T
E
L
K
O
MN
I
K
A
T
el
eco
m
m
u
n
i
ca
t
i
o
n
,
C
o
m
p
u
t
i
n
g
,
E
l
ect
ro
n
i
cs
a
n
d
C
o
n
t
ro
l
Vo
l
.
1
9
, N
o
.
5
,
O
ct
o
b
er
202
1
, p
p
.
17
07
~
1714
I
S
S
N
:
1693
-
6930,
a
c
c
r
e
di
t
e
d F
i
r
s
t
G
r
a
de
by K
e
m
e
nr
i
s
t
e
kdi
kt
i
,
D
e
c
r
e
e
N
o:
21/
E
/
K
P
T
/
2018
D
O
I
:
10.
12928/
T
E
L
K
O
M
N
I
K
A
.
v1
9
i
5
.
15981
1707
Jou
r
n
al
h
om
e
page
:
ht
t
p:
/
/
j
our
nal
.
uad
.
ac
.
i
d/
i
nde
x
.
php/
T
E
L
K
O
M
N
I
K
A
A
m
a
x
i
m
u
m
e
nt
ro
p
y
cl
a
s
s
i
f
i
c
a
t
i
o
n s
c
hem
e f
o
r p
hi
s
hi
ng
det
ec
t
i
o
n us
i
ng
pa
rs
i
m
o
ni
o
us
f
ea
t
ur
es
E
m
m
a
nue
l
O
.
A
s
a
ni
1
,
A
d
e
b
ayo O
m
ot
os
h
o
2
,
P
a
u
l
A
.
D
a
nqua
h
3
,
Joyc
e
A
.
A
yool
a
4
,
P
ea
ce O
.
A
y
eg
b
a
5
,
Olu
m
id
e
B
.
L
o
n
g
e
6
1,
4,
5
De
pa
r
tm
e
n
t of
C
om
pu
te
r
S
c
ie
nc
e
,
L
a
ndm
a
r
k U
ni
ve
r
sit
y,
Om
u
-
Ar
a
n,
Ni
ge
r
ia
1
S
DG
11: S
u
sta
ina
bl
e
C
it
ie
s a
nd C
om
m
un
it
ie
s R
e
se
a
r
c
h
gr
ou
p
,
L
a
ndm
a
r
k Un
ive
r
si
ty
,
O
mu
-
Ar
a
n,
Ni
ge
r
ia
2
I
nte
r
ne
t Te
c
hn
ol
og
ie
s a
nd I
n
te
r
n
e
t S
y
ste
m
s R
e
se
a
r
c
h G
r
ou
p,
Ha
s
so P
la
tt
ne
r
I
n
st
it
ute
,
P
o
ts
da
m
,
G
e
r
m
a
ny
3
Co
un
c
i
l f
or
S
c
ie
nt
if
ic
a
n
d I
n
du
str
ia
l R
e
se
a
r
c
h
-
I
ns
ti
tu
te
f
or
S
c
ie
n
tif
ic
a
n
d
Te
c
h
no
lo
gic
a
l
I
nf
or
m
a
t
io
n,
Ac
c
r
a
,
G
ha
na
6
S
c
hoo
l of
C
om
pu
ta
ti
ona
l S
c
ie
nc
e
s a
n
d I
nf
or
m
a
tic
s,
Ac
a
de
m
ic
C
i
ty
Un
ive
r
si
ty C
o
lle
ge
,
Ac
c
r
a
,
G
ha
na
A
rt
i
cl
e I
n
f
o
AB
S
T
RACT
A
r
tic
le
h
i
st
o
ry
:
R
ecei
v
ed
Ma
r
0
3
,
2020
R
ev
i
s
ed
J
un 29
,
202
1
A
ccep
t
ed
J
ul
13
,
202
1
Ove
r
t
he
ye
a
r
s,
e
le
c
tr
on
ic
m
a
il (
e
-
m
a
il)
ha
s be
e
n t
he
ta
r
ge
t of
se
ve
r
a
l
m
a
lic
io
us a
t
ta
c
k
s.
