,
a
i
d
i
n
g
o
r
g
a
n
iza
ti
o
n
s
i
n
s
tr
e
n
g
t
h
e
n
i
n
g
c
y
b
e
rse
c
u
r
i
ty
m
e
a
s
u
re
s
.
K
ey
w
o
r
d
s
:
E
x
p
lain
ab
le
ar
tific
ial
in
tellig
en
ce
Featu
r
e
en
g
in
ee
r
i
n
g
Ma
ch
in
e
lear
n
in
g
Ph
is
h
in
g
d
etec
tio
n
Ph
is
h
in
g
web
s
ites
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
:
No
r
ah
Als
u
q
ay
h
I
n
f
o
r
m
atio
n
Sy
s
tem
s
Dep
ar
tm
en
t,
C
o
lleg
e
o
f
C
o
m
p
u
ter
an
d
I
n
f
o
r
m
atio
n
Scien
ce
Kin
g
Sau
d
Un
iv
e
r
s
ity
R
iy
ad
h
1
1
5
4
3
,
Sau
d
i A
r
ab
ia
E
m
ail:
n
.
alsu
q
ay
h
@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
T
o
d
ay
,
th
e
wid
esp
r
ea
d
u
s
e
o
f
tech
n
o
lo
g
y
in
m
an
y
a
ctiv
ities
co
n
d
u
cted
b
y
in
d
iv
id
u
als
an
d
o
r
g
an
izatio
n
s
h
as
g
r
ea
tly
s
im
p
lifie
d
life
an
d
c
o
n
tr
o
lled
tr
an
s
ac
tio
n
s
a
n
d
th
er
ef
o
r
e
h
as
r
esu
lted
in
a
s
im
u
ltan
eo
u
s
r
is
e
in
th
e
s
o
p
h
is
ticatio
n
an
d
r
ate
o
f
cy
b
er
th
r
ea
ts
[
1
]
.
A
cy
b
er
attac
k
is
d
ef
i
n
ed
as
th
e
m
alicio
u
s
ex
p
lo
itatio
n
o
f
co
m
p
u
ter
n
etwo
r
k
s
,
in
f
o
r
m
atio
n
s
y
s
tem
s
,
an
d
in
f
r
astru
ctu
r
e
[
2
]
.
T
h
is
m
alicio
u
s
v
io
latio
n
o
f
co
m
p
u
tin
g
r
eso
u
r
ce
s
is
ac
co
m
p
lis
h
ed
b
y
u
s
in
g
v
ar
i
o
u
s
m
eth
o
d
s
to
s
teal,
alter
,
o
r
d
e
s
tr
o
y
f
in
a
n
cial
d
ata,
d
is
ab
le
s
y
s
tem
s
an
d
n
etwo
r
k
s
,
an
d
co
m
m
it id
en
tity
th
e
f
t
[
3
]
.
On
e
cy
b
er
th
r
ea
t
is
p
h
is
h
in
g
,
wh
ich
h
as
em
er
g
ed
a
s
ig
n
if
ican
t
co
n
ce
r
n
r
ec
e
n
tly
d
u
e
to
its
in
cr
ea
s
in
g
o
cc
u
r
r
e
n
ce
[
1
]
,
[
2
]
.
Ph
is
h
in
g
em
p
lo
y
s
s
o
cial
en
g
in
ee
r
in
g
an
d
tech
n
ical
m
eth
o
d
s
to
s
te
al
p
er
s
o
n
al
id
en
tity
in
f
o
r
m
atio
n
(
PII
)
an
d
f
in
a
n
cial
cr
ed
en
tials
.
So
cial
en
g
i
n
ee
r
i
n
g
d
ec
ei
v
es
v
ictim
s
in
to
tr
u
s
ti
n
g
th
e
s
o
u
r
ce
wh
ile
d
ir
ec
tin
g
th
em
to
f
r
au
d
u
len
t
web
s
ites
.
A
p
h
is
h
in
g
s
ce
n
ar
io
is
illu
s
tr
ated
in
Fig
u
r
e
1
,
in
wh
ich
a
m
alicio
u
s
ac
to
r
f
ab
r
icate
s
a
web
s
ite
th
at
m
im
ics
a
r
esp
ec
tab
le
an
d
wel
l
-
k
n
o
wn
co
m
p
an
y
,
s
u
ch
as
A
m
az
o
n
.
Nex
t,
u
s
in
g
a
v
ar
iety
o
f
p
latf
o
r
m
s
,
in
clu
d
in
g
s
o
cial
m
ed
ia
an
d
e
-
m
ai
ls
,
th
e
attac
k
er
s
en
d
s
th
e
r
elate
d
lin
k
to
m
an
y
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
6
,
Decem
b
e
r
20
25
:
5
8
6
3
-
5
8
7
8
5864
p
o
s
s
ib
le
tar
g
ets.
I
f
a
v
ictim
f
alls
f
o
r
th
e
s
ca
m
,
th
ey
m
ay
ac
ce
s
s
th
e
f
ak
e
web
s
ite
an
d
p
r
o
v
id
e
v
ital
in
f
o
r
m
atio
n
,
an
d
th
e
u
s
er
’
s
c
r
ed
en
tials
ar
e
ef
f
ec
tiv
ely
ac
q
u
ir
ed
b
y
th
e
attac
k
er
.
T
h
e
att
ac
k
er
th
en
u
s
es
th
e
s
to
len
lo
g
in
cr
e
d
en
tials
to
g
ain
ac
ce
s
s
to
th
e
in
ten
d
ed
we
b
s
ite
an
d
co
m
m
it f
r
au
d
[
4
]
.
Fig
u
r
e
1
.
Step
s
in
v
o
lv
ed
in
a
c
o
m
m
o
n
p
h
is
h
in
g
s
ce
n
ar
io
[
4
]
Ph
is
h
in
g
attac
k
s
ar
e
n
o
t
lim
it
ed
to
in
f
o
r
m
atio
n
co
llectio
n
;
th
ey
ar
e
also
th
e
m
o
s
t
p
o
p
u
l
ar
way
to
d
is
s
em
in
ate
r
an
s
o
m
war
e
an
d
o
th
er
m
alicio
u
s
s
o
f
twar
e.
Acc
o
r
d
in
g
to
s
u
r
v
ey
s
,
th
e
f
i
n
an
cial
d
am
a
g
e
o
f
cy
b
er
cr
im
e
s
u
r
p
ass
ed
$
6
tr
il
lio
n
ea
ch
y
ea
r
,
an
d
is
co
n
ti
n
u
ally
in
cr
ea
s
in
g
.
Fu
r
th
er
m
o
r
e,
d
ata
ab
o
u
t
th
e
B
u
s
in
ess
E
m
ail
C
o
m
p
r
o
m
is
e
(
B
E
C
)
r
ev
ea
led
in
2
0
1
9
t
h
at
p
h
is
h
in
g
ass
au
lts
wer
e
r
esp
o
n
s
ib
le
f
o
r
a
n
asto
u
n
d
in
g
$
2
6
b
illi
o
n
in
f
in
a
n
cial
d
am
ag
es
[
1
]
.
Mo
r
e
o
v
er
,
a
r
ec
o
r
d
was
s
et
f
o
r
p
h
is
h
in
g
in
2
0
2
2
wh
en
th
e
an
ti
-
p
h
is
h
in
g
wo
r
k
in
g
g
r
o
u
p
(
APW
G)
r
ec
o
r
d
ed
m
o
r
e
th
a
n
4
.
7
m
illi
o
n
attac
k
s
,
ac
co
r
d
in
g
to
its
Ph
is
h
in
g
Activ
ity
T
r
en
d
s
R
ep
o
r
t
f
r
o
m
th
e
f
o
u
r
th
q
u
ar
ter
o
f
th
at
y
ea
r
[
2
]
.
Ad
d
itio
n
ally
,
th
e
Oc
to
b
er
2
0
2
2
s
am
p
le
s
h
o
win
g
1
0
1
,
1
0
4
p
h
is
h
in
g
e
-
m
ail
s
u
b
jects wa
s
th
e
lar
g
est s
u
ch
s
am
p
le
th
at
APW
G
h
ad
e
v
er
s
ee
n
.
Acc
o
r
d
in
g
to
r
ec
en
t
s
tu
d
ies,
XAI
co
m
b
in
ed
with
f
ea
t
u
r
e
en
g
in
ee
r
in
g
tech
n
iq
u
es
ca
n
en
h
an
ce
p
h
is
h
in
g
web
s
ite
d
etec
tio
n
s
y
s
tem
s
.
T
h
is
is
b
ec
au
s
e
X
AI
ca
n
p
r
o
v
id
e
b
o
th
ac
cu
r
a
te
p
r
ed
ictio
n
s
an
d
in
ter
p
r
etab
le
in
s
ig
h
ts
in
to
m
o
d
el
b
eh
a
v
io
r
.
XAI
tec
h
n
iq
u
es,
s
u
ch
as
SHAP,
en
ab
le
u
s
er
s
to
r
ea
lize
th
e
im
p
o
r
tan
ce
o
f
in
d
i
v
id
u
al
f
ea
t
u
r
es
(
e.
g
.
,
u
n
if
o
r
m
r
eso
u
r
ce
l
o
ca
to
r
(
UR
L
)
-
b
ased
a
n
d
c
o
n
t
en
t
-
b
ased
)
t
o
m
o
d
el
d
ec
is
io
n
s
,
in
cr
ea
s
in
g
co
n
f
id
e
n
ce
an
d
tr
an
s
p
ar
en
c
y
in
p
h
i
s
h
in
g
d
etec
tio
n
s
y
s
tem
s
.
So
m
e
o
f
th
e
cu
r
r
e
n
t
ap
p
r
o
ac
h
es
h
av
e
b
ee
n
s
ee
n
to
p
r
o
v
id
e
r
ea
s
o
n
ab
le
s
o
lu
tio
n
s
to
th
e
p
r
o
b
lem
o
f
p
h
is
h
in
g
;
h
o
wev
er
,
th
e
y
h
a
v
e
s
o
m
e
d
r
awb
ac
k
s
in
clu
d
in
g
th
e
ab
ilit
y
to
ad
ap
t
to
ze
r
o
-
d
ay
attac
k
s
[
3
]
,
th
e
is
s
u
e
o
f
in
ter
p
r
etab
ilit
y
[
4
]
an
d
th
e
p
r
o
b
lem
s
o
f
d
ea
lin
g
with
im
b
alan
ce
d
d
atasets
an
d
th
e
s
ca
lab
ilit
y
q
u
esti
o
n
.
T
o
o
v
er
co
m
e
t
h
ese
lim
itatio
n
s
an
d
s
in
ce
th
e
n
at
u
r
e
o
f
p
h
is
h
in
g
t
h
r
ea
ts
is
ev
er
ch
an
g
in
g
,
r
esea
r
ch
er
s
m
u
s
t
l
o
o
k
f
o
r
n
ew
wa
y
s
an
d
tech
n
iq
u
es.
So
m
e
o
f
th
e
f
u
tu
r
e
r
ec
o
m
m
en
d
atio
n
s
f
o
r
t
h
e
en
h
a
n
ce
m
en
t
o
f
th
e
p
h
is
h
in
g
d
etec
tio
n
s
y
s
tem
s
in
clu
d
e
in
teg
r
atio
n
o
f
f
ea
tu
r
e
en
g
in
e
er
in
g
with
XAI
f
o
r
p
h
is
h
in
g
d
etec
tio
n
to
ad
d
r
ess
th
ese
ch
allen
g
es.
T
h
ese
ap
p
r
o
ac
h
es
im
p
r
o
v
e
t
h
e
ef
f
ici
en
cy
o
f
d
etec
tio
n
in
ad
d
itio
n
to
o
f
f
er
in
g
im
p
o
r
tan
t
in
f
o
r
m
a
tio
n
r
eg
ar
d
in
g
th
e
d
ec
is
io
n
s
m
ad
e
b
y
th
e
d
etec
tio
n
s
y
s
tem
s
[
5
]
.
T
h
r
o
u
g
h
th
e
i
n
co
r
p
o
r
atio
n
o
f
s
tab
le
f
ea
tu
r
e
s
elec
tio
n
with
th
e
in
ter
p
r
etab
le
m
o
d
els,
th
ey
e
n
h
an
ce
th
e
p
er
f
o
r
m
a
n
ce
o
f
th
e
s
y
s
tem
as we
ll a
s
th
e
tr
u
s
t o
f
th
e
u
s
er
s
.
T
h
is
r
esear
ch
co
n
n
ec
ts
th
e
f
ield
s
o
f
cy
b
er
s
ec
u
r
ity
,
m
ac
h
i
n
e
lear
n
in
g
,
an
d
XAI
b
y
p
r
esen
tin
g
a
s
y
s
tem
atic
r
ev
iew
co
n
ce
n
t
r
ated
o
n
b
o
th
f
ea
tu
r
e
e
n
g
in
ee
r
in
g
an
d
in
ter
p
r
etab
ilit
y
in
p
h
is
h
in
g
d
etec
tio
n
.
T
h
e
p
r
ac
tical
s
ig
n
if
ican
ce
s
o
f
th
is
s
tu
d
y
ar
e
s
ig
n
if
ican
t,
c
o
n
tr
ib
u
tin
g
to
o
n
g
o
in
g
d
is
cu
s
s
io
n
s
in
th
e
f
ield
o
f
s
ec
u
r
e
co
m
p
u
tin
g
s
y
s
tem
s
.
I
ts
r
esu
lt
s
will
b
e
v
alu
ab
le
to
r
esear
c
h
er
s
,
d
ev
elo
p
e
r
s
,
an
d
p
o
licy
m
ak
er
s
,
en
s
u
r
in
g
its
r
elev
an
ce
a
n
d
p
o
te
n
tial
f
o
r
f
u
t
u
r
e
citatio
n
.
T
h
e
r
esear
c
h
s
ee
k
s
to
u
n
d
e
r
s
tan
d
r
ec
en
t
d
ev
el
o
p
m
en
ts
in
p
h
is
h
in
g
web
s
ite
d
etec
tio
n
u
s
in
g
f
ea
tu
r
e
en
g
in
ee
r
in
g
a
n
d
XAI
,
an
aly
zin
g
th
eir
ad
v
an
ta
g
es,
d
r
awb
a
ck
s
,
an
d
p
o
ten
tial
p
ath
s
f
o
r
war
d
f
o
r
d
e
v
elo
p
in
g
ac
cu
r
ate
a
n
d
i
n
ter
p
r
etab
le
d
etec
tio
n
s
y
s
tem
s
.
Mu
ltip
le
s
tu
d
ies
h
av
e
b
ee
n
co
n
d
u
cte
d
o
n
m
an
y
ca
teg
o
r
ie
s
o
f
p
h
is
h
in
g
d
etec
tio
n
s
u
c
h
as
m
ac
h
in
e
lear
n
in
g
an
d
d
ee
p
lear
n
in
g
m
eth
o
d
s
;
y
et,
to
o
u
r
k
n
o
wled
g
e,
th
e
r
e
h
as
b
ee
n
a
s
h
o
r
tag
e
o
f
r
ese
ar
ch
th
at
f
o
cu
s
es
o
n
th
e
co
m
b
in
atio
n
o
f
f
ea
t
u
r
e
en
g
in
ee
r
in
g
m
et
h
o
d
s
with
XAI
to
im
p
r
o
v
e
d
etec
tio
n
ac
c
u
r
ac
y
a
n
d
in
ter
p
r
etab
ilit
y
.
T
h
is
u
n
d
e
r
s
co
r
es
th
e
im
p
o
r
tan
ce
o
f
c
o
n
d
u
ctin
g
d
ee
p
er
in
v
esti
g
atio
n
s
to
an
aly
ze
an
d
ass
ess
th
e
s
ig
n
if
ican
t
f
o
r
im
p
r
o
v
in
g
p
h
is
h
in
g
d
etec
tio
n
s
y
s
tem
s
.
T
h
is
wo
r
k
r
ep
r
esen
ts
th
e
f
ir
s
t
SLR
th
at
co
m
p
r
eh
en
s
iv
ely
ex
p
lo
r
es
th
e
im
p
ac
t
o
f
f
ea
tu
r
e
e
n
g
in
ee
r
in
g
an
d
XAI
o
n
im
p
r
o
v
in
g
th
e
a
cc
u
r
ac
y
an
d
in
ter
p
r
etab
ilit
y
o
f
p
h
is
h
in
g
web
s
ite
d
etec
tio
n
s
y
s
tem
s
.
T
h
e
p
ap
er
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
E
xp
lo
r
in
g
fea
tu
r
e
en
g
in
ee
r
in
g
a
n
d
ex
p
la
in
a
b
le
A
I
fo
r
p
h
is
h
in
g
w
eb
s
ite
d
etec
tio
n
…
(
N
o
r
a
h
A
ls
u
q
a
yh
)
5865
p
r
esen
ts
n
o
v
el
co
n
t
r
ib
u
tio
n
s
,
in
clu
d
in
g
:
th
e
class
if
icatio
n
o
f
p
h
is
h
in
g
d
etec
tio
n
r
esear
ch
u
tili
zin
g
h
y
b
r
id
f
ea
tu
r
e
en
g
i
n
ee
r
in
g
m
et
h
o
d
s
,
an
d
th
e
ev
alu
atio
n
o
f
XAI
'
s
r
o
le
in
elu
cid
atin
g
m
o
d
el
o
u
tp
u
ts
.
T
h
e
r
ev
iew
id
en
tifie
s
r
esear
ch
n
ee
d
s
,
in
cl
u
d
in
g
s
ca
lab
ilit
y
,
th
e
s
h
if
tin
g
n
atu
r
e
o
f
p
h
is
h
in
g
s
tr
ateg
ies,
an
d
th
e
tr
a
d
e
-
o
f
f
b
etwe
en
m
o
d
el
c
o
m
p
lex
ity
a
n
d
in
te
r
p
r
etab
ilit
y
,
s
o
p
r
o
v
id
i
n
g
u
s
ef
u
l
ac
a
d
em
ic
in
s
ig
h
ts
a
n
d
a
f
r
a
m
ewo
r
k
f
o
r
f
u
tu
r
e
r
esear
ch
.
I
t
p
r
o
v
id
es
p
r
ac
tical
d
ir
ec
tio
n
f
o
r
d
ev
elo
p
in
g
p
r
ec
is
e
an
d
tr
a
n
s
p
ar
en
t
p
h
is
h
in
g
d
etec
tio
n
s
y
s
tem
s
,
as
s
is
tin
g
en
ter
p
r
is
es
in
en
h
an
ci
n
g
th
eir
c
y
b
er
s
ec
u
r
ity
f
r
am
ew
o
r
k
s
an
d
ag
g
r
ess
iv
ely
ad
d
r
ess
in
g
ad
v
an
ce
d
p
h
is
h
in
g
attac
k
s
.
T
h
e
o
r
g
an
izatio
n
o
f
th
is
r
ev
iew
is
as
f
o
llo
ws.
I
n
s
ec
tio
n
two
,
a
b
r
ief
b
ac
k
g
r
o
u
n
d
o
n
f
ea
tu
r
e
en
g
in
ee
r
in
g
,
XAI
an
d
p
h
is
h
in
g
d
etec
tio
n
m
eth
o
d
s
is
p
r
esen
ted
.
Sectio
n
t
h
r
ee
v
iews
th
e
m
eth
o
d
o
l
o
g
y
o
f
th
e
s
tu
d
y
wh
ile
s
ec
tio
n
f
o
u
r
co
n
d
u
cts
a
s
y
s
tem
atic
liter
atu
r
e
r
ev
iew
o
f
th
e
s
tate
-
of
-
th
e
-
ar
t
wo
r
k
s
r
elate
d
to
p
h
is
h
in
g
d
etec
tio
n
.
I
n
s
ec
tio
n
f
iv
e,
a
d
is
cu
s
s
io
n
o
f
th
e
m
ain
r
esear
ch
s
tu
d
ies
in
th
is
ar
ea
is
p
r
o
v
id
ed
,
alo
n
g
with
an
ex
p
lo
r
atio
n
o
f
n
ew
ch
allen
g
es.
Sectio
n
s
ix
p
r
esen
ts
th
e
s
u
g
g
ested
d
ir
ec
tio
n
s
f
o
r
f
u
tu
r
e
s
tu
d
ies
b
ased
o
n
th
e
f
i
n
d
in
g
s
a
n
d
f
in
ally
th
e
co
n
clu
s
io
n
.
2.
B
ACK
G
RO
UND
Ph
is
h
in
g
d
etec
tio
n
is
a
cr
u
cial
p
ar
t
o
f
th
e
cy
b
er
s
ec
u
r
ity
d
o
m
ain
with
th
e
g
o
al
o
f
id
e
n
t
if
y
in
g
an
d
p
r
ev
en
tin
g
f
r
au
d
u
len
t
attem
p
t
s
at
s
tealin
g
s
en
s
itiv
e
in
f
o
r
m
a
tio
n
.
Featu
r
e
en
g
in
e
er
in
g
an
d
XAI
ar
e
im
p
o
r
tan
t
co
n
tr
ib
u
to
r
s
to
im
p
r
o
v
in
g
th
e
r
o
b
u
s
tn
ess
an
d
r
eliab
ilit
y
o
f
p
h
is
h
in
g
d
etec
tio
n
a
p
p
r
o
ac
h
es.
Fo
r
s
y
s
tem
s
to
p
r
o
p
er
l
y
d
is
tin
g
u
is
h
b
etwe
en
leg
itima
te
an
d
p
h
is
h
in
g
ac
tiv
ities
,
f
ea
tu
r
es
s
h
o
u
ld
b
e
s
el
ec
ted
,
cr
af
ted
,
an
d
o
p
tim
ized
.
T
h
is
is
s
u
p
p
o
r
ted
b
y
XAI
,
wh
ich
en
s
u
r
es
th
at
th
ese
s
y
s
tem
s
r
em
ain
u
n
d
er
s
tan
d
ab
le
an
d
tr
u
s
two
r
th
y
th
r
o
u
g
h
h
ig
h
lig
h
tin
g
th
e
d
ec
is
io
n
-
m
ak
i
n
g
p
r
o
ce
s
s
.
T
h
er
ef
o
r
e,
f
ea
tu
r
e
en
g
in
ee
r
in
g
a
n
d
XAI
im
p
r
o
v
e
t
h
e
ef
f
icien
c
y
,
ex
p
lain
ab
ilit
y
,
an
d
u
s
er
f
r
ie
n
d
lin
ess
o
f
p
h
is
h
in
g
d
etec
tio
n
s
y
s
tem
s
.
2
.
1
.
F
e
a
t
ure
eng
ineering
Ph
is
h
in
g
d
etec
tio
n
tech
n
iq
u
e
s
im
p
r
o
v
ed
b
y
f
ea
tu
r
e
en
g
in
ee
r
in
g
wh
ich
tr
a
n
s
f
o
r
m
s
r
aw
d
ata
in
to
r
elate
d
f
ea
tu
r
es
th
at
en
h
an
ce
th
e
b
eh
av
i
o
r
o
f
m
o
d
els.
T
o
e
f
f
ec
tiv
ely
d
is
tin
g
u
is
h
b
etwe
en
b
en
ig
n
an
d
f
ak
e
web
s
ites
,
p
h
is
h
in
g
d
etec
tio
n
m
ec
h
an
is
m
s
ca
n
b
e
b
ased
o
n
UR
L
ch
ar
ac
ter
is
tic
s
an
d
co
n
ten
t
-
b
ased
in
d
icato
r
s
th
at
ar
e
d
er
iv
ed
.
Ad
v
a
n
ce
d
tech
n
iq
u
es,
s
u
c
h
as
Gen
etic
Alg
o
r
ith
m
s
an
d
Prin
cip
al
C
o
m
p
o
n
e
n
t
An
aly
s
is
(
PC
A)
ca
n
b
e
u
s
ed
to
f
ilter
th
e
f
ea
tu
r
e
s
p
ac
e
to
m
in
im
ize
t
h
e
d
im
en
s
io
n
ality
an
d
at
th
e
s
am
e
tim
e
p
r
eser
v
e
m
o
s
t
o
f
t
h
e
in
f
o
r
m
atio
n
co
n
te
n
t
[
6
]
.
T
h
is
g
u
ar
an
tees
th
at
ea
ch
o
f
th
e
f
ea
tu
r
es h
as
its
o
wn
u
n
iq
u
e
co
n
tr
i
b
u
tio
n
to
th
e
p
r
ed
ictiv
e
ab
ilit
y
o
f
th
e
m
o
d
el,
b
y
s
elec
tin
g
an
d
elim
in
atin
g
r
ed
u
n
d
an
t
f
ea
tu
r
es
p
r
o
p
er
ly
.
T
h
is
m
ak
es
f
ea
tu
r
e
en
g
in
ee
r
in
g
a
p
o
wer
f
u
l
p
r
o
ce
s
s
f
o
r
en
h
an
ci
n
g
th
e
ac
cu
r
ac
y
an
d
ef
f
icien
cy
o
f
p
h
is
h
in
g
d
etec
tio
n
s
y
s
tem
s
,
esp
ec
ially
wh
en
co
m
p
o
s
ite
f
ea
tu
r
es a
n
d
d
o
m
ain
-
s
p
ec
if
ic
attr
ib
u
tes ar
e
cr
ea
ted
.
2
.
2
.
Art
if
ici
a
l
i
nte
llig
ence
(
AI)
AI
is
a
f
ield
th
at
r
e
p
licates
h
u
m
an
in
tellig
en
ce
to
e
n
ab
le
s
y
s
tem
s
to
d
o
th
in
g
s
lik
e
p
r
ed
ictin
g
f
u
tu
r
e
tr
en
d
s
in
th
e
s
to
ck
m
ar
k
et.
I
t
h
as
r
ep
lace
d
tr
ad
itio
n
al
m
e
th
o
d
s
an
d
is
co
m
p
r
is
ed
o
f
s
u
b
f
ield
s
in
clu
d
in
g
m
ac
h
in
e
lear
n
in
g
an
d
n
atu
r
al
lan
g
u
ag
e
p
r
o
ce
s
s
in
g
th
at
h
a
v
e
alter
ed
th
e
f
ac
e
o
f
in
d
u
s
tr
ies
lik
e
h
ea
lth
ca
r
e,
f
in
an
ce
an
d
au
t
o
n
o
m
o
u
s
s
y
s
tem
s
.
T
h
e
ac
ce
ler
atio
n
in
AI
d
e
v
elo
p
m
en
t
ca
n
b
e
attr
ib
u
te
d
t
o
th
e
r
a
p
id
p
r
o
g
r
ess
m
ad
e
in
r
esear
ch
,
esp
ec
ially
i
n
co
m
p
u
ter
v
is
io
n
(
C
V)
an
d
s
p
ee
ch
r
ec
o
g
n
itio
n
,
wh
ich
h
ig
h
lig
h
t
th
e
im
p
ac
t
o
f
AI
o
n
tr
an
s
f
o
r
m
in
g
th
e
f
u
tu
r
e
o
f
tech
n
o
lo
g
y
an
d
s
o
ciety
.
2
.
3
.
E
x
pla
ina
ble
a
rt
if
icia
l in
t
ellig
ence
(
XAI)
I
n
r
ec
en
t
y
ea
r
s
,
lear
n
in
g
m
o
d
e
ls
r
ev
o
lu
tio
n
ized
th
e
lan
d
s
ca
p
e
o
f
au
to
m
ated
p
r
e
d
ictio
n
an
d
d
ec
is
io
n
-
m
ak
in
g
.
Ar
tific
ial
n
eu
r
al
n
etwo
r
k
s
(
ANN)
an
d
d
ee
p
lear
n
in
g
m
o
d
els h
av
e
p
r
o
v
en
h
ig
h
ly
e
f
f
ec
tiv
e
in
h
an
d
lin
g
co
m
p
lex
task
s
an
d
ac
h
iev
in
g
h
ig
h
p
e
r
f
o
r
m
an
ce
[
7
]
.
Desp
it
e
th
eir
p
er
f
o
r
m
a
n
ce
g
ain
s
,
th
ese
m
o
d
els
ten
d
to
lack
tr
an
s
p
ar
en
c
y
an
d
ar
e
d
if
f
icu
lt
to
in
ter
p
r
et.
I
n
co
r
p
o
r
ati
n
g
in
ter
p
r
etab
ilit
y
as
an
ad
d
i
tio
n
al
lay
er
d
u
r
in
g
m
o
d
el
d
ev
elo
p
m
en
t
ca
n
en
h
a
n
ce
p
r
ac
tical
im
p
lem
e
n
tatio
n
an
d
h
elp
id
e
n
tify
an
d
ad
d
r
e
s
s
d
ef
icien
cies
f
o
r
th
r
ee
k
ey
r
ea
s
o
n
s
[
6
]
:
−
I
t
h
elp
s
e
n
s
u
r
e
in
te
g
r
ity
i
n
d
e
cisi
o
n
-
m
ak
in
g
b
y
en
a
b
lin
g
t
h
e
d
etec
tio
n
an
d
co
r
r
ec
tio
n
o
f
b
iases
p
r
esen
t
in
th
e
tr
ain
in
g
d
ataset.
−
I
t
f
ac
ilit
ates
m
o
d
el
r
o
b
u
s
tn
e
s
s
b
y
id
en
tify
in
g
p
o
te
n
tial
p
er
tu
r
b
atio
n
s
th
at
m
ay
s
ig
n
if
i
ca
n
tly
alter
th
e
m
o
d
el’
s
p
r
e
d
ictio
n
s
.
−
I
t
en
s
u
r
es
th
at
o
n
ly
m
ea
n
in
g
f
u
l
v
ar
iab
les
co
n
tr
ib
u
te
to
th
e
o
u
tp
u
t,
p
r
o
m
o
tin
g
tr
u
th
f
u
l
ca
u
s
ality
an
d
tr
an
s
p
ar
en
cy
i
n
th
e
m
o
d
el’
s
r
e
aso
n
in
g
p
r
o
ce
s
s
.
Ap
p
ly
in
g
XAI
tech
n
i
q
u
es
in
f
ea
tu
r
e
en
g
in
ee
r
in
g
f
o
r
p
h
is
h
in
g
web
s
ite
d
etec
tio
n
is
cr
u
cial
f
o
r
s
ev
er
al
r
ea
s
o
n
s
[
8
]
,
[
9
]
.
First,
it
s
u
p
p
lies
v
is
ib
le
in
s
ig
h
ts
in
to
h
o
w
an
d
wh
y
d
ec
is
io
n
s
ar
e
m
a
d
e
b
e
h
in
d
a
m
o
d
el'
s
p
r
ed
ictio
n
o
f
a
web
s
ite
as
p
h
is
h
in
g
o
r
au
th
e
n
tic.
Seco
n
d
,
r
aises
u
s
er
tr
u
s
t,
wh
ich
i
s
ess
en
tial
f
o
r
th
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
6
,
Decem
b
e
r
20
25
:
5
8
6
3
-
5
8
7
8
5866
ac
ce
p
tan
ce
o
f
AI
-
d
r
i
v
en
s
ec
u
r
ity
m
ea
s
u
r
es.
T
h
ir
d
,
XAI
p
r
o
v
id
es
an
u
n
d
er
s
tan
d
in
g
o
f
h
o
w
d
if
f
er
en
t
f
ea
tu
r
es
s
u
ch
as
UR
L
an
d
Hy
p
e
r
T
ex
t
m
ar
k
u
p
la
n
g
u
a
g
e
(
HT
ML
)
f
ea
tu
r
es
in
f
lu
en
ce
p
r
ed
ictio
n
s
th
at
r
esu
lt
in
m
o
r
e
p
r
ec
is
e
an
d
tr
u
s
two
r
th
y
p
h
is
h
in
g
d
etec
tio
n
.
L
astl
y
,
XAI
id
e
n
tifie
s
th
e
ef
f
ec
t
o
f
in
d
i
v
id
u
al
f
ea
tu
r
es
a
n
d
t
h
eir
in
ter
ac
tio
n
s
an
d
allo
ws
o
p
ti
m
ized
f
ea
tu
r
e
e
n
g
in
ee
r
i
n
g
b
y
f
o
cu
s
in
g
o
n
t
h
e
m
o
s
t
in
f
l
u
en
tial
f
ac
to
r
s
an
d
d
is
ca
r
d
in
g
r
e
d
u
n
d
an
t o
r
ir
r
ele
v
an
t d
ata.
2
.
4
.
P
his
hin
g
det
ec
t
io
n m
et
h
o
ds
Ph
is
h
in
g
d
etec
tio
n
m
eth
o
d
s
em
b
r
ac
e
d
iv
e
r
s
e
tech
n
iq
u
es
an
d
s
tr
ateg
ies
to
id
en
tify
an
d
m
itig
ate
p
h
is
h
in
g
attac
k
s
o
n
web
s
ites
th
at
ar
e
aim
ed
at
s
tealin
g
s
en
s
itiv
e
in
f
o
r
m
atio
n
o
r
c
r
e
d
en
tials
.