P
hi
sh
in
g is
one
of
the
m
o
st r
e
c
og
niz
a
ble
f
or
m
s of
m
a
ni
pu
la
t
io
n a
im
e
d a
t e
-
m
a
i
l u
se
r
s a
nd
us
ua
l
ly,
e
m
pl
o
ys s
oc
ia
l e
n
gi
ne
e
r
i
ng
to tr
ic
k in
noc
e
nt u
se
r
s in
to s
up
pl
yi
ng s
e
ns
it
ive
inf
or
m
a
tio
n in
to a
n im
pos
te
r
we
b
si
te
.
At
ta
c
k
s f
r
om
ph
is
hi
ng e
m
a
il
s c
a
n r
e
su
lt
in t
he
e
x
po
sur
e
of
c
onf
ide
nt
ia
l
inf
or
m
a
t
io
n,
f
ina
nc
ia
l l
os
s,
da
ta
m
i
su
se
,
a
nd
o
the
r
s.
T
hi
s pa
pe
r
pr
e
se
nt
s
the
im
p
le
m
e
n
ta
t
io
n
of
a
m
a
x
im
um
e
n
tr
o
py
(ME
)
c
la
s
sif
ic
a
t
io
n
met
h
od f
or
a
n e
f
f
ic
ie
n
t a
p
pr
oa
c
h t
o t
he
i
de
n
tif
ic
a
t
io
n of
phi
sh
in
g e
m
a
i
ls.
Our
r
e
su
lt sh
owe
d tha
t m
a
xim
um
e
ntr
op
y wit
h pa
r
s
im
o
ni
ou
s f
e
a
tur
e
spa
c
e
g
ive
s
a
be
t
te
r
c
la
ss
if
ic
a
ti
on
pr
e
c
i
si
on
tha
n b
ot
h t
he
Na
ïve
B
a
y
e
s a
n
d s
up
por
t ve
c
tor
mach
i
n
e
(S
V
M).
Ke
y
wo
r
d
s
:
C
la
s
s
if
ic
a
tio
n
P
a
r
s
i
m
onous
f
e
a
t
ur
e
s
P
hi
s
hi
ng
S
o
ci
al
en
g
i
n
eer
i
n
g
T
his
is
a
n
o
pe
n
ac
c
e
s
s
ar
tic
le
u
nde
r
the
CC
B
Y
-
SA
lic
e
n
se
.
C
or
r
e
s
pon
di
n
g A
u
t
h
or
:
E
m
m
an
u
el
O
.
A
s
an
i
D
ep
ar
t
m
en
t
o
f
C
o
m
p
u
t
er
S
ci
en
ce
L
a
ndm
a
r
k U
ni
ve
r
s
i
t
y
Om
u
-
A
r
an
,
N
ig
e
r
ia
E
ma
il:
as
an
i
.
em
m
an
u
el
@
l
m
u
.
ed
u
.
n
g
1.
I
NT
RO
DUC
T
I
O
N
O
w
i
ng pe
r
ha
ps
,
t
o i
t
s
ubi
qui
t
y a
nd
l
i
m
i
t
l
e
s
s
pot
e
nt
i
a
l
s
f
or
c
om
m
uni
c
a
t
i
ons
a
nd i
n
t
e
r
c
onne
c
t
i
vi
t
y,
t
he
i
nt
e
r
ne
t
ha
s
be
c
om
e
a
hom
e
f
or
di
ve
r
ge
nt
t
e
nde
nc
i
e
s
t
ha
t
e
nge
nde
r
bot
h pos
i
t
i
ve
a
nd ne
ga
t
i
ve
p
r
a
c
t
i
c
e
s
;
w
hi
l
e
i
nt
e
r
ne
t
t
e
c
hnol
ogy i
s
a
t
t
he
c
ut
t
i
ng e
dge
of
gr
e
a
t
i
nnova
t
i
ons
a
nd r
e
vol
ut
i
ona
r
y f
i
ndi
ngs
,
c
r
i
m
i
na
l
s
a
r
e
e
qua
l
l
y a
bl
e
t
o de
pl
oy t
hi
s
t
e
c
hnol
ogy f
or
e
a
s
y p
r
opa
ga
t
i
on a
nd pe
r
pe
t
r
a
t
i
on of
t
h
ei
r
cr
i
m
i
n
al
ag
en
d
a w
i
t
h
u
n
i
v
er
s
al
r
each
[
1]
.
T
h
e e
-
ma
il,
b
e
in
g
th
e
p
r
e
d
o
min
a
n
t me
a
n
s
o
f
c
o
mmu
n
ic
a
tio
n
w
ith
o
v
e
r
3
bi
l
l
i
on a
c
t
i
ve
u
s
er
s
,
h
as
b
eco
m
e a v
er
i
t
ab
l
e m
ed
i
u
m
o
f
ch
o
i
ce f
o
r
cy
b
er
cr
i
m
i
n
al
s
[
2
]
.
T
h
u
s
,
cy
b
er
cr
i
m
es
p
r
o
l
i
f
er
at
e
ve
r
y r
a
pi
dl
y a
nd ha
ve
t
he
pot
e
nt
i
a
l
s
t
o c
a
us
e
i
m
m
e
ns
e
da
m
a
ge
t
o bot
h i
ndi
vi
dua
l
s
a
nd
c
or
por
a
t
e
or
ga
ni
z
a
t
i
ons
[
3
] [
4
]
;
phi
s
hi
ng i
s
pe
r
ha
ps
t
he
m
os
t
popul
a
r
of
t
he
s
e
c
r
i
m
e
s
.
P
hi
s
hi
ng i
s
a
n a
t
t
a
c
k ve
c
t
or
t
ha
t
de
pl
oys
t
e
c
hni
c
a
l
s
ubt
e
r
f
uge
a
nd s
oc
i
a
l
e
ngi
ne
e
r
i
ng t
o
s
ur
r
e
pt
i
t
i
ous
l
y ob
t
a
i
n ot
he
r
w
i
s
e
pe
r
s
ona
l
a
nd s
e
ns
i
t
i
ve
i
nf
or
m
a
t
i
on s
uc
h a
s
c
r
e
di
t
c
a
r
d pi
ns
,
a
nd
u
s
e
r ID
s
[
5
]
.
U
n
s
u
s
p
ect
i
n
g
u
s
er
s
ar
e l
u
r
ed
b
y
cr
i
m
i
n
al
el
em
en
t
s
,
m
as
q
u
er
ad
i
n
g
as
l
eg
i
t
i
m
at
e
en
t
i
t
i
es
v
i
a
el
ect
r
o
n
i
c
c
om
m
uni
c
a
t
i
on m
e
di
a
t
o di
vul
ge
vi
t
a
l
,
pe
r
s
ona
l
,
of
t
e
n,
f
i
na
nc
i
a
l
i
nf
or
m
a
t
i
on
,
w
hi
c
h
m
a
y,
i
n
t
ur
n
,
be
us
e
d
ille
g
a
lly
b
y
th
e
c
r
imin
a
ls
w
i
th
o
u
t th
e
k
n
o
w
le
dge
or
c
ons
e
nt
of
t
he
r
e
a
l
ow
ne
r
s
.
P
hi
s
hi
ng i
s
a
n
i
ns
t
a
nc
e
of
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SSN
:
1693
-
6930
T
E
L
KOM
NI
KA
T
el
eco
m
m
u
n
C
om
put
E
l
C
ont
r
o
l
,
Vo
l
.
1
9
, N
o
.
5
,
O
ct
o
b
er
2021
:
17
07
-
17
14
1708
id
e
n
tity
th
e
f
t [
6
]
.