T
y
p
ically
,
m
u
ltip
le
tech
n
iq
u
es
ar
e
c
o
m
b
i
n
ed
to
d
etec
t
a
n
d
p
r
ev
en
t
th
is
attac
k
b
ec
au
s
e
p
h
is
h
in
g
is
co
m
p
licated
an
d
th
e
r
e
is
n
o
s
p
ec
if
ic
s
o
lu
tio
n
to
co
m
p
letely
p
r
e
v
en
t
th
is
th
r
ea
t.
Fig
u
r
e
2
illu
s
tr
ates
th
e
p
h
is
h
in
g
d
etec
tio
n
ap
p
r
o
ac
h
es
—
u
s
er
awa
r
en
ess
a
n
d
s
o
f
twar
e
-
b
ased
d
etec
tio
n
.
I
n
th
e
f
o
llo
win
g
s
ec
tio
n
s
,
we
f
o
cu
s
o
n
d
is
cu
s
s
in
g
th
e
s
o
f
twar
e
-
b
ased
tech
n
iq
u
es
in
d
etail.
Fig
u
r
e
2
.
Ph
is
h
in
g
d
etec
tio
n
m
eth
o
d
s
[
1
0
]
–
[
1
2
]
2
.
4
.
1
.
L
is
t
-
ba
s
ed
a
pp
ro
a
ch
L
is
t
-
b
ased
d
etec
tio
n
ca
n
b
e
i
m
p
lem
en
ted
in
two
f
o
r
m
s
:
b
l
ac
k
lis
t
d
etec
tio
n
an
d
wh
itelis
t
d
etec
tio
n
[
1
3
]
,
[
1
4
]
.
T
h
is
ap
p
r
o
ac
h
is
c
h
ar
ac
ter
ized
b
y
ea
s
e
o
f
im
p
le
m
en
tatio
n
an
d
s
tr
o
n
g
o
p
er
ati
o
n
al
ef
f
ec
tiv
e
n
ess
.
Ho
wev
er
,
it
ca
n
n
o
t
ef
f
icien
tl
y
id
en
tify
a
p
h
is
h
in
g
attac
k
d
u
e
to
p
r
o
b
lem
s
with
th
e
u
p
d
ate
m
ec
h
an
is
m
s
o
f
th
ese
lis
t
s
[
1
]
,
wh
ich
r
eq
u
i
r
es
a
lo
t
o
f
h
u
m
a
n
ef
f
o
r
t
an
d
ti
m
e
to
u
p
d
ate
th
e
lis
ts
[
1
0
]
.
T
h
e
m
eth
o
d
f
ails
to
d
etec
t
th
r
ea
ts
f
r
o
m
n
ew
an
d
u
n
k
n
o
wn
UR
L
s
,
th
u
s
m
a
k
in
g
it
p
r
o
n
e
to
ze
r
o
-
d
ay
attac
k
s
[
1
1
]
,
[
1
3
]
.
T
h
e
r
ef
o
r
e
,
th
e
b
lack
-
a
n
d
wh
itelis
t d
etec
tio
n
m
eth
o
d
s
ar
e
cu
r
r
en
tly
less
u
tili
ze
d
.
2
.
4
.
2
.
H
euristic
-
ba
s
ed
a
pp
ro
a
ch
T
h
e
h
eu
r
is
tic
ap
p
r
o
ac
h
ca
n
id
en
tify
s
u
s
p
icio
u
s
co
n
te
n
t
b
ase
d
o
n
i
n
d
icativ
e
cu
es,
th
e
r
eb
y
en
h
an
cin
g
d
etec
tio
n
ef
f
icien
cy
an
d
m
i
n
im
izin
g
p
h
is
h
in
g
-
r
elate
d
l
o
s
s
es
in
a
tim
ely
m
an
n
e
r
.
U
n
lik
e
th
e
lis
t
-
b
ased
ap
p
r
o
ac
h
,
th
is
tech
n
iq
u
e
h
as
a
h
ig
h
lev
el
o
f
p
er
f
o
r
m
an
ce
i
n
d
etec
tin
g
th
r
ea
ts
f
r
o
m
n
ew
an
d
u
n
k
n
o
wn
UR
L
s
[
1
1
]
.
Ho
we
v
er
,
it
o
f
ten
h
as
a
r
elativ
ely
h
ig
h
er
f
alse
p
o
s
itiv
e
r
ate
(
FP
R
)
an
d
ten
d
s
to
b
e
tim
e
-
co
n
s
u
m
in
g
,
as
it
d
ep
en
d
s
o
n
s
ea
r
ch
en
g
i
n
es
an
d
th
ir
d
-
p
ar
ty
s
er
v
ices
s
u
ch
as
DNS
q
u
er
ies
[
1
2
]
.
I
n
ad
d
itio
n
,
th
e
f
o
r
m
u
latio
n
o
f
h
eu
r
is
tic
s
tr
ateg
ies
is
s
u
b
jec
tiv
e
an
d
d
e
p
en
d
s
o
n
ex
p
er
t
k
n
o
wled
g
e
o
r
o
b
s
er
v
a
b
le
p
atter
n
s
in
p
h
is
h
in
g
attem
p
ts
.
T
h
is
tech
n
iq
u
e
is
p
er
f
o
r
m
ed
b
y
c
h
ec
k
in
g
a
w
eb
p
ag
e'
s
co
n
ten
t,
th
e
web
s
ite
UR
L
,
o
r
v
is
u
al
s
im
ilar
ities
.
2
.
4
.
3
.
M
a
chine
l
ea
rning
(
M
L
)
a
pp
ro
a
ch
ML
ap
p
r
o
ac
h
es
f
o
r
d
etec
tin
g
p
h
is
h
in
g
we
b
p
a
g
es
h
av
e
p
r
e
v
io
u
s
ly
b
ee
n
ex
te
n
s
iv
ely
d
is
c
u
s
s
ed
[
1
5
]
,
[
1
6
]
.
Sin
ce
p
h
is
h
in
g
d
etec
tio
n
in
v
o
lv
es
ca
teg
o
r
izin
g
web
p
ag
es
as
eith
er
b
en
ig
n
o
r
p
h
is
h
in
g
,
th
e
m
o
d
els
em
p
lo
y
ed
a
r
e
ty
p
ically
b
i
n
ar
y
class
if
ier
s
[
1
1
]
.
E
ac
h
d
ata
p
o
in
t
in
th
e
in
p
u
t
d
ataset
—
s
u
ch
as
a
UR
L
—
i
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
E
xp
lo
r
in
g
fea
tu
r
e
en
g
in
ee
r
in
g
a
n
d
ex
p
la
in
a
b
le
A
I
fo
r
p
h
is
h
in
g
w
eb
s
ite
d
etec
tio
n
…
(
N
o
r
a
h
A
ls
u
q
a
yh
)
5867
lab
eled
as
eith
er
b
en
ig
n
o
r
p
h
is
h
in
g
to
e
n
ab
le
t
h
e
m
o
d
el
to
lear
n
t
h
e
d
is
tin
g
u
is
h
in
g
f
ea
t
u
r
es
o
f
b
o
th
class
es
[
1
7
]
.
Var
io
u
s
f
ea
t
u
r
e
en
g
in
ee
r
in
g
tech
n
i
q
u
es
ar
e
em
p
lo
y
e
d
t
o
r
ed
u
ce
th
e
n
u
m
b
er
o
f
f
ea
tu
r
es
an
d
en
h
an
ce
th
e
ef
f
icien
cy
an
d
in
ter
p
r
etab
ilit
y
o
f
d
ataset
v
is
u
aliza
tio
n
[
1
8
]
.
Desp
ite
s
u
b
s
tan
tial
p
r
o
g
r
ess
in
th
e
id
en
tific
atio
n
o
f
p
h
is
h
in
g
UR
L
s
u
s
in
g
ML
tech
n
iq
u
es,
s
ev
er
al
cr
itical
ch
allen
g
es
r
em
ain
.
On
e
m
aj
o
r
c
o
n
ce
r
n
lies
in
th
e
s
elec
tio
n
o
f
ef
f
ec
tiv
e
tr
ain
i
n
g
d
atasets
th
at
ac
c
u
r
ately
r
ep
r
esen
t
b
o
th
p
h
is
h
in
g
an
d
b
en
ig
n
web
s
ites
.
R
esear
ch
er
s
m
u
s
t
ca
r
ef
u
lly
b
a
lan
ce
th
e
q
u
a
n
tity
o
f
UR
L
s
u
s
ed
f
o
r
t
r
ain
in
g
with
th
e
co
m
p
u
tatio
n
al
ef
f
icien
cy
an
d
s
ca
lab
ilit
y
o
f
t
h
e
ap
p
lied
ML
alg
o
r
ith
m
s
,
en
s
u
r
in
g
b
o
th
p
e
r
f
o
r
m
an
ce
a
n
d
p
r
ac
ticality
in
r
ea
l
-
wo
r
ld
d
ep
lo
y
m
e
n
t
[
1
9
]
.
An
o
th
er
k
e
y
o
b
s
tacle
is
f
ea
tu
r
in
g
e
x
tr
ac
t
io
n
,
as
m
ac
h
in
e
lea
r
n
in
g
m
o
d
els
ty
p
ically
r
ely
o
n
m
an
u
al
en
g
i
n
ee
r
in
g
o
f
f
ea
tu
r
es
to
ca
p
tu
r
e
r
elev
a
n
t
p
atte
r
n
s
[
1
7
]
.
C
o
llectin
g
ce
r
tain
ty
p
es
o
f
f
ea
tu
r
es,
p
ar
ticu
lar
ly
h
o
s
t
-
b
ased
f
ea
tu
r
es,
is
a
ls
o
tim
e
-
co
n
s
u
m
in
g
,
wh
ich
ca
n
h
in
d
er
th
e
ef
f
icie
n
cy
o
f
th
e
p
h
is
h
in
g
d
etec
tio
n
p
r
o
ce
s
s
[
2
0
]
.
On
e
o
f
th
e
k
e
y
ch
allen
g
es
ass
o
ciate
d
with
h
an
d
cr
a
f
ted
f
ea
tu
r
es
is
th
eir
lim
ited
g
en
er
aliza
b
ilit
y
to
u
n
s
ee
n
d
at
a.
Ad
v
er
s
ar
ies,
s
u
ch
as
p
h
is
h
er
s
,
m
ay
e
x
p
lo
it
th
is
b
y
id
e
n
t
if
y
in
g
th
e
s
p
ec
if
ic
f
ea
tu
r
es a
m
o
d
el
r
elies o
n
an
d
in
ten
tio
n
ally
cr
af
tin
g
UR
L
s
o
r
web
p
ag
es to
ev
a
d
e
d
etec
tio
n
.
2
.
4
.
4
.
Dee
p
l
ea
rning
(
DL
)
a
pp
ro
a
ch
T
h
e
r
o
b
u
s
tn
ess
o
f
DL
alg
o
r
ith
m
s
h
as
en
c
o
u
r
ag
e
d
r
esear
c
h
er
s
to
ex
p
lo
r
e
a
r
an
g
e
o
f
tech
n
iq
u
es
f
o
r
web
s
ite
class
if
icatio
n
,
in
clu
d
i
n
g
th
e
ex
tr
a
ctio
n
o
f
b
o
th
n
o
v
el
an
d
estab
lis
h
ed
f
ea
tu
r
es
—
s
u
ch
as
k
ey
wo
r
d
f
r
eq
u
e
n
cy
with
in
UR
L
s
[
2
1
]
.
I
n
p
h
is
h
in
g
d
etec
tio
n
,
DL
tech
n
iq
u
es
o
f
f
e
r
th
e
p
o
ten
tial
t
o
d
ev
elo
p
d
y
n
am
ic
f
ea
tu
r
e
r
ep
r
esen
tatio
n
s
th
at
ca
n
ad
ap
t
to
co
n
ce
p
t
d
r
if
t
co
m
m
o
n
l
y
o
b
s
er
v
ed
in
p
h
is
h
in
g
d
ata
[
1
1
]
.
DL
alg
o
r
ith
m
s
r
ed
u
ce
th
e
lo
a
d
o
f
f
ea
tu
r
e
ex
tr
ac
tio
n
an
d
s
elec
tio
n
.
I
n
co
n
t
r
ast
to
ML
,
DL
p
r
esen
ts
s
ev
er
al
d
if
f
icu
lties
in
co
n
tr
ast
to
ML
,
it
n
ec
ess
itate
s
a
len
g
th
y
tr
a
in
in
g
p
er
io
d
[
2
2
]
,
[
2
3
]
a
n
d
e
x
ce
s
s
iv
e
co
m
p
u
ter
r
eso
u
r
ce
s
[
2
4
]
.
Fu
r
t
h
er
m
o
r
e,
b
ec
au
s
e
th
ese
m
o
d
els
wo
r
k
a
s
"
B
lack
b
o
x
"
tech
n
iq
u
es,
it
is
d
if
f
icu
lt
to
e
x
p
lain
h
o
w
th
e
m
o
d
el
ar
r
iv
ed
at
a
r
e
s
u
lt
[
2
5
]
.
An
o
t
h
er
p
r
o
b
lem
wi
th
p
h
is
h
in
g
d
etec
tio
n
th
at
h
as
n
'
t
b
ee
n
th
o
r
o
u
g
h
ly
d
is
cu
s
s
ed
y
et
is
r
ea
l
-
tim
e
d
etec
tio
n
[
2
5
]
.
DL
-
b
ased
p
h
is
h
in
g
d
etec
tio
n
m
o
d
els
also
f
ac
e
th
e
p
r
o
b
lem
o
f
o
v
er
f
itti
n
g
,
in
wh
ich
a
m
o
d
el
p
er
f
o
r
m
s
well
o
n
th
e
tr
ain
in
g
d
ata
b
u
t
f
ails
to
g
en
e
r
alize
to
n
ew,
u
n
s
ee
n
d
ata,
s
u
ch
as th
at
r
eq
u
ir
ed
to
d
etec
t p
h
is
h
in
g
web
s
ites
th
at
wer
e
n
o
t p
ar
t o
f
th
e
t
r
ain
in
g
[
2
6
]
.
Als
o
,
th
e
d
atasets
m
ay
co
n
tain
s
o
m
e
d
u
p
licate
p
o
in
ts
,
an
d
it
is
ch
allen
g
in
g
t
o
f
i
n
d
en
o
u
g
h
lab
elle
d
d
ata,
a
n
d
t
h
e
d
is
tr
ib
u
tio
n
o
f
r
ea
l
d
ata
an
d
th
e
d
ataset
m
ig
h
t
b
e
d
if
f
er
en
t,
r
esu
ltin
g
in
th
e
p
o
te
n
tial
r
eq
u
ir
em
e
n
t
f
o
r
a
d
ap
tatio
n
s
.
Mo
s
t
m
alicio
u
s
web
s
ites
ar
e
s
h
o
r
t
-
liv
ed
an
d
ar
e
o
f
ten
o
f
f
lin
e
b
y
th
e
tim
e
t
h
e
y
ar
e
an
aly
ze
d
[
2
7
]
.
2
.
4
.
5
.
H
y
brid ba
s
ed
a
pp
ro
a
ch
Hy
b
r
id
d
etec
tio
n
tech
n
iq
u
es
r
ely
o
n
th
e
in
te
g
r
atio
n
o
f
two
o
r
m
o
r
e
ex
is
tin
g
ap
p
r
o
ac
h
es
t
o
en
h
an
ce
th
e
p
er
f
o
r
m
an
ce
o
f
p
h
is
h
in
g
s
ite
d
etec
tio
n
[
1
2
]
.
Fo
r
ex
am
p
l
e,
co
m
b
in
in
g
h
eu
r
is
tics
an
d
ML
ca
n
h
elp
f
o
r
m
a
b
etter
s
y
s
tem
[
2
8
]
.
An
o
th
er
t
y
p
e
o
f
h
y
b
r
id
m
o
d
el
in
v
o
lv
e
s
th
e
co
m
b
in
atio
n
o
f
m
u
ltip
l
e
m
ac
h
in
e
lear
n
in
g
alg
o
r
ith
m
s
,
wh
e
r
e
th
e
d
ataset
is
in
itially
tr
ain
ed
u
s
in
g
o
n
e
al
g
o
r
ith
m
,
an
d
th
e
r
esu
ltin
g
o
u
t
p
u
t
is
s
u
b
s
eq
u
e
n
tly
f
ed
in
to
a
s
ec
o
n
d
al
g
o
r
ith
m
f
o
r
f
u
r
t
h
er
tr
ain
in
g
[
2
9
]
,
[
3
0
]
.
Fu
r
th
er
m
o
r
e
,
DL
m
eth
o
d
s
c
an
b
e
m
ix
ed
(
e.
g
.
,
cr
ea
tin
g
a
co
n
v
o
lu
tio
n
al
n
e
u
r
al
n
etwo
r
k
(
C
NN)
–
lo
n
g
s
h
o
r
t
-
ter
m
m
em
o
r
y
[
L
STM
]
m
o
d
el
f
o
r
p
h
is
h
in
g
d
etec
tio
n
)
[
3
1
]
.
3.
M
E
T
H
O
DO
L
O
G
Y
T
h
e
s
tu
d
y
'
s
m
ain
g
o
al
is
t
o
s
y
s
tem
atica
lly
an
aly
ze
h
o
w
f
ea
tu
r
e
en
g
in
ee
r
in
g
tech
n
iq
u
es
an
d
ex
p
lain
ab
le
AI
m
eth
o
d
s
m
ig
h
t
en
h
an
ce
p
h
is
h
in
g
web
s
ite
d
etec
tio
n
.
T
h
e
m
eth
o
d
o
lo
g
y
co
m
p
r
is
es
a
P
R
I
SMA
-
g
u
id
ed
s
y
s
tem
atic
r
ev
iew
o
f
r
ec
en
t
s
ch
o
lar
ly
liter
atu
r
e,
in
c
o
r
p
o
r
ati
n
g
q
u
an
titativ
e
an
d
q
u
alitativ
e
ev
alu
atio
n
o
f
ML
an
d
DL
m
o
d
els,
XAI
f
r
am
ewo
r
k
s
(
e
.
g
.
,
SHAP,
L
I
ME
)
,
an
d
h
y
b
r
id
f
ea
tu
r
e
s
elec
tio
n
tech
n
iq
u
es.
I
n
o
r
d
er
to
id
en
tif
y
co
m
m
o
n
th
em
es,
m
eth
o
d
o
lo
g
ical
ad
v
an
ce
m
en
ts
,
an
d
cu
r
r
e
n
t
p
r
o
b
lem
s
,
th
e
r
ev
iew
s
y
n
th
esizes m
o
r
e
th
an
th
ir
t
y
in
v
esti
g
atio
n
s
.
T
h
is
r
esear
ch
u
s
es
an
SLR
m
eth
o
d
o
l
o
g
y
to
d
is
cu
s
s
th
e
r
o
les
o
f
f
ea
tu
r
e
en
g
in
ee
r
in
g
a
n
d
XAI
in
p
h
is
h
in
g
web
s
ite
d
etec
tio
n
b
y
in
v
esti
g
atin
g
th
e
r
ec
en
t
tech
n
iq
u
es
f
o
r
p
h
is
h
in
g
web
s
ite
d
etec
tio
n
.
Mo
r
eo
v
e
r
,
h
o
w
f
ea
tu
r
e
en
g
in
ee
r
in
g
an
d
XAI
ca
n
en
h
an
ce
th
e
ac
cu
r
ac
y
an
d
in
ter
p
r
etab
ilit
y
o
f
p
h
is
h
in
g
web
s
ite
d
etec
tio
n
.
Fin
ally
,
th
e
is
s
u
es
an
d
lim
itatio
n
s
ar
e
ass
o
ciate
d
with
p
h
is
h
in
g
web
s
ite
d
etec
ti
o
n
.
I
n
ad
d
itio
n
,
we
id
en
tifie
d
an
a
p
p
r
o
p
r
iate
d
ata
b
ase
to
d
eliv
er
r
elev
an
t
r
esu
lt
s
th
at
ar
e
lim
ited
to
a
5
-
y
ea
r
p
er
io
d
b
etwe
en
2
0
1
9
an
d
2
0
2
4
lo
ca
ted
in
AC
M
Di
g
ital,
I
E
E
E
E
x
p
lo
r
e,
E
ls
ev
ier
,
Sp
r
in
g
er
,
MD
PI
an
d
Go
o
g
le
Sch
o
lar
.
L
iter
atu
r
e
lim
ited
to
r
ev
iew
a
r
ticles,
co
n
f
er
en
ce
p
r
o
ce
e
d
in
g
s
a
n
d
r
esear
ch
er
s
’
th
eses
.
T
o
id
en
tify
r
elev
an
t
s
tu
d
ies
an
d
n
ar
r
o
w
d
o
w
n
th
e
n
u
m
b
e
r
o
f
r
esu
lts
in
clu
d
ed
in
th
is
r
ev
iew,
we
f
o
llo
wed
th
e
s
y
s
tem
ati
c
r
ev
iew
p
r
o
ce
s
s
a
s
illu
s
tr
ated
in
Fig
u
r
e
3
.
T
h
e
r
e
v
iew
p
r
o
ce
s
s
was
d
iv
id
ed
in
t
o
th
r
ee
s
eq
u
e
n
tial
s
tep
s
:
id
en
tific
atio
n
,
s
cr
ee
n
in
g
an
d
s
elec
tio
n
.
T
h
e
f
o
llo
win
g
s
ea
r
ch
s
tr
in
g
was
u
s
ed
to
r
etr
iev
e
r
elev
a
n
t
ar
ticles:
(
“Fea
tu
r
e
en
g
in
ee
r
in
g
”)
OR
(
“XA
I
”
OR
“e
x
p
lain
ab
le
AI
”
OR
“e
x
p
lain
ab
le
ar
tific
ial
in
tellig
en
ce
”)
AND
(
“p
h
is
h
in
g
d
etec
tio
n
”
OR
“p
h
is
h
in
g
we
b
s
ite
d
etec
tio
n
”
)
.
Fro
m
th
e
i
n
itial
s
ea
r
ch
,
1
0
2
p
a
p
er
s
th
at
in
v
o
l
v
ed
f
ea
t
u
r
e
en
g
in
ee
r
in
g
f
o
r
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
6
,
Decem
b
e
r
20
25
:
5
8
6
3
-
5
8
7
8
5868
p
h
is
h
in
g
web
s
ite
d
etec
tio
n
w
er
e
o
b
tain
ed
.
I
n
th
e
s
cr
ee
n
in
g
p
r
o
ce
s
s
,
6
0
s
tu
d
ies
th
at
wer
e
n
o
t
in
co
n
f
o
r
m
ity
with
th
e
r
eq
u
ir
em
en
ts
wer
e
ex
clu
d
ed
.
I
n
th
e
f
in
al
s
elec
ti
o
n
s
tag
e,
we
in
clu
d
ed
3
4
p
a
p
er
s
th
at
m
et
th
e
in
clu
s
io
n
cr
iter
ia
f
o
r
th
is
s
y
s
tem
atic
liter
atu
r
e
r
ev
iew.
Fig
u
r
e
3
.
Ph
ases
o
f
SLR s
elec
tio
n
p
r
o
ce
s
s
.
N
d
en
o
tes th
e
n
u
m
b
er
o
f
p
ap
e
r
s
at
ea
ch
s
tag
e
4.
RE
L
AT
E
D
WO
RK
S
Ph
is
h
in
g
d
etec
tio
n
ap
p
r
o
ac
h
e
s
,
as
d
is
cu
s
s
ed
in
s
ec
tio
n
two
,
ca
n
b
e
im
p
r
o
v
ed
b
y
co
m
b
in
i
n
g
f
ea
tu
r
e
en
g
in
ee
r
in
g
an
d
XAI
tech
n
i
q
u
es.
Featu
r
e
en
g
in
ee
r
in
g
p
er
m
its
s
y
s
tem
s
to
s
elec
t
th
e
m
o
s
t
r
elev
an
t
attr
ib
u
tes,
wh
ich
r
aise
d
etec
tio
n
ac
cu
r
a
cy
.
Mo
r
eo
v
er
,
in
teg
r
atin
g
X
AI
to
o
ls
,
s
u
ch
as
S
h
ap
ley
a
d
d
itiv
e
ex
p
lan
atio
n
s
(
SHAP)
o
r
lo
ca
l
in
ter
p
r
etab
le
m
o
d
el
-
ag
n
o
s
tic
ex
p
lan
atio
n
s
(
L
I
ME
)
,
s
u
p
p
l
y
in
ter
p
r
et
ab
ilit
y
to
en
h
an
ce
d
ec
is
io
n
-
m
ak
in
g
p
r
o
ce
s
s
es,
en
co
u
r
a
g
e
tr
u
s
t
a
m
o
n
g
u
s
er
s
a
n
d
s
tak
eh
o
ld
er
s
.
T
h
is
s
ec
tio
n
is
d
iv
id
ed
i
n
to
f
i
v
e
s
u
b
-
s
ec
tio
n
s
co
r
r
esp
o
n
d
in
g
to
th
e
r
ec
en
t
p
h
is
h
in
g
d
etec
tio
n
tech
n
iq
u
es
o
u
tlin
ed
i
n
s
ec
tio
n
two
,
co
m
b
in
ed
with
f
ea
tu
r
e
en
g
in
ee
r
in
g
m
et
h
o
d
s
.
I
t
also
d
em
o
n
s
tr
ates
h
o
w
th
e
in
teg
r
atio
n
o
f
f
ea
tu
r
e
e
n
g
in
ee
r
in
g
a
n
d
XAI
im
p
r
o
v
es
p
h
is
h
in
g
d
etec
tio
n
m
o
d
els
b
y
ad
d
r
ess
in
g
ch
all
en
g
es
lik
e
a
d
ap
tab
ilit
y
t
o
ev
o
lv
in
g
th
r
ea
ts
an
d
b
alan
cin
g
co
m
p
lex
ity
with
in
ter
p
r
etab
ilit
y
.
Fin
ally
,
th
is
s
ec
tio
n
h
ig
h
lig
h
ts
th
e
lim
itatio
n
s
an
d
u
n
r
eso
lv
e
d
is
s
u
es a
s
s
o
ciate
d
with
th
ese
ap
p
r
o
ac
h
es,
p
av
in
g
th
e
way
f
o
r
f
u
tu
r
e
r
esear
ch
an
d
p
r
ac
tical
ad
v
an
ce
m
en
ts
.
4
.
1
.
L
is
t
-
ba
s
ed
a
pp
ro
a
ch
Stu
d
y
[
1
6
]
in
tr
o
d
u
ce
d
th
e
au
t
o
m
ated
in
d
iv
id
u
al
wh
itelis
t
,
a
u
n
iq
u
e
an
ti
-
p
h
is
h
in
g
s
tr
ateg
y
b
ased
o
n
th
e
Naïv
e
B
ay
es
(
NB
)
class
if
ier
.
B
y
l
o
g
g
in
g
t
h
e
I
P
ad
d
r
ess
es
o
f
all
well
-
k
n
o
wn
l
o
g
in
u
s
er
in
ter
f
ac
es
(
L
UI
s
)
th
at
th
e
u
s
er
h
as
v
is
ited
,
t
h
is
tech
n
iq
u
e
cr
ea
tes
a
c
u
s
to
m
ized
wh
itelis
t.
T
h
e
s
y
s
tem
cr
ea
tes
a
war
n
in
g
ab
o
u
t
a
p
o
s
s
ib
le
p
h
is
h
in
g
attem
p
t
wh
en
th
e
u
s
er
tr
ies
to
s
en
d
p
r
iv
ate
in
f
o
r
m
atio
n
to
a
L
UI
th
at
is
n
o
t
o
n
th
e
wh
itelis
t.
I
n
co
n
tr
ast,
Stu
d
y
[
3
2
]
ev
al
u
ated
th
eir
s
u
g
g
ested
b
lack
lis
t
-
b
ased
ap
p
r
o
ac
h
u
s
in
g
a
s
et
o
f
3
8
ch
ar
ac
ter
is
tics
an
d
r
an
d
o
m
f
o
r
est
(
R
F)
an
d
lin
ea
r
r
eg
r
ess
io
n
(
L
R
)
class
if
ier
s
.
T
h
e
m
eth
o
d
s
u
cc
ess
f
u
lly
d
is
tin
g
u
is
h
ed
b
etwe
en
f
r
au
d
u
len
tly
r
eg
is
ter
ed
d
o
m
ai
n
s
an
d
v
alid
o
n
es with
a
9
7
% a
cc
u
r
ac
y
r
ate
an
d
a
2
.
5
% FP
R
.
I
n
o
r
d
e
r
to
im
p
r
o
v
e
th
e
d
etec
tio
n
ac
cu
r
ac
y
o
f
p
h
is
h
in
g
att
ac
k
s
,
B
ar
r
ac
lo
u
g
h
et
a
l.
[
3
3
]
co
m
b
in
ed
h
eu
r
is
tic
m
eth
o
d
s
,
web
co
n
te
n
t
an
aly
s
is
,
an
d
b
lack
lis
ts
in
a
m
ac
h
in
e
lear
n
in
g
f
r
am
ew
o
r
k
th
at
m
ad
e
u
s
e
o
f
ex
ten
s
iv
e
f
ea
tu
r
e
s
ets.
T
h
e
m
ac
h
in
e
lear
n
in
g
alg
o
r
ith
m
s
th
at
wer
e
ass
ess
ed
wer
e
J
4
8
,
J
R
ip
,
NB
,
PAR
T
,
an
d
th
e
ad
ap
tiv
e
n
eu
r
o
-
f
u
zz
y
in
f
e
r
en
ce
s
y
s
tem
(
ANFI
S).
T
h
e
b
est
p
er
f
o
r
m
a
n
ce
was
g
iv
en
b
y
PAR
T
,
wh
ich
h
ad
an
ex
ec
u
tio
n
tim
e
o
f
0
.
0
0
6
s
ec
o
n
d
s
an
d
an
ac
c
u
r
ac
y
o
f
9
9
.
3
3
%.
A
th
r
ee
-
p
h
ase
attac
k
d
etec
tio
n
tech
n
i
q
u
e
u
s
in
g
web
t
r
af
f
ic,
we
b
co
n
ten
t
,
an
d
UR
L
f
ea
tu
r
es
as
in
p
u
t
w
as
p
r
o
p
o
s
ed
b
y
Nath
ez
h
t
h
a
et
a
l
.
[
3
4
]
.
Acc
o
r
d
in
g
to
ex
p
er
im
e
n
tal
d
ata,
th
e
s
u
g
g
ested
m
eth
o
d
d
etec
ted
b
o
th
p
h
is
h
in
g
a
n
d
ze
r
o
-
d
a
y
p
h
is
h
in
g
attem
p
ts
with
a
n
ac
cu
r
ac
y
o
f
9
8
.
9
%.
C
lass
if
y
in
g
XM
L
-
b
ased
UR
L
s
ac
co
r
d
i
n
g
to
th
eir
s
em
an
tic
s
tr
u
ctu
r
al
o
r
ien
tatio
n
was
th
e
s
u
b
ject
o
f
a
s
ep
ar
ate
s
tu
d
y
b
y
Mu
r
th
y
et
a
l
.
[
3
5
]
.
An
ac
c
u
r
ac
y
o
f
9
7
.
3
6
% wa
s
attain
ed
b
y
t
h
eir
m
eth
o
d
.
4
.
2
.
H
euristic
-
ba
s
ed
a
pp
ro
a
ch
Stu
d
y
[
2
9
]
s
u
g
g
ested
a
web
p
h
is
h
in
g
d
etec
tio
n
m
eth
o
d
th
at
u
tili
ze
d
in
teg
r
ated
f
ea
tu
r
es
f
r
o
m
a
web
s
ite
'
s
tex
t,
g
r
ap
h
ics,
a
n
d
f
r
am
es.
T
h
e
y
u
tili
ze
d
ANFI
S
,
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
e
(
S
VM
)
,
an
d
k
-
n
ea
r
est
Id
e
n
ti
f
ica
ti
o
n
•
Rec
o
rd
s
id
en
tif
ied
(N=1
0
2
)
S
c
re
e
n
in
g
•
Re
c
o
rd
s
a
f
ter
e
li
m
in
a
ted
n
o
n
f
o
rm
in
g
p
a
p
e
rs
(N=
6
0
)
S
e
lec
ti
o
n
•
F
u
ll
tex
t
p
ap
eres
assessed
(N=3
4
)
St
u
d
y
in
c
lu
d
e
d
3
4
p
a
p
e
rs
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
E
xp
lo
r
in
g
fea
tu
r
e
en
g
in
ee
r
in
g
a
n
d
ex
p
la
in
a
b
le
A
I
fo
r
p
h
is
h
in
g
w
eb
s
ite
d
etec
tio
n
…
(
N
o
r
a
h
A
ls
u
q
a
yh
)
5869
n
eig
h
b
o
r
(K
-
NN)
class
if
ier
s
,
with
an
ac
cu
r
ac
y
o
f
9
8
.
3
%.