T
h
e
p
h
is
h
in
g
c
y
c
le
o
f
te
n
s
ta
r
t
s
w
ith
a
n
e
ma
il th
a
t r
e
p
lic
a
te
s
th
e
id
e
n
tity
o
f
a
tr
u
s
te
d
a
s
s
oc
i
a
t
e
or
or
ga
ni
z
a
t
i
on of
t
e
n w
i
t
h a
bogus
but
j
ui
c
y c
l
a
i
m
t
o a
r
e
w
a
r
d f
or
t
he
uns
us
pe
c
t
i
ng r
e
c
i
pi
e
nt
,
or
i
n
ot
he
r
i
ns
t
a
nc
e
s
,
a
dubi
ous
r
e
va
l
i
da
t
i
on e
xe
r
c
i
s
e
b
y e
l
e
m
e
nt
s
pos
i
ng a
s
f
i
na
nc
i
a
l
i
ns
t
i
t
ut
i
ons
,
de
m
a
n
di
ng t
ha
t
us
e
r
s
s
uppl
y t
he
i
r
a
ut
he
nt
i
c
a
t
i
on de
t
a
i
l
s
.
F
o
r
i
n
s
t
a
n
ce,
an
at
t
ack
er
m
ay
d
es
cr
i
b
e a
p
r
o
b
l
em
(
u
s
u
al
l
y
g
en
er
i
c)
,
w
hi
c
h i
n s
om
e
c
a
s
e
s
m
a
y a
ppl
y t
o t
he
uns
us
pe
c
t
i
ng r
e
c
i
pi
e
nt
.
T
he
y t
he
n pr
oc
e
e
d t
o pr
opos
e
a
s
ol
ut
i
o
n,
w
hi
c
h
u
s
u
a
lly
in
c
lu
d
e
a
lin
k
f
o
r
f
illin
g
o
u
t s
e
n
s
itiv
e
d
e
t
a
ils
,
a
n
d
a
lin
k
to
r
e
je
c
t
th
e
o
f
f
e
r
,
to
g
iv
e
th
e
e
m
a
il s
o
me
m
odi
c
um
of
a
ut
he
nt
i
c
i
t
y.
H
a
vi
ng
t
a
ke
n t
he
ba
i
t
,
t
he
us
e
r
i
s
m
a
de
t
o
f
i
l
l
out
pe
r
s
ona
l
da
t
a
s
uc
h a
s
ba
nk
a
c
c
ount
P
I
N
,
s
oc
i
a
l
s
e
c
ur
i
t
y num
be
r
,
or
s
om
e
ot
h
e
r
us
e
f
ul
a
ut
he
nt
i
c
a
t
i
on de
t
a
i
l
s
,
w
hi
c
h m
a
y be
us
e
d
by t
h
e
cr
i
m
i
n
a
l
s
t
o
p
er
p
et
r
a
t
e i
l
l
eg
al
t
r
a
n
s
a
ct
i
o
n
s
l
a
t
er
.
F
i
g
u
r
e 1
s
h
ow
s
t
he
ph
i
s
hi
ng
e
-
ma
il lif
e
cy
cl
e,
w
h
i
l
e
F
i
g
ur
e
2
s
how
s a
of
t
yp
i
c
a
l
i
nf
or
m
a
t
i
o
n c
o
m
p
on
e
n
t
i
n
a
ph
i
s
h
i
ng
e
m
a
il.
P
hi
s
hi
ng a
t
t
a
c
ks
pos
e
s
e
r
i
ous
r
i
s
ks
t
o bot
h i
n
di
vi
dua
l
s
a
nd c
or
por
a
t
e
e
nt
i
t
i
e
s
a
nd ha
ve
di
r
e
c
ons
e
que
nc
e
s
on gl
oba
l
s
e
c
ur
i
t
y a
nd
t
he
eco
n
o
m
y
[
7
]
.