Featu
r
e
s
elec
tio
n
was
co
n
d
u
c
ted
v
ia
I
n
f
o
r
m
atio
n
Gain
an
d
C
h
i
-
Sq
u
ar
e
m
eth
o
d
o
lo
g
ies
to
im
p
r
o
v
e
m
o
d
el
ef
f
icac
y
.
Ad
d
itio
n
ally
,
R
ao
et
a
l.
in
[
3
6
]
i
n
tr
o
d
u
ce
d
a
h
eu
r
is
tic
-
b
ased
p
h
is
h
in
g
d
ete
ctio
n
m
eth
o
d
th
at
an
aly
ze
s
th
e
lo
g
in
p
a
g
e
an
d
t
h
e
h
o
m
ep
a
g
e
o
f
a
web
s
ite
b
y
u
tili
zin
g
h
y
p
e
r
lin
k
an
d
UR
L
-
b
ased
ch
ar
ac
ter
is
tics
.
T
h
e
m
et
h
o
d
em
p
lo
y
ed
a
T
win
SVM
class
if
ier
to
id
en
tify
in
ten
tio
n
ally
r
eg
is
ter
ed
p
h
is
h
i
n
g
web
s
ites
.
E
x
p
er
im
en
tal
r
esu
lts
in
d
icate
d
th
at
th
e
T
win
S
VM
s
u
r
p
ass
ed
o
th
er
v
ar
iatio
n
s
,
attain
in
g
a
r
ec
all
o
f
9
8
.
3
3
% a
n
d
a
n
ac
cu
r
ac
y
o
f
9
8
.
0
5
%.
Fu
r
th
er
m
o
r
e
,
th
e
s
tu
d
y
in
[
3
7
]
s
o
u
g
h
t to
ass
es
s
1
2
s
tatic
ele
m
en
ts
,
in
clu
d
in
g
k
ey
w
o
r
d
s
an
d
s
tr
u
ctu
r
al
p
atter
n
s
,
in
s
elec
ted
p
h
is
h
in
g
UR
L
s
an
d
m
o
n
ito
r
th
eir
p
r
ev
ale
n
ce
th
r
o
u
g
h
o
u
t
co
n
tem
p
o
r
ar
y
p
h
is
h
in
g
web
s
ites
.
Alo
n
g
s
id
e
th
is
in
v
esti
g
atio
n
,
th
e
r
esear
ch
er
s
p
er
f
o
r
m
e
d
b
o
th
q
u
an
titativ
e
an
d
q
u
alitativ
e
ev
alu
atio
n
s
o
f
b
eh
av
io
r
al
p
att
er
n
s
.
T
h
is
e
n
ab
led
th
e
id
en
tif
icatio
n
o
f
cr
itical
c
o
m
p
o
n
en
ts
,
in
clu
d
i
n
g
f
ea
tu
r
e
s
ig
n
if
ican
ce
,
in
ter
-
f
ea
tu
r
e
co
r
r
elatio
n
s
,
an
d
s
im
ilar
ities
,
wh
ich
ca
n
f
ac
ilit
ate
th
e
cr
ea
tio
n
o
f
n
o
v
el
h
eu
r
is
tic
m
eth
o
d
s
o
r
im
p
r
o
v
e
ex
is
tin
g
o
n
es.
I
n
o
r
d
e
r
to
c
h
o
o
s
e
th
e
b
est
class
if
ier
,
R
am
an
a
et
a
l.
[
3
8
]
p
r
esen
ted
a
n
en
s
em
b
le
-
b
ased
p
h
is
h
in
g
d
et
ec
tio
n
m
o
d
el
th
at
co
m
b
in
es
m
an
y
m
ac
h
in
e
lear
n
i
n
g
m
et
h
o
d
s
,
s
u
ch
as
R
F,
d
ec
is
io
n
tr
ee
(
DT
)
,
a
n
d
XGBo
o
s
t
.
T
o
im
p
r
o
v
e
class
if
icatio
n
p
er
f
o
r
m
an
ce
,
th
e
s
tu
d
y
also
u
s
ed
a
n
u
m
b
er
o
f
f
ea
tu
r
e
s
elec
tio
n
s
tr
ateg
ies,
in
clu
d
in
g
ANOV
A,
I
n
f
o
r
m
atio
n
Gain
,
Fis
h
er
Sco
r
e,
R
elief
-
F,
an
d
r
ec
u
r
s
iv
e
f
ea
tu
r
e
elim
in
atio
n
.
W
h
en
test
ed
o
n
th
e
Me
n
d
eley
p
h
is
h
in
g
d
ataset,
th
e
m
o
d
el'
s
ac
cu
r
ac
y
was
9
8
.
4
5
%,
b
u
t
it
was
9
7
.
5
1
%
o
n
th
e
UC
I
p
h
i
s
h
in
g
d
ataset.
L
astl
y
,
Do
o
r
e
m
aa
l
et
a
l.
[
3
9
]
p
r
esen
ted
a
n
o
v
el
m
eth
o
d
f
o
r
d
etec
tin
g
p
h
is
h
in
g
attac
k
s
b
y
co
m
b
in
in
g
tex
tu
al
d
ata
f
r
o
m
t
h
e
d
o
cu
m
e
n
t
o
b
ject
m
o
d
el
(
D
OM
)
s
tr
u
ctu
r
e
with
v
is
u
al
f
ea
tu
r
es
tak
en
f
r
o
m
s
cr
ee
n
s
h
o
ts
o
f
web
p
ag
es.
W
ith
an
o
v
er
all
d
etec
tio
n
ac
cu
r
ac
y
o
f
9
9
.
6
6
%,
th
is
h
y
b
r
id
ap
p
r
o
ac
h
d
r
am
atica
lly
d
ec
r
ea
s
ed
th
e
p
h
is
h
in
g
m
is
cla
s
s
if
icatio
n
r
ate
b
y
6
7
%,
f
r
o
m
1
.
0
2
% to
0
.
3
4
%.
4
.
3
.
M
L
a
pp
ro
a
ch
Stu
d
y
[
3
0
]
r
elate
d
a
m
u
ltis
tag
e
p
h
is
h
in
g
d
etec
tio
n
m
o
d
el
an
d
p
r
esen
ted
an
ex
te
n
s
iv
e
C
AS
E
f
ea
tu
r
e
ar
ch
itectu
r
e,
class
if
y
in
g
f
ea
t
u
r
es
in
to
f
o
u
r
p
r
in
cip
al
ca
teg
o
r
ies:
C
o
u
n
ter
f
eitin
g
,
Af
f
iliatio
n
,
Stealin
g
,
an
d
E
v
alu
atio
n
.
T
h
e
s
u
g
g
ested
m
e
th
o
d
e
x
h
ib
ited
r
o
b
u
s
t
ef
f
icac
y
in
p
r
ac
tical
p
h
is
h
in
g
d
etec
tio
n
co
n
tex
ts
,
y
ield
in
g
ef
f
icien
t
o
u
tco
m
es
with
m
in
im
ized
ex
ec
u
tio
n
d
u
r
atio
n
s
.
A
p
h
is
h
in
g
d
etec
tio
n
alg
o
r
ith
m
with
h
y
b
r
id
cu
m
u
lativ
e
f
ea
tu
r
e
s
elec
tio
n
was
p
r
o
p
o
s
ed
i
n
[
3
1
]
.
T
h
e
m
eth
o
d
o
lo
g
y
u
tili
ze
s
v
ar
io
u
s
f
ea
tu
r
e
s
elec
tio
n
ap
p
r
o
ac
h
es,
s
u
ch
as
C
h
i
-
Sq
u
ar
e,
g
ain
r
atio
,
i
n
f
o
r
m
atio
n
g
ain
,
Pear
s
o
n
co
r
r
elatio
n
co
ef
f
icien
t
,
an
d
PC
A,
to
d
iv
id
e
th
e
d
ataset
in
to
n
s
u
b
s
ets
ac
co
r
d
in
g
to
th
e
ch
o
s
en
f
ea
tu
r
es.
A
v
ar
iety
o
f
class
if
ier
s
is
em
p
lo
y
ed
f
o
r
ea
ch
p
ar
titi
o
n
,
in
clu
d
in
g
SVM,
NB
,
C
4
.
5
,
R
F,
J
R
ip
,
PAR
T
,
an
d
KNN.
T
h
e
R
F
class
if
i
er
attain
ed
t
h
e
b
est
ac
cu
r
ac
y
,
with
9
8
.
2
4
%.
Stu
d
y
[
4
0
]
p
r
o
v
id
ed
a
p
h
is
h
in
g
d
ete
ctio
n
f
r
am
ewo
r
k
u
tili
zin
g
a
cl
ass
if
ier
to
f
ac
ilit
at
e
th
e
co
m
p
ar
ativ
e
ass
ess
m
en
t
o
f
d
etec
tio
n
s
y
s
tem
s
b
ased
o
n
8
7
d
is
tin
ct
f
ea
tu
r
es.
T
o
m
iti
g
ate
th
e
ep
h
em
er
al
n
atu
r
e
o
f
p
h
is
h
in
g
web
s
ites
,
th
e
au
th
o
r
s
cr
ea
ted
a
d
y
n
a
m
ic
d
ataset
th
at
m
ay
ad
ap
t
o
v
er
tim
e.
T
h
ei
r
in
v
esti
g
atio
n
in
d
icate
d
th
at
web
p
ag
e
co
n
te
n
t
was
th
e
least
d
is
cr
im
in
ativ
e
f
ea
tu
r
e
g
r
o
u
p
,
b
u
t
ex
ter
n
al
f
ea
tu
r
es
—
s
u
ch
as
d
o
m
ain
an
d
h
o
s
tin
g
attr
ib
u
tes
—
wer
e
th
e
m
o
s
t
in
f
o
r
m
ativ
e.
A
m
ax
im
u
m
ac
c
u
r
ac
y
o
f
9
6
.
6
1
%
was
attain
ed
b
y
th
e
u
t
ilizatio
n
o
f
h
y
b
r
id
f
ea
tu
r
es.
Fu
r
th
er
m
o
r
e,
ap
p
ly
i
n
g
f
ilter
-
b
a
s
ed
r
an
k
i
n
g
m
eth
o
d
with
p
r
o
g
r
ess
iv
e
elim
in
atio
n
o
f
less
s
ig
n
if
ican
t f
ea
tu
r
es im
p
r
o
v
ed
th
e
ac
cu
r
ac
y
b
y
9
6
.
8
3
%
I
n
co
n
tr
ast,
Gu
p
ta
et
a
l.
[
4
1
]
d
ev
is
ed
a
s
tr
ea
m
lin
ed
p
h
is
h
in
g
d
etec
tio
n
tech
n
iq
u
e
th
at
u
tili
ze
s
m
er
ely
n
in
e
lex
ical
p
ar
am
eter
s
,
in
cl
u
d
in
g
UR
L
len
g
th
,
f
o
r
class
if
ic
atio
n
p
u
r
p
o
s
es.
Af
ter
ass
ess
in
g
th
e
s
tr
ateg
y
with
m
an
y
m
ac
h
i
n
e
lear
n
in
g
class
if
ier
s
,
th
e
R
F
alg
o
r
ith
m
attain
ed
th
e
g
r
ea
test
ac
cu
r
ac
y
o
f
9
9
.
5
7
%.
An
u
p
a
m
a
n
d
Kar
[
4
2
]
em
p
lo
y
ed
d
iv
er
s
e
U
R
L
-
b
ased
ch
ar
ac
ter
is
tics
—
s
u
ch
as
th
e
len
g
th
o
f
th
e
I
P
ad
d
r
ess
an
d
th
e
v
alid
ity
o
f
th
e
HT
T
PS
r
eq
u
est
—
to
ca
teg
o
r
ize
web
s
ites
as
p
h
is
h
in
g
o
r
r
ea
l.
A
b
in
a
r
y
SVM
class
i
f
ier
was
u
tili
ze
d
to
d
eter
m
in
e
an
a
p
p
r
o
p
r
iate
h
y
p
er
p
lan
e
f
o
r
class
if
icatio
n
p
u
r
p
o
s
es.
Fo
u
r
o
p
tim
izatio
n
s
tr
ate
g
ies
wer
e
em
p
lo
y
ed
to
im
p
r
o
v
e
SVM
p
er
f
o
r
m
an
ce
:
th
e
b
at
alg
o
r
ith
m
,
th
e
f
ir
e
f
ly
alg
o
r
ith
m
,
t
h
e
g
r
e
y
wo
lf
o
p
tim
izer
(
GW
O)
,
an
d
th
e
wh
ale
o
p
tim
izatio
n
alg
o
r
i
th
m
.
T
h
e
GW
O
alg
o
r
ith
m
s
u
r
p
ass
ed
th
e
f
ir
ef
ly
alg
o
r
ith
m
r
eg
ar
d
in
g
d
etec
tio
n
ac
cu
r
ac
y
.
4
.
4
.
DL
a
pp
ro
a
ch
I
n
[
1
2
]
,
d
ee
p
lear
n
in
g
-
b
ased
p
h
is
h
in
g
d
etec
tio
n
m
o
d
el
was
p
r
o
p
o
s
ed
u
s
in
g
a
C
NN
ar
ch
itectu
r
e
th
at
r
elies
s
o
lely
o
n
th
e
web
s
ite'
s
UR
L
an
d
v
ar
io
u
s
f
ea
t
u
r
e
r
e
p
r
esen
tatio
n
s
.
T
h
ese
in
clu
d
e
h
a
n
d
-
cr
af
te
d
ch
ar
ac
te
r
em
b
ed
d
in
g
s
,
ch
ar
ac
ter
-
le
v
el
T
F
-
I
DF,
an
d
ch
ar
ac
ter
-
lev
el
co
u
n
t
v
ec
to
r
f
ea
tu
r
es.
No
tab
ly
,
th
e
m
o
d
el
d
o
es
n
o
t
r
eq
u
ir
e
ac
ce
s
s
to
web
p
ag
e
c
o
n
ten
t
o
r
an
y
th
ir
d
-
p
ar
ty
s
er
v
ices,
n
o
r
d
o
es
it
d
ep
e
n
d
o
n
p
r
io
r
k
n
o
wled
g
e
o
f
p
h
is
h
in
g
tech
n
iq
u
es.
I
n
s
tead
,
i
t
ca
p
tu
r
es
in
f
o
r
m
ativ
e
an
d
s
eq
u
en
tial
p
atter
n
s
with
in
UR
L
s
tr
in
g
s
f
o
r
e
f
f
ec
tiv
e
d
etec
tio
n
.
T
h
e
p
r
o
p
o
s
ed
m
o
d
e
l
ac
h
iev
ed
an
ac
c
u
r
ac
y
o
f
9
5
.
0
2
%
o
n
a
cu
s
to
m
d
ataset
an
d
r
ec
o
r
d
e
d
ac
cu
r
ac
ies
o
f
9
8
.
5
8
%,
9
5
.
4
6
%,
an
d
9
5
.
2
2
%
o
n
th
r
ee
b
en
c
h
m
ar
k
d
atasets
,
o
u
tp
er
f
o
r
m
in
g
ex
is
tin
g
p
h
is
h
in
g
UR
L
d
etec
tio
n
m
o
d
els.
I
n
co
n
tr
ast,
Vr
b
an
čič
et
a
l.
[
3
2
]
p
r
o
p
o
s
e
d
an
an
ti
-
p
h
is
h
in
g
s
y
s
tem
th
at
in
teg
r
ates
UR
L
-
b
ased
,
n
atu
r
al
lan
g
u
a
g
e
p
r
o
c
ess
in
g
(
NL
P)
-
b
ased
,
an
d
h
o
s
t
-
b
ased
f
ea
tu
r
es
t
o
tr
ain
a
r
an
g
e
o
f
ML
an
d
DL
m
o
d
els,
in
clu
d
in
g
K
-
NN,
L
R
,
SVM,
g
r
ad
ien
t
b
o
o
s
tin
g
(
GB
)
,
Ad
aBo
o
s
t,
R
F,
an
d
n
e
u
r
al
n
etwo
r
k
(
NN)
.
Am
o
n
g
th
ese,
th
e
NN
m
o
d
el
a
ch
iev
ed
th
e
h
ig
h
est ac
cu
r
ac
y
,
r
ea
ch
in
g
9
4
.
8
9
% in
p
h
is
h
in
g
UR
L
d
etec
tio
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
6
,
Decem
b
e
r
20
25
:
5
8
6
3
-
5
8
7
8
5870
I
n
[
4
3
]
,
a
ch
ar
ac
ter
-
lev
el
co
n
v
o
lu
tio
n
al
au
to
en
c
o
d
er
(
C
A
E
)
was
d
ev
elo
p
ed
with
in
an
an
o
m
aly
d
etec
tio
n
f
r
a
m
ewo
r
k
f
o
r
p
h
is
h
in
g
d
etec
tio
n
.