I
t
i
s
e
ve
n m
or
e
s
o da
nge
r
o
u
s
,
as
i
t
ap
p
ear
s
t
h
at
phi
s
he
r
s
c
ont
i
nue
t
o
pe
r
f
e
c
t
m
e
a
ns
t
o
out
m
a
ne
u
ve
r
e
ve
n
t
he
know
l
e
dge
a
bl
e
a
nd
s
e
c
ur
i
t
y
-
c
ons
c
i
ous
[
8
]
;
t
e
c
hnol
ogy gi
a
nt
s
s
uc
h a
s
G
oogl
e
a
nd
F
a
c
e
book ha
ve
l
os
t
a
bout
$100
m
i
l
l
i
on
t
o
phi
s
hi
ng e
m
a
i
l
s
f
r
om
ha
c
ke
r
s
w
ho i
m
pe
r
s
ona
t
e
d a
s
ha
r
dw
a
r
e
ve
ndor
s
i
n
2017.
T
he
e
c
onom
i
c
e
f
f
e
c
t
of
phi
s
hi
ng
a
t
t
a
c
k i
s
e
n
or
m
ous
;
r
e
por
t
s
ha
ve
s
how
n
t
ha
t
f
i
na
nc
i
a
l
l
os
s
oc
c
a
s
i
one
d by phi
s
hi
ng a
t
t
a
c
ks
e
xc
e
e
ds
$
5
b
illio
n
g
lo
b
a
lly
[
9
]
.
F
i
gur
e
1.
L
i
f
e
c
yc
l
e
of
a
phi
s
hi
ng e
m
a
i
l
[
10]
F
i
gur
e
2.
T
yp
i
c
a
l
i
nf
o
r
m
a
t
i
on c
om
pone
nt
i
n a
phi
s
hi
ng
e
ma
il [
1
1
]
P
hi
s
hi
ng a
t
t
a
c
ke
r
s
a
r
e
i
nc
r
e
a
s
i
ngl
y be
c
om
i
ng m
or
e
r
e
s
i
l
i
e
nt
ove
r
t
he
ye
a
r
s
,
due
t
o t
he
a
l
a
r
m
i
ng
i
nc
r
e
a
s
e
i
n t
he
vol
um
e
of
a
t
t
a
c
k a
nd
t
he
i
nnova
t
i
ve
ne
s
s
w
i
t
h w
hi
c
h t
he
a
t
t
a
c
ks
a
r
e
be
i
ng
i
m
p
le
me
n
te
d
.
S
e
c
ur
i
t
y s
pe
c
i
a
l
i
s
t
s
a
nd phi
s
he
r
s
a
r
e
i
n a
v
i
c
i
ous
c
i
r
c
l
e
be
c
a
us
e
a
ppr
e
he
ndi
ng phi
s
he
r
s
ha
ve
be
c
om
e
m
or
e
an
d
m
o
r
e co
m
p
l
i
cat
ed
.
P
h
i
s
h
er
s
ar
e co
n
s
t
an
t
l
y
ch
an
g
i
n
g
t
h
ei
r
t
act
i
cs
t
o
d
ef
eat
an
t
i
-
phi
s
hi
ng t
e
c
hni
que
s
[
1
2
]
.
T
he
a
ggr
e
ga
t
e
num
be
r
o
f
di
s
t
i
nc
t
l
y r
e
c
ogni
z
e
d phi
s
hi
ng a
t
t
a
c
ks
r
e
a
c
he
d a
pe
a
k of
263,
538 a
t
t
a
c
ks
i
n
t
he
f
i
r
s
t
qua
r
t
e
r
of
2018
;
a
n a
l
a
r
m
i
ng
ups
ur
ge
f
r
om
180,
57
7 r
e
por
t
e
d i
n
t
he
l
a
s
t
qua
r
t
e
r
of
2017
(
A
P
W
G
,
20
18)
[
1
3
].
T
he
e
m
a
i
l
ha
s
a
l
s
o be
e
n i
de
nt
i
f
i
e
d a
s
t
he
t
op
phi
s
hi
ng t
a
r
ge
t
;
c
ons
e
que
n
tly
,
a
p
h
is
h
in
g
e
ma
il
a
tta
c
k
a
ime
d
a
t
i
ndi
vi
dua
l
s
a
nd c
or
por
a
t
e
bodi
e
s
i
s
on
t
he
r
i
s
e
[
1
4
].
S
e
ve
r
a
l
i
nt
e
r
ve
nt
i
ons
ha
ve
be
e
n m
a
de
ove
r
t
he
ye
a
r
s
t
o c
om
ba
t
t
he
ph
i
s
hi
ng m
e
na
c
e
.
Q
a
ba
j
e
h
e
t
a
l.
[1
5
]
i
de
nt
i
f
i
e
d s
om
e
t
e
c
hni
que
s
bot
h
‘
t
r
a
di
t
i
ona
l
’
a
nd
‘
c
om
put
e
r
iz
e
d
’
in
li
te
r
a
tu
r
e
.
S
o
me
t
r
a
d
itio
n
a
l
a
n
ti
-
phi
s
hi
ng
t
e
c
hni
que
s
l
i
ke
e
nf
or
c
i
ng l
a
w
s
,
e
qui
ppi
ng
us
e
r
s
w
i
t
h know
l
e
dge
,
a
nd e
duc
a
t
i
ng
t
he
publ
i
c
w
e
r
e
m
e
nt
i
one
d.
C
o
mp
u
te
r
iz
e
d
e
f
f
o
r
ts
in
c
lu
d
e
b
la
c
k
lis
ts
,
f
ilte
r
in
g
,
as
s
o
ci
at
i
v
e cl
as
s
i
f
i
cat
i
o
n
,
a
nd
r
ul
e
i
nduc
t
i
on
as
w
el
l
as
t
h
e
us
e
of
m
a
c
hi
ne
l
e
a
r
ni
ng
a
ppr
oa
c
he
s
vi
a
di
f
f
e
r
e
nt
c
l
a
s
s
i
f
i
c
a
t
i
on a
nd m
ode
l
-
ba
s
e
d t
e
c
hni
que
s
.
A
va
r
i
e
t
y of
s
ur
ve
ys
a
nd r
e
vi
e
w
s
of
a
nt
i
-
phi
s
hi
ng t
e
c
hni
que
s
ha
ve
a
l
s
o be
e
n doc
um
e
nt
e
d
i
n t
he
l
i
t
e
r
a
t
u
r
e
,
to
pr
ovi
d
e
be
t
t
e
r
unde
r
s
t
a
ndi
ng a
nd e
nha
nc
e
t
he
de
ve
l
opm
e
nt
of
be
t
t
e
r
a
nt
i
-
phi
s
hi
ng s
ys
t
e
m
s
.
A
l
e
r
oud a
nd Z
hou
[1
6
]
doc
um
e
nt
e
d s
om
e
an
t
i
-
phi
s
hi
ng t
e
c
hni
que
s
i
n e
m
a
i
l
s
,
w
e
bs
i
t
e
s
,
m
obi
l
e
d
ev
i
ces
as
w
el
l
as
s
o
ci
al
n
et
w
o
r
k
i
n
g
s
i
t
es
.
T
he
y t
he
n pr
opos
e
d a
ne
w
t
a
xonom
y of
phi
s
hi
ng a
t
t
a
c
ks
w
i
t
h a
n
e
m
pha
s
i
s
on t
he
t
a
r
ge
t
e
nvi
r
onm
e
nt
,
a
t
t
a
c
ki
ng
t
e
c
hni
que
s
,
c
om
m
uni
c
a
t
i
on m
e
di
a
,
a
nd
c
ount
e
r
m
e
a
s
ur
e
s
.
T
he
i
r
w
o
r
k of
f
e
r
e
d a
r
obus
t
a
pp
r
oa
c
h
to
i
de
nt
i
f
yi
ng phi
s
hi
ng a
t
t
a
c
ks
.
G
oe
l
a
nd J
a
i
n
[1
7
]
pr
o
vi
de
d a
c
l
a
s
s
i
f
i
c
a
t
i
on of
m
obi
l
e
phi
s
hi
ng a
t
t
a
c
ks
a
nd
s
u
gge
s
t
e
d be
t
t
e
r
m
e
t
hods
t
o
i
de
nt
i
f
y a
nd
e
ns
ur
e
p
r
ot
e
c
t
i
on
ag
ai
n
s
t
t
h
es
e at
t
a
ck
s
.