E
x
p
er
im
e
n
tal
ev
alu
atio
n
,
c
o
n
d
u
cted
u
s
in
g
R
OC
cu
r
v
e
an
aly
s
is
an
d
1
0
-
f
o
ld
c
r
o
s
s
-
v
alid
atio
n
,
d
em
o
n
s
tr
ated
th
at
th
e
p
r
o
p
o
s
ed
m
o
d
el
im
p
r
o
v
e
d
s
en
s
itiv
ity
b
y
3
.
9
8
%
co
m
p
ar
e
d
to
th
e
m
o
s
t
r
ec
en
t
d
ee
p
lear
n
i
n
g
m
o
d
el.
T
h
ese
r
esu
lts
co
n
f
ir
m
th
e
ef
f
ec
tiv
en
ess
o
f
th
e
C
AE
-
b
ased
ap
p
r
o
ac
h
in
id
en
tif
y
in
g
p
h
is
h
in
g
th
r
e
ats.
Xiao
et
a
l.
[
4
4
]
in
tr
o
d
u
ce
d
a
s
elf
-
atten
tio
n
-
b
ased
C
NN
m
o
d
el
t
h
at
in
co
r
p
o
r
ates
a
g
en
er
ativ
e
ad
v
e
r
s
ar
ial
n
etwo
r
k
(
GAN)
to
s
y
n
th
esize
p
h
is
h
in
g
UR
L
s
f
o
r
tr
ai
n
in
g
p
u
r
p
o
s
es.
T
h
e
p
r
o
p
o
s
ed
ar
ch
itectu
r
e
c
o
n
s
is
ts
o
f
f
o
u
r
m
ain
co
m
p
o
n
en
ts
:
th
e
in
p
u
t
b
lo
c
k
,
atten
tio
n
b
l
o
ck
,
f
ea
tu
r
e
b
lo
c
k
,
an
d
o
u
tp
u
t
b
lo
ck
.
B
y
co
m
b
in
in
g
C
NN
with
m
u
lti
-
h
ea
d
s
elf
-
att
en
tio
n
m
ec
h
a
n
is
m
s
,
th
e
m
o
d
e
l
co
n
s
tr
u
cts
a
r
o
b
u
s
t
class
if
ier
ca
p
ab
le
o
f
ac
cu
r
ately
d
etec
tin
g
p
r
ev
io
u
s
ly
u
n
s
ee
n
p
h
is
h
in
g
UR
L
s
.
T
h
e
class
if
ier
ac
h
iev
ed
an
ac
cu
r
ac
y
o
f
9
5
.
6
%,
o
u
tp
er
f
o
r
m
in
g
b
aselin
e
m
o
d
els
—
C
NN
-
L
STM
,
s
tan
d
alo
n
e
C
N
N,
an
d
s
tan
d
alo
n
e
L
STM
—
b
y
m
ar
g
i
n
s
o
f
1
.
4
%,
4
.
6
%,
a
n
d
2
.
1
%,
r
esp
ec
tiv
ely
.
AlE
r
o
u
d
et
a
l.
[
4
5
]
em
p
lo
y
e
d
GAN
to
g
e
n
er
ate
UR
L
-
b
ased
p
h
is
h
in
g
ex
am
p
l
es
ca
p
ab
le
o
f
ev
ad
in
g
d
etec
tio
n
.
T
h
e
s
y
n
th
esized
ex
am
p
l
es
wer
e
s
h
o
wn
to
ef
f
ec
tiv
ely
d
ec
eiv
e
b
o
th
s
im
p
le
an
d
ad
v
an
ce
d
b
lack
-
b
o
x
m
a
ch
in
e
lear
n
in
g
-
b
ased
p
h
is
h
in
g
d
etec
tio
n
m
o
d
els.
4
.
5
.
H
y
brid a
pp
ro
a
ch
R
ao
an
d
Pais
in
[
4
6
]
s
u
g
g
ested
an
e
n
s
em
b
le
p
h
is
h
in
g
d
etec
tio
n
m
o
d
el
th
at
in
co
r
p
o
r
ates
ex
tr
a
tr
ee
s
,
R
F,
an
d
XGBo
o
s
t
c
lass
if
ier
s
.
T
h
e
m
o
d
el
ass
ess
es
th
e
s
y
n
er
g
is
tic
ef
f
icac
y
o
f
h
eu
r
is
tic
an
d
b
lack
lis
t
f
ilter
in
g
s
tr
ateg
ies
as
a
co
h
esiv
e
s
tr
ateg
y
,
with
a
n
ac
cu
r
a
cy
o
f
9
8
.
7
2
%.
Fu
r
th
er
m
o
r
e
,
Ko
r
k
m
az
et
a
l.
in
[
4
7
]
cr
ea
ted
a
p
h
is
h
in
g
d
etec
tio
n
s
y
s
tem
with
a
C
NN
th
at
em
p
lo
y
s
n
-
g
r
am
ch
ar
ac
ter
is
tics
d
er
iv
ed
f
r
o
m
UR
L
s
.
E
x
p
er
im
en
tal
f
in
d
in
g
s
d
em
o
n
s
tr
ated
th
at
u
n
ig
r
am
s
p
r
o
d
u
ce
d
t
h
e
g
r
ea
test
ca
teg
o
r
izatio
n
ac
cu
r
ac
y
.
T
h
e
m
o
d
el
attain
e
d
a
n
ac
cu
r
ac
y
o
f
8
8
.
9
0
%
o
n
th
e
U
R
L
d
ataset
b
y
u
tili
zin
g
a
s
p
ec
if
ic
s
et
o
f
7
0
ch
ar
a
cter
s
.
Ad
d
i
tio
n
ally
,
Or
u
n
s
o
lu
et
a
l.
in
[
4
8
]
s
u
g
g
ested
a
p
h
is
h
in
g
d
etec
tio
n
m
eth
o
d
th
at
in
clu
d
es
a
f
ea
tu
r
e
s
elec
tio
n
m
o
d
u
le
to
ex
tr
ac
t
p
er
tin
en
t
in
f
o
r
m
atio
n
f
r
o
m
UR
L
s
tr
u
ctu
r
e,
web
p
ag
e
att
r
ib
u
tes,
an
d
web
p
a
g
e
ac
tiv
it
y
u
s
in
g
f
r
eq
u
en
cy
ass
es
s
m
en
t
an
aly
s
is
.
T
h
e
m
et
h
o
d
o
lo
g
y
was
ass
es
s
ed
with
NB
an
d
SVM
clas
s
if
ier
s
.
E
x
p
er
im
en
tal
r
esu
lts
in
d
icate
d
an
ef
f
icie
n
t
r
u
n
tim
e
o
f
u
n
d
er
2
,
0
0
0
m
illi
s
ec
o
n
d
s
,
ac
co
m
p
an
ied
b
y
r
o
b
u
s
t
p
e
r
f
o
r
m
a
n
ce
m
etr
ics:
9
9
.
9
6
%
tr
u
e
p
o
s
itiv
es,
9
9
.
9
6
%
tr
u
e
n
eg
ativ
es,
0
.
0
4
%
f
alse
p
o
s
itiv
es,
an
d
0
.
0
4
%
f
alse
n
eg
ativ
es.
Als
o
,
Yu
et
a
l.
[
4
9
]
cr
ea
ted
a
h
y
b
r
id
p
h
is
h
in
g
d
etec
tio
n
m
o
d
el
th
at
co
m
b
in
es
v
ar
io
u
s
d
ee
p
lear
n
i
n
g
ar
ch
itectu
r
es
f
o
r
f
ea
tu
r
e
e
x
tr
ac
tio
n
a
n
d
class
i
f
icatio
n
.
A
m
u
ltil
ay
er
p
er
ce
p
tr
o
n
(
ML
P)
p
r
o
ce
s
s
ed
cu
s
t
o
m
f
ea
tu
r
es,
C
NN
h
an
d
led
im
ag
e
-
b
ased
f
ea
tu
r
e
s
,
an
d
a
r
ec
u
r
r
e
n
t
n
e
u
r
al
n
e
two
r
k
(
R
NN)
m
an
ag
e
d
tex
t
-
b
ased
f
ea
tu
r
e
.
T
h
e
r
etr
iev
ed
f
ea
tu
r
e
v
ec
to
r
s
wer
e
s
u
b
s
eq
u
en
tly
in
te
g
r
ated
u
s
in
g
a
class
if
icatio
n
n
etwo
r
k
to
g
e
t
f
in
al
p
r
ed
ictio
n
s
.
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
attain
ed
an
o
v
er
all
ac
c
u
r
ac
y
o
f
9
7
%.
Fu
r
th
er
m
o
r
e
,
Ar
iy
a
d
asa
et
a
l.
[
5
0
]
s
u
g
g
ested
a
p
h
is
h
in
g
d
e
tectio
n
m
eth
o
d
th
at
in
te
g
r
ates
lo
n
g
-
ter
m
r
ec
u
r
r
en
t
co
n
v
o
lu
tio
n
al
n
etw
o
r
k
s
with
g
r
ap
h
co
n
v
o
lu
tio
n
al
n
etwo
r
k
s
,
em
p
lo
y
in
g
b
o
th
UR
L
an
d
HT
ML
ch
ar
ac
ter
is
tics
.
T
h
e
ap
p
r
o
ac
h
lev
er
ag
es
th
e
s
o
p
h
is
ticated
an
aly
tical
p
o
wer
s
o
f
g
r
ap
h
n
e
u
r
al
n
etwo
r
k
s
in
th
e
an
ti
-
p
h
is
h
in
g
s
ec
to
r
.
E
x
p
er
im
en
tal
r
esu
lts
in
d
icate
d
a
d
etec
t
io
n
ac
cu
r
ac
y
o
f
9
6
.
4
2
%
an
d
a
f
alse
-
n
eg
ativ
e
r
ate
o
f
0
.
0
3
6
.
Als
o
,
s
tu
d
y
[
2
4
]
s
u
g
g
ested
a
p
h
is
h
in
g
web
s
ite
d
et
ec
tio
n
m
eth
o
d
th
at
ex
clu
s
iv
el
y
u
tili
ze
s
th
e
UR
L
,
en
ca
p
s
u
latin
g
its
in
f
o
r
m
atio
n
in
to
a
two
-
d
im
en
s
io
n
al
t
en
s
o
r
.
T
h
is
ten
s
o
r
is
in
itial
ly
p
r
o
ce
s
s
ed
b
y
a
b
id
ir
ec
tio
n
al
lo
n
g
s
h
o
r
t
-
te
r
m
m
em
o
r
y
(
B
i
-
L
STM
)
n
et
wo
r
k
to
ex
t
r
ac
t
g
lo
b
al
c
o
n
tex
tu
al
in
f
o
r
m
ati
o
n
,
s
u
b
s
eq
u
en
tly
f
o
llo
wed
b
y
C
NN
to
au
to
m
atica
lly
i
d
en
tify
th
e
m
o
s
t
p
er
tin
en
t
c
o
m
p
o
n
e
n
ts
o
f
th
e
UR
L
.
T
h
e
s
u
g
g
ested
m
o
d
el,
PDR
C
NN,
attain
ed
a
d
etec
tio
n
ac
c
u
r
ac
y
o
f
9
7
%
a
n
d
an
AUC
v
al
u
e
o
f
9
9
%
in
ex
p
er
im
en
tal
ass
es
s
m
en
ts
.
Stu
d
y
[
1
0
]
co
m
b
in
ed
C
NN
an
d
R
F
b
y
em
p
l
o
y
i
n
g
c
h
ar
ac
ter
em
b
ed
d
in
g
tech
n
iq
u
es
to
t
r
an
s
f
o
r
m
UR
L
s
in
to
f
ix
ed
-
s
ize
m
atr
ic
es,
ex
tr
ac
tin
g
f
e
atu
r
es
at
v
a
r
io
u
s
lev
els
with
C
NN
m
o
d
els,
s
u
b
s
eq
u
en
tly
class
if
y
in
g
th
ese
f
ea
tu
r
es
u
s
in
g
m
u
ltip
le
R
F
class
if
ier
s
,
an
d
u
ltima
tely
p
r
o
d
u
cin
g
p
r
ed
ictio
n
r
esu
lts
th
r
o
u
g
h
a
win
n
er
-
tak
e
-
all
m
eth
o
d
.
A
p
r
ec
is
e
r
ate
o
f
9
9
.
2
6
%
was
attain
ed
o
n
th
e
b
en
ch
m
a
r
k
d
ata.
Fin
ally
,
Stu
d
y
[
5
1
]
p
r
esen
ted
HT
ML
Ph
is
h
,
a
p
h
is
h
in
g
d
etec
tio
n
m
o
d
el
th
at
an
aly
ze
s
th
e
HT
ML
co
n
ten
t
o
f
web
p
ag
es
th
r
o
u
g
h
C
NN
to
d
is
ce
r
n
s
em
an
tic
r
ela
tio
n
s
h
ip
s
with
in
th
e
tex
tu
al
s
tr
u
ctu
r
e,
elim
in
atin
g
th
e
n
ee
d
f
o
r
m
an
u
al
f
ea
tu
r
e
en
g
in
ee
r
in
g
.
T
h
is
m
eth
o
d
o
l
o
g
y
allo
ws
th
e
m
o
d
el
to
a
d
a
p
tiv
ely
m
an
a
g
e
n
o
v
el
f
ea
tu
r
es
an
d
g
en
e
r
alize
p
r
o
f
icien
tly
to
p
r
ev
io
u
s
ly
u
n
o
b
s
er
v
ed
test
d
ata.
HT
ML
Ph
is
h
attain
ed
a
d
etec
tio
n
ac
cu
r
ac
y
an
d
tr
u
e
p
o
s
itiv
e
r
ate
o
f
9
3
%,
illu
s
tr
atin
g
its
ef
f
icac
y
in
r
ec
o
g
n
izin
g
p
h
is
h
in
g
web
s
ites
ju
s
t
th
r
o
u
g
h
HT
ML
co
n
ten
t.
T
a
b
le
1
(
in
ap
p
en
d
i
x
)
s
h
o
ws a
s
u
m
m
atio
n
o
f
r
ec
e
n
t r
esear
ch
o
n
p
h
is
h
in
g
d
etec
tio
n
m
o
d
els.
4
.
6
.
XAI in phis
hi
ng
websi
t
e
det
ec
t
io
n
T
o
th
e
b
est
o
f
o
u
r
k
n
o
w
le
d
g
e
,
t
h
e
ap
p
l
ic
ati
o
n
o
f
XA
I
i
n
p
h
is
h
in
g
d
et
ec
ti
o
n
r
e
m
a
in
s
r
el
ati
v
e
l
y
u
n
d
er
e
x
p
lo
r
e
d
.
T
h
e
w
o
r
k
i
n
[
5
2
]
ex
p
l
o
r
e
d
t
h
e
i
n
te
r
p
r
et
ab
ili
ty
o
f
p
h
is
h
i
n
g
d
et
ec
ti
o
n
m
o
d
e
ls
b
y
a
p
p
l
y
i
n
g
R
F,
an
d
SV
M
i
n
co
m
b
i
n
at
io
n
wit
h
XAI
m
et
h
o
d
s
,
i
n
c
lu
d
i
n
g
L
I
M
E
a
n
d
e
x
p
l
ai
n
a
b
le
b
o
o
s
ti
n
g
m
a
ch
i
n
es
(
E
B
M
)
.
T
h
e
an
al
y
s
is
s
h
o
we
d
t
h
at
t
h
e
m
o
s
t
i
n
f
lu
e
n
ti
al
UR
L
f
e
at
u
r
es,
as
id
e
n
ti
f
i
ed
b
y
t
h
es
e
t
ec
h
n
i
q
u
e
s
,
c
lo
s
el
y
m
a
tc
h
e
d
ty
p
ic
al
p
h
is
h
i
n
g
-
r
elat
e
d
a
tt
r
i
b
u
tes.
W
h
i
le
s
tu
d
y
[
1
3
]
e
x
p
lo
r
ed
th
e
ap
p
l
ica
ti
o
n
o
f
XA
I
t
ec
h
n
i
q
u
es to
e
n
h
a
n
c
e
t
h
e
d
et
ec
ti
o
n
o
f
p
h
is
h
i
n
g
a
tte
m
p
ts
in
e
m
ai
ls
.
T
h
eir
s
t
u
d
y
em
p
h
as
i
ze
d
th
e
im
p
o
r
ta
n
c
e
o
f
s
p
ec
i
f
i
c
wo
r
d
s
a
n
d
p
h
r
as
es
th
a
t s
ig
n
i
f
i
ca
n
tl
y
i
n
f
lu
e
n
ce
t
h
e
class
i
f
i
ca
t
io
n
d
ec
is
i
o
n
s
m
a
d
e
b
y
p
h
is
h
i
n
g
d
e
tec
ti
o
n
m
o
d
els.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
E
xp
lo
r
in
g
fea
tu
r
e
en
g
in
ee
r
in
g
a
n
d
ex
p
la
in
a
b
le
A
I
fo
r
p
h
is
h
in
g
w
eb
s
ite
d
etec
tio
n
…
(
N
o
r
a
h
A
ls
u
q
a
yh
)
5871
Ad
d
itio
n
ally
,
s
tu
d
y
[
1
4
]
p
r
o
p
o
s
ed
a
m
u
lti
-
m
o
d
al
h
ier
ar
ch
ical
atten
tio
n
m
o
d
el
d
esig
n
ed
to
lear
n
d
ee
p
p
h
is
h
in
g
in
d
icato
r
s
f
r
o
m
UR
L
,
tex
tu
al,
a
n
d
v
is
u
al
m
o
d
alitie
s
.
T
h
e
m
o
d
el
in
co
r
p
o
r
ates
two
lev
els
o
f
atten
tio
n
m
ec
h
an
is
m
s
to
f
ac
ilit
ate
th
e
ex
tr
ac
tio
n
o
f
r
elev
an
t
f
ea
tu
r
es
an
d
to
p
r
o
v
id
e
in
f
o
r
m
ativ
e
i
n
ter
p
r
etab
ilit
y
ac
r
o
s
s
d
if
f
er
en
t
m
o
d
alities
.
E
x
p
er
im
en
tal
r
esu
lts
d
em
o
n
s
tr
ated
th
at
th
e
m
o
d
el
n
o
t
o
n
ly
en
h
an
ce
s
p
h
is
h
in
g
d
etec
tio
n
p
er
f
o
r
m
an
ce
b
u
t
also
o
f
f
er
s
h
ier
ar
ch
ical
in
ter
p
r
eta
b
ilit
y
,
im
p
r
o
v
i
n
g
tr
an
s
p
ar
e
n
cy
in
th
e
d
ec
is
io
n
-
m
ak
in
g
p
r
o
ce
s
s
.
T
o
im
p
r
o
v
e
in
te
r
p
r
e
tab
ilit
y
,
s
tu
d
y
[
5
3
]
u
s
ed
a
h
y
b
r
id
d
ee
p
lear
n
in
g
-
b
ased
m
o
d
el
th
at
in
clu
d
ed
ex
p
lain
ab
le
v
is
u
al
an
n
o
tatio
n
s
s
u
p
er
im
p
o
s
ed
o
n
s
cr
ee
n
s
h
o
ts
o
f
p
h
is
h
in
g
web
s
ites
.
A
two
-
s
tag
e
s
tack
ed
en
s
em
b
le
lear
n
in
g
tech
n
iq
u
e
was
u
s
ed
b
y
s
tu
d
y
[
5
4
]
,
wh
o
ap
p
lied
GB
an
d
R
F
class
if
ier
s
to
2
1
s
elec
ted
f
ea
tu
r
es
f
r
o
m
a
d
ataset
o
f
6
5
1
,
1
9
1
UR
L
s
.
T
h
e
ac
cu
r
ac
y
o
f
th
e
s
u
g
g
ested
m
o
d
el
was
9
7
%.
T
h
e
m
o
d
el'
s
d
ec
is
io
n
-
m
ak
in
g
p
r
o
ce
s
s
was
th
en
in
ter
p
r
eted
u
s
in
g
XAI
ap
p
r
o
ac
h
es,
wh
ic
h
wer
e
also
u
s
ed
to
ex
a
m
in
e
ea
ch
f
ea
tu
r
e'
s
co
n
tr
ib
u
tio
n
to
th
e
f
o
u
r
-
class
p
r
e
d
ictio
n
c
h
allen
g
e
,
wh
ich
in
clu
d
e
d
m
alwa
r
e,
p
h
is
h
in
g
,
d
e
f
ac
em
en
t
,
an
d
b
e
n
ig
n
class
if
icatio
n
s
.
Stu
d
y
[
5
5
]
SHAP
v
alu
es
wer
e
em
p
lo
y
ed
to
in
ter
p
r
et
b
o
t
h
i
n
d
iv
id
u
al
m
ac
h
in
e
lear
n
in
g
m
o
d
els
an
d
en
s
em
b
le
m
o
d
els
—
in
clu
d
in
g
K
-
Me
an
s
,
R
F,
DT
,
C
atB
o
o
s
t,
L
ig
h
tGB
M,
Ad
aBo
o
s
t,
an
d
a
v
o
tin
g
class
if
ier
—
f
o
r
p
h
is
h
in
g
UR
L
d
etec
tio
n
class
if
icatio
n
.
Am
o
n
g
th
ese
,
th
e
C
atB
o
o
s
t
class
if
ier
d
em
o
n
s
tr
ated
s
u
p
e
r
io
r
p
er
f
o
r
m
an
ce
ac
r
o
s
s
ev
alu
atio
n
m
etr
ics.
T
h
e
u
s
e
o
f
SHAP
v
alu
es
p
lay
ed
a
p
iv
o
tal
r
o
le
in
id
en
tify
in
g
th
e
m
o
s
t
in
f
lu
en
tial
f
ea
tu
r
es
an
d
u
n
d
er
s
tan
d
in
g
th
eir
e
f
f
ec
ts
o
n
th
e
m
o
d
el'
s
o
u
tp
u
ts
,
th
er
e
b
y
en
h
a
n
c
in
g
in
ter
p
r
etab
ilit
y
an
d
tr
u
s
t
in
th
e
class
if
icatio
n
p
r
o
ce
s
s
.
T
ab
le
2
s
h
o
ws
a
s
u
m
m
ar
y
o
f
XAI
an
d
f
ea
tu
r
e
en
g
i
n
ee
r
in
g
ap
p
r
o
ac
h
es
f
o
r
p
h
is
h
in
g
web
s
ites
d
etec
tio
n
.
T
ab
le
2
.
Su
m
m
a
r
y
o
f
XAI
an
d
f
ea
tu
r
e
en
g
in
ee
r
in
g
ap
p
r
o
ac
h
es f
o
r
p
h
is
h
in
g
web
s
ites
d
etec
tio
n
Li
t
e
r
a
t
u
r
e
Ty
p
e
o
f
f
e
a
t
u
r
e
s
F
e
a
t
u
r
e
e
n
g
i
n
e
e
r
i
n
g
m
e
t
h
o
d
X
A
I
t
e
c
h
n
i
q
u
e
P
e
r
f
o
r
ma
n
c
e
me
t
r
i
c
s
[
5
3
]
U
R
L
N
LP t
e
c
h
n
i
q
u
e
s
LI
M
E
a
n
d
EB
M
.
P
r
e
c
i
s
i
o
n
,
r
e
c
a
l
l
,
F
1
s
c
o
r
e
a
n
d
a
c
c
u
r
a
c
y
[
8
]
Emai
l
Lo
c
a
l
f
e
a
t
u
r
e
i
m
p
o
r
t
a
n
c
e
,
t
e
x
t
h
i
g
h
l
i
g
h
t
s a
s
e
x
p
l
a
n
a
t
i
o
n
s
mo
d
e
l
-
a
g
n
o
st
i
c
p
r
i
n
c
i
p
l
e
s
,
l
o
c
a
l
f
e
a
t
u
r
e
i
mp
o
r
t
a
n
c
e
,
a
n
d
s
e
a
r
c
h
-
b
a
se
d
e
x
p
l
a
n
a
t
i
o
n
g
e
n
e
r
a
t
i
o
n
F
a
l
se
p
o
si
t
i
v
e
r
a
t
e
a
n
d
c
l
a
ss
i
f
i
c
a
t
i
o
n
t
h
r
e
s
h
o
l
d
s
[
9
]
U
R
L,
w
e
b
p
a
g
e
t
e
x
t
a
n
d
w
e
b
p
a
g
e
i
ma
g
e
S
h
a
r
e
d
d
i
c
t
i
o
n
a
r
y
l
e
a
r
n
i
n
g
a
p
p
r
o
a
c
h
H
i
e
r
a
r
c
h
i
c
a
l
A
t
t
e
n
t
i
o
n
M
e
c
h
a
n
i
sm,
A
t
t
e
n
t
i
o
n
S
c
o
r
e
V
i
s
u
a
l
i
z
a
t
i
o
n
P
r
e
c
i
s
i
o
n
,
r
e
c
a
l
l
,
F
1
s
c
o
r
e
a
n
d
a
c
c
u
r
a
c
y
[
5
4
]
U
R
L,
c
o
n
t
e
n
t
a
n
d
v
i
s
u
a
l
f
e
a
t
u
r
e
s
-
v
i
s
u
a
l
c
o
m
p
a
r
i
so
n
s
a
n
d
l
o
g
o
r
e
c
o
g
n
i
t
i
o
n
I
d
e
n
t
i
f
i
c
a
t
i
o
n
r
a
t
e
,
d
e
t
e
c
t
i
o
n
r
a
t
e
,
p
r
e
c
i
si
o
n
a
n
d
r
e
c
a
l
l
[
5
5
]
U
R
L
-
A
LE
(
A
c
c
u
m
u
l
a
t
e
d
Lo
c
a
l
Ef
f
e
c
t
s)
P
r
e
c
i
s
i
o
n
,
r
e
c
a
l
l
,
F
1
s
c
o
r
e
a
n
d
a
c
c
u
r
a
c
y
[
5
6
]
U
R
L,
c
o
n
t
e
n
t
a
n
d
b
e
h
a
v
i
o
r
a
l
f
e
a
t
u
r
e
s
-
S
H
A
P
P
r
e
c
i
s
i
o
n
,
r
e
c
a
l
l
,
F
1
s
c
o
r
e
a
n
d
a
c
c
u
r
a
c
y
5.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
R
ec
en
tly
,
p
h
is
h
in
g
b
ec
am
e
a
th
r
ea
t
in
th
e
cy
b
er
s
ec
u
r
ity
lan
d
s
ca
p
e,
tar
g
etin
g
u
s
er
s
b
y
m
im
ick
in
g
leg
itima
te
web
s
ites
to
s
tea
l
s
e
n
s
itiv
e
in
f
o
r
m
atio
n
.
T
h
is
r
esear
ch
r
ec
o
g
n
izes
th
e
ef
f
o
r
t
o
n
f
ea
tu
r
e
en
g
in
ee
r
in
g
an
d
XAI
in
to
p
h
is
h
in
g
web
s
i
te
d
etec
tio
n
,
with
a
n
o
tab
le
in
cr
ea
s
e
in
s
tu
d
ies
s
in
ce
2
0
1
9
.
T
h
ese
m
o
d
els
n
o
t
o
n
ly
e
n
h
an
ce
th
e
ac
cu
r
ac
y
o
f
d
etec
tio
n
b
u
t
also
im
p
r
o
v
e
th
e
in
ter
p
r
eta
b
ilit
y
,
wh
ic
h
ar
e
cr
i
tical
in
h
ig
h
-
s
tak
es
cy
b
er
s
ec
u
r
ity
ap
p
licatio
n
s
.
T
h
is
r
ev
iew
en
h
an
ce
s
ex
is
tin
g
k
n
o
wled
g
e
th
r
o
u
g
h
in
teg
r
ati
n
g
f
ea
tu
r
e
s
elec
tio
n
o
p
tim
izatio
n
an
d
m
o
d
el
in
te
r
p
r
etab
ilit
y
—
two
elem
en
ts
f
r
eq
u
en
tly
e
x
am
in
ed
in
d
ep
en
d
en
tly
.
I
t
o
f
f
er
s
a
co
m
p
r
eh
e
n
s
iv
e
v
iewp
o
in
t
cr
u
cial
f
o
r
d
ev
elo
p
in
g
r
esil
ien
t
an
d
tr
an
s
p
ar
en
t
p
h
is
h
in
g
d
ete
ctio
n
s
y
s
tem
s
.
T
h
e
s
tu
d
y
p
r
io
r
itizes th
e
ex
p
lain
ab
ilit
y
o
f
d
ec
is
io
n
s
an
d
t
h
eir
r
eli
ab
ilit
y
in
ess
en
tial secu
r
ity
s
y
s
tem
s
,
in
co
n
tr
ast to
p
r
ev
io
u
s
s
tu
d
ies th
at
co
n
ce
n
tr
ated
ex
clu
s
iv
ely
o
n
m
o
d
el
ac
c
u
r
ac
y
.
A
p
r
ev
io
u
s
r
ev
iew
b
y
Saf
i
an
d
Sin
g
h
[
4
0
]
d
iv
id
e
d
p
h
is
h
in
g
d
etec
tio
n
tech
n
i
q
u
es
in
to
f
iv
e
ap
p
r
o
ac
h
es;
lis
t
s
b
ased
,
v
is
u
a
l
s
im
ilar
ity
,
Heu
r
is
tic,
ML
,
a
n
d
DL
b
ased
tech
n
iq
u
es
an
d
am
o
n
g
th
ese,
ML
tech
n
iq
u
es
h
av
e
b
ee
n
ap
p
lied
th
e
m
o
s
t.
I
n
ad
d
itio
n
,
m
o
s
t
s
tu
d
ies
b
ased
o
n
s
tu
d
y
u
s
ed
ML
tech
n
iq
u
es
s
u
ch
as
R
F
wh
ile
C
N
N
ac
h
iev
ed
th
e
h
ig
h
est
ac
cu
r
ac
y
f
o
r
d
etec
tin
g
p
h
is
h
in
g
web
s
ites
.
Similar
l
y
,
C
atal
et
a
l.
[
5
7
]
,
th
r
o
u
g
h
a
co
m
p
r
eh
en
s
iv
e
liter
atu
r
e
r
ev
iew,
r
ec
o
g
n
ized
d
ee
p
lear
n
in
g
m
ec
h
a
n
is
m
s
f
o
r
p
h
is
h
in
g
d
etec
tio
n
.
T
h
e
s
tu
d
y
d
em
o
n
s
tr
ated
th
at
all
m
o
d
els
em
p
lo
y
ed
s
u
p
e
r
v
is
ed
d
e
ep
lear
n
in
g
alg
o
r
ith
m
s
an
d
u
t
ilized
d
ata
s
o
u
r
ce
s
s
u
ch
as
UR
L
-
an
d
co
n
ten
t
-
r
elate
d
f
ea
tu
r
es,
th
ir
d
-
p
a
r
ty
m
et
ad
ata
ab
o
u
t
th
e
web
s
ite,
an
d
em
ail
d
ata.
Am
o
n
g
th
ese,
DNNs a
n
d
C
NNs e
m
er
g
ed
as th
e
m
o
s
t w
id
ely
a
d
o
p
te
d
ar
ch
itectu
r
e.
Desp
ite
th
e
g
r
o
win
g
r
elian
ce
o
n
a
d
v
an
ce
d
lear
n
i
n
g
al
g
o
r
i
th
m
s
,
it
is
n
o
tewo
r
th
y
th
at
7
2
%
o
f
th
e
an
aly
ze
d
s
tu
d
ies
d
id
n
o
t
im
p
lem
en
t
a
n
y
f
o
r
m
o
f
f
ea
tu
r
e
s
elec
tio
n
d
u
r
in
g
m
o
d
el
c
o
n
s
tr
u
ctio
n
th
at
m
ay
co
m
p
r
o
m
is
e
b
o
th
m
o
d
el
ef
f
ici
en
cy
an
d
in
ter
p
r
etab
ilit
y
.
Ad
d
itio
n
ally
,
Su
b
ash
in
i
et
a
l.
[
5
8
]
h
ig
h
lig
h
ted
s
ev
er
al
ch
allen
g
es
in
p
h
is
h
in
g
d
etec
t
io
n
,
in
clu
d
in
g
im
b
alan
ce
d
d
a
tasets
th
at
ca
n
lead
to
b
iased
class
if
ier
s
an
d
an
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
6
,
Decem
b
e
r
20
25
:
5
8
6
3
-
5
8
7
8
5872
in
cr
ea
s
ed
r
is
k
o
f
f
alse
n
e
g
ativ
es.
Mo
r
eo
v
er
,
attac
k
e
r
s
o
f
te
n
lev
er
ag
e
en
cr
y
p
ted
tr
af
f
ic
t
o
co
n
ce
al
m
alicio
u
s
ac
tiv
ities
.
E
v
asio
n
tech
n
iq
u
es,
s
u
ch
as
UR
L
o
b
f
u
s
ca
tio
n
an
d
ad
v
er
s
ar
ial
tactics,
f
u
r
th
er
c
o
m
p
licate
d
etec
tio
n
b
y
en
a
b
lin
g
p
h
is
h
in
g
attem
p
ts
to
b
y
p
ass
ML
m
o
d
els.
T
h
is
s
tu
d
y
f
o
cu
s
es
o
n
t
h
r
ee
f
u
n
d
am
e
n
tal
asp
ec
ts
o
f
p
h
is
h
in
g
web
s
ite
d
etec
tio
n
:
id
e
n
tify
in
g
s
tate
-
of
-
th
e
-
ar
t
tech
n
iq
u
es
f
o
r
p
h
is
h
in
g
web
s
ite
d
etec
tio
n
,
e
v
alu
atin
g
th
e
co
n
tr
i
b
u
tio
n
o
f
f
ea
t
u
r
e
e
n
g
in
ee
r
in
g
a
n
d
XAI
in
im
p
r
o
v
i
n
g
th
e
p
e
r
f
o
r
m
an
ce
an
d
in
ter
p
r
eta
b
ilit
y
o
f
p
h
is
h
in
g
web
s
ite
d
etec
tio
n
s
y
s
tem
s
,
a
n
d
d
elin
ea
tin
g
k
ey
ch
allen
g
es
an
d
lim
itatio
n
s
o
f
ap
p
ly
in
g
f
ea
tu
r
e
e
n
g
in
ee
r
i
n
g
an
d
XAI
in
p
h
is
h
in
g
web
s
ites
d
etec
tio
n
.
T
h
ese
f
o
ca
l
ar
ea
s
f
ac
ilit
ate
a
c
o
m
p
r
e
h
en
s
iv
e
u
n
d
er
s
tan
d
i
n
g
o
f
th
e
d
o
m
ain
an
d
in
f
o
r
m
f
u
tu
r
e
a
d
v
an
ce
m
en
ts
.
As
y
o
u
s
ee
in
s
ec
tio
n
f
o
u
r
,
s
tate
-
of
-
th
e
-
ar
t
ap
p
r
o
ac
h
es
f
o
r
p
h
is
h
in
g
web
s
ite
d
etec
tio
n
lev
er
ag
e
im
p
r
o
v
em
e
n
ts
in
ML
,
DL
,
an
d
h
y
b
r
id
tech
n
i
q
u
es
to
r
eso
lv
e
t
h
e
c
o
m
p
licatio
n
o
f
cu
r
r
en
t
p
h
is
h
in
g
s
tr
ateg
ies.
ML
m
o
d
els,
s
u
ch
as
SVM
[
5
9
]
a
n
d
R
F
[
5
2
]
em
p
lo
y
f
ea
tu
r
es
li
k
e
UR
L
len
g
th
,
s
u
s
p
icio
u
s
k
ey
wo
r
d
s
f
o
r
b
in
a
r
y
class
if
icatio
n
task
s
.
Featu
r
e
en
g
in
ee
r
in
g
tec
h
n
iq
u
es
s
u
ch
as
PC
A
an
d
R
FE
r
aise
th
e
ac
cu
r
ac
y
o
f
d
etec
tio
n
b
y
elim
in
atin
g
r
ed
u
n
d
an
t
o
r
ir
r
elev
a
n
t
attr
ib
u
tes.
On
th
e
o
th
e
r
h
an
d
,
DL
ap
p
r
o
ac
h
es,
in
clu
d
i
n
g
C
NNs
[
2
1
]
a
n
d
L
T
SM
[
6
0
]
,
ar
e
ca
p
ab
le
o
f
lear
n
in
g
h
ig
h
-
d
im
en
s
io
n
al
r
ep
r
esen
tatio
n
s
d
ir
ec
tly
f
r
o
m
r
aw
in
p
u
ts
,
o
b
v
i
atin
g
th
e
n
ee
d
f
o
r
ex
ten
s
iv
e
m
an
u
al
f
ea
t
u
r
e
en
g
in
ee
r
in
g
.
XAI
to
o
ls
wer
e
r
ec
en
tly
ad
d
e
d
in
to
th
ese
s
y
s
tem
s
to
im
p
r
o
v
e
tr
an
s
p
ar
en
cy
.
Mo
r
eo
v
er
,
o
u
r
f
in
d
in
g
s
ad
d
n
u
an
ce
b
y
s
h
o
win
g
h
o
w
in
te
g
r
atio
n
o
f
f
ea
tu
r
e
en
g
in
ee
r
in
g
an
d
XAI
m
ec
h
an
is
m
s
im
p
r
o
v
es
th
e
ac
c
u
r
ac
y
a
n
d
in
ter
p
r
eta
b
ilit
y
o
f
p
h
is
h
in
g
we
b
s
ite
d
etec
tio
n
b
y
h
an
d
lin
g
is
s
u
es
in
th
e
d
o
m
ain
th
at
d
is
cu
s
s
ed
in
s
ec
tio
n
f
o
u
r
.
Featu
r
e
en
g
in
ee
r
in
g
en
ab
les
th
e
s
elec
tio
n
o
f
r
el
ev
an
t
f
ea
tu
r
es
u
s
in
g
tech
n
iq
u
es
s
u
ch
as
PC
A
[
6
1
]
an
d
FS
M
[
4
8
]
,
th
e
r
eb
y
r
e
d
u
ci
n
g
d
im
en
s
io
n
ality
a
n
d
im
p
r
o
v
in
g
g
en
er
aliza
tio
n
.
B
y
f
o
cu
s
in
g
o
n
r
ele
v
an
t
attr
ib
u
tes,
p
h
is
h
in
g
we
b
s
ites
d
etec
tio
n
r
ea
c
h
im
p
r
o
v
ed
ac
cu
r
ac
y
[
6
1
]
.
XAI
m
eth
o
d
s
,
s
u
ch
as
SHAP
[
5
6
]
an
d
L
I
M
E
[
6
2
]
o
f
f
er
i
n
g
in
ter
p
r
etab
ilit
y
to
u
n
d
er
s
tan
d
th
e
m
o
d
el
d
e
cisi
o
n
s
an
d
th
er
ef
o
r
e
in
cr
ea
s
e
tr
u
s
t
in
th
e
s
y
s
tem
.
I
n
teg
r
atio
n
tr
ea
ts
ch
allen
g
es
lik
e
ze
r
o
-
d
a
y
attac
k
s
,
an
d
i
m
p
r
o
v
e
r
eliab
ilit
y
,
lead
in
g
to
m
o
r
e
r
o
b
u
s
t,
in
ter
p
r
etab
le,
an
d
u
s
er
-
f
r
ien
d
ly
p
h
is
h
in
g
web
s
ites
d
etec
tio
n
s
y
s
tem
s
[
5
3
]
.
Ou
r
r
ev
iew
also
id
en
tifie
s
r
ec
u
r
r
in
g
ch
allen
g
es.
On
e
o
f
th
e
m
ain
ch
allen
g
es
is
f
ea
tu
r
e
s
el
ec
tio
n
,
as
s
p
ec
if
y
in
g
th
e
e
f
f
ec
tiv
e
attr
ib
u
tes is
co
m
p
lex
d
u
e
to
th
e
n
atu
r
e
o
f
p
h
is
h
in
g
web
s
ites
tech
n
iq
u
es,
an
d
ir
r
ele
v
an
t
f
ea
tu
r
es
ca
n
af
f
ec
t
t
h
e
p
e
r
f
o
r
m
an
ce
o
f
m
o
d
el
[
6
1
]
.
Mo
r
e
o
v
er
,
th
e
co
m
p
le
x
ity
o
f
h
i
g
h
-
d
i
m
en
s
io
n
al
d
ata
lead
s
to
in
cr
ea
s
ed
co
m
p
u
tatio
n
al
co
s
ts
an
d
r
ed
u
ce
d
m
o
d
el
ef
f
icien
cy
[
5
8
]
.
Ad
d
itio
n
ally
,
attac
k
e
r
s
d
ev
elo
p
th
e
latest
tech
n
iq
u
es,
s
u
ch
as
o
b
f
u
s
ca
ti
o
n
attac
k
s
,
m
ak
i
n
g
it
d
if
f
icu
l
t
f
o
r
s
tatic
f
ea
tu
r
e
s
ets
to
r
e
m
ain
ef
f
ec
tiv
e
o
v
er
tim
e.
Mo
r
e
o
v
er
,
tr
ad
e
-
o
f
f
ex
i
s
ts
b
etwe
en
ac
cu
r
ac
y
a
n
d
in
t
er
p
r
etab
ilit
y
:
wh
ile
DL
m
o
d
e
ls
o
f
ten
f
u
n
ctio
n
as
b
lack
b
o
x
es,
m
ak
in
g
th
eir
d
ec
i
s
io
n
s
d
if
f
icu
lt
t
o
e
x
p
lain
;
c
o
n
v
er
s
ely
s
im
p
ler
m
o
d
els
ar
e
m
o
r
e
in
ter
p
r
eta
b
le
b
u
t
m
ay
lack
d
etec
tio
n
p
r
ec
is
io
n
[
6
2
]
.
An
o
th
er
is
s
u
e
is
s
ca
lab
ilit
y
an
d
r
ea
l
-
tim
e
p
r
o
ce
s
s
in
g
,
as
f
ea
tu
r
e
en
g
in
ee
r
in
g
an
d
e
x
p
lain
ab
ilit
y
tech
n
iq
u
es
s
h
o
u
ld
wo
r
k
wi
th
in
h
ig
h
-
t
r
af
f
ic
d
o
m
ain
s
with
o
u
t
an
y
d
etec
tio
n
d
elay
s
[
4
1
]
.
Fin
ally
,
th
e
a
b
s
en
ce
o
f
s
tan
d
ar
d
ized
d
at
asets
an
d
ev
alu
atio
n
p
r
o
to
co
ls
also
h
in
d
er
s
r
ep
r
o
d
u
cib
ilit
y
an
d
co
n
s
is
ten
t
b
en
ch
m
ar
k
in
g
ac
r
o
s
s
s
tu
d
ies
[
6
3
]
.
T
h
is
s
tu
d
y
estab
lis
h
es
a
r
o
b
u
s
t
b
asis
f
o
r
ac
ad
em
ic
r
esea
r
ch
a
n
d
o
r
g
a
n
izatio
n
al
ap
p
li
ca
tio
n
b
y
in
teg
r
atin
g
tech
n
o
lo
g
ical
im
p
r
o
v
em
en
ts
with
p
r
ac
tical
s
ec
u
r
ity
r
eq
u
ir
em
e
n
ts
.
Fro
m
an
ac
ad
em
ic
s
tan
d
p
o
in
t,
th
e
am
alg
am
atio
n
o
f
f
ea
tu
r
e
e
n
g
in
ee
r
in
g
with
XAI
cr
ea
tes
o
p
p
o
r
tu
n
ities
f
o
r
th
e
a
d
v
an
ce
m
en
t
o
f
in
ter
p
r
etab
le
m
ac
h
in
e
lear
n
in
g
m
o
d
els
th
at
r
ec
o
n
cile
p
er
f
o
r
m
an
ce
with
tr
an
s
p
ar
en
cy
—
an
im
p
er
ativ
e
f
ac
to
r
in
cr
itical
f
ield
s
s
u
ch
as
cy
b
er
s
ec
u
r
ity
.
T
h
e
in
teg
r
atio
n
o
f
h
y
b
r
id
t
ec
h
n
iq
u
es,
class
if
icatio
n
o
f
p
h
is
h
in
g
d
etec
tio
n
m
ea
s
u
r
es,
an
d
ex
am
in
atio
n
o
f
XAI
m
eth
o
d
s
s
u
ch
as
SH
A
P
an
d
L
I
ME
en
h
an
ce
co
m
p
r
eh
en
s
io
n
o
f
m
o
d
el
b
eh
av
io
r
a
n
d
s
y
s
tem
wea
k
n
ess
es.
T
h
e
f
in
d
in
g
s
b
r
in
g
u
s
ef
u
l
in
s
ig
h
ts
f
o
r
cy
b
er
s
ec
u
r
ity
ex
p
er
ts
,
em
p
h
asizin
g
th
e
s
ig
n
if
ican
ce
o
f
d
ev
elo
p
in
g
d
etec
tio
n
s
y
s
tem
s
th
at
ar
e
b
o
th
r
eliab
le
a
n
d
co
m
p
r
e
h
en
s
ib
le,
as
well
as
ad
ap
tiv
e
to
em
er
g
in
g
th
r
ea
ts
.
T
h
ese
in
s
ig
h
ts
en
ab
le
tr
u
s
t
am
o
n
g
s
tak
eh
o
ld
er
s
,
b
o
o
s
t
in
cid
en
t
r
esp
o
n
s
e
m
eth
o
d
s
,
an
d
g
u
a
r
an
tee
co
m
p
lian
ce
with
r
eg
u
latio
n
s
o
f
cy
b
e
r
s
ec
u
r
ity
.
T
h
e
r
ev
iew
u
ltima
tely
r
ec
o
m
m
en
d
s
f
o
r
th
e
f
u
r
th
er
d
ev
elo
p
m
en
t
o
f
r
e
liab
le
n
ex
t
-
g
en
er
atio
n
AI
s
y
s
tem
s
an
d
p
r
o
v
i
d
es
a
f
r
am
ewo
r
k
f
o
r
o
r
g
an
izatio
n
s
aim
in
g
to
en
h
an
ce
th
ei
r
d
ig
ital
d
ef
en
s
es a
g
ain
s
t p
r
o
g
r
ess
iv
ely
in
tr
icate
p
h
is
h
in
g
th
r
ea
ts
.
6.
F
UT
UR
E
DIR
E
C
T
I
O
NS
T
h
e
co
m
b
in
atio
n
o
f
f
ea
tu
r
e
e
n
g
in
ee
r
in
g
an
d
XAI
in
p
h
is
h
in
g
web
s
ite
d
etec
tio
n
is
p
r
o
g
r
e
s
s
in
g
,
with
s
o
m
e
f
u
tu
r
e
d
ir
e
ctio
n
s
to
en
h
an
ce
ac
cu
r
ac
y
a
n
d
tr
an
s
p
a
r
en
cy
.
First
d
o
m
ain
is
ap
p
ly
in
g
au
to
m
ated
f
ea
tu
r
e
en
g
in
ee
r
in
g
tech
n
iq
u
es
an
d
r
e
d
u
cin
g
th
e
d
e
p
en
d
e
n
ce
o
f
m
a
n
u
al
f
ea
tu
r
e
en
g
in
ee
r
in
g
in
r
e
s
p
o
n
s
e
to
ev
o
lv
in
g
p
h
is
h
in
g
tactics
in
em
p
l
o
y
in
g
ML
m
o
d
els
[
1
]
.
A
d
d
itio
n
ally
,
im
p
r
o
v
in
g
th
e
ex
p
lain
a
b
ilit
y
o
f
th
ese
m
o
d
els
is
also
cr
itical;
ap
p
ly
in
g
ad
v
a
n
ce
d
XAI
m
eth
o
d
s
,
s
u
ch
a
s
SHAP
an
d
L
I
ME
,
ca
n
p
r
o
v
id
e
in
te
r
p
r
etab
le
class
if
icatio
n
s
wi
th
o
u
t
af
f
ec
tin
g
s
ec
u
r
ity
[
6
4
]
.
An
o
th
er
cr
i
tical
f
o
cu
s
is
m
ak
in
g
s
u
r
e
th
a
t
d
etec
tio
n
s
y
s
tem
s
o
p
er
ate
ef
f
icien
tly
in
r
ea
l
-
ti
m
e
en
v
ir
o
n
m
en
ts
with
m
in
i
m
u
m
laten
cy
wh
ic
h
r
eq
u
ir
ed
lig
h
tweig
h
t
f
ea
tu
r
e
en
g
in
ee
r
in
g
alg
o
r
ith
m
s
an
d
o
p
tim
ized
XAI
m
eth
o
d
s
.
Mo
r
e
o
v
er
,
t
h
e
s
tu
d
ies
will
co
n
ce
n
t
r
ate
o
n
im
p
r
o
v
in
g
p
h
is
h
in
g
d
etec
tio
n
tech
n
i
q
u
e
s
f
o
r
s
o
cial
n
etwo
r
k
i
n
g
a
n
d
m
o
b
ile
p
latf
o
r
m
s
b
y
d
ev
el
o
p
in
g
s
o
p
h
is
ticated
Evaluation Warning : The document was created with Spire.PDF for Python.