I
t
w
as
s
how
n t
ha
t
i
ndi
vi
dua
l
s
t
ha
t
us
e
m
obi
l
e
de
vi
c
e
s
we
r
e
m
or
e
l
i
ke
l
y t
o be
e
xpos
e
d t
o
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
KA
T
el
eco
m
m
u
n
C
om
put
E
l
C
ont
r
o
l
A
m
ax
i
m
um
e
nt
r
opy
c
l
as
s
i
f
i
c
at
i
on s
c
he
m
e
f
or
phi
s
hi
ng de
t
e
c
t
i
on us
i
ng
…
(
E
m
m
anue
l
O
.
A
s
ani
)
1709
phi
s
hi
ng a
t
t
a
c
ks
t
ha
n de
s
kt
op us
e
r
s
.
A
l
s
o,
due
t
o
di
f
f
e
r
e
nc
e
s
i
n
f
unc
t
i
ona
l
i
t
y a
nd
l
a
yout
o
f
t
he
de
vi
c
e
s
,
t
he
y
pr
opos
e
d
a d
ev
i
s
e
-
c
en
t
r
i
c
m
e
t
hod
t
ha
t
c
ons
i
de
r
s
t
he
de
vi
c
e
i
n us
e
an
d
w
as
ab
l
e t
o c
ount
e
r
m
obi
l
e
phi
s
hi
ng
at
t
ack
s
.
S
u
m
an
t
h
i
an
d
D
am
o
d
ar
am
[
1
8
]
s
ur
ve
ye
d
s
e
ve
nt
e
e
n phi
s
hi
ng d
et
ect
i
o
n
s
ch
em
es
w
i
t
h
a p
er
f
o
r
m
an
ce
ev
al
u
at
i
o
n
b
as
ed
o
n
s
ev
er
al
p
ar
am
et
er
s
w
h
i
ch
i
n
cl
u
d
e
d
accu
r
acy
,
p
r
eci
s
i
o
n
,
r
ecal
l
,
t
r
u
e n
eg
at
i
v
e
es
t
i
m
at
e,
tr
u
e
p
o
s
itiv
e
e
s
tima
te
,
f
a
ls
e
-
n
eg
at
i
v
e e
s
t
i
m
at
e,
an
d
f
al
s
e
-
p
o
s
itiv
e
e
s
tima
te
.
R
e
s
u
lt
s
in
d
ic
a
te
d
t
h
e
ta
r
g
e
t
va
l
i
da
t
i
on m
e
t
hod ha
d t
he
hi
ghe
s
t
a
c
c
ur
a
c
y of
99
.
54%
,
t
he
phi
s
hi
ng a
l
a
r
m
ha
d
t
he
hi
ghe
s
t
pr
e
c
i
s
i
on
of
100%
a
nd t
he
s
m
a
r
t
w
e
bs
i
t
e
c
om
bi
ne
d w
i
t
h
t
he
c
a
t
e
go
r
i
z
a
t
i
on m
ode
l
m
e
t
hod
f
o
r
phi
s
hi
ng de
t
e
c
t
i
on
pr
oduc
e
d a
98.
72%
r
e
c
a
l
l
va
l
ue
.
C
hi
e
w
e
t
a
l.
[1
9
]
p
r
es
en
t
ed
a r
o
bus
t
,
s
ys
t
e
m
a
t
i
c
r
e
vi
e
w
of
phi
s
hi
ng a
t
t
a
c
k a
nd t
he
i
r
as
s
o
ci
at
ed
v
e
ct
o
r
s
.
T
he
r
e
vi
e
w
s
how
e
d t
ha
t
phi
s
he
r
s
m
a
ke
us
e
of
t
e
c
hni
c
a
l
a
ppr
oa
c
he
s
s
u
c
h a
s
c
l
oud
c
om
put
i
ng,
c
l
i
c
kj
a
c
ki
ng,
a
nd m
a
l
ve
r
t
i
s
i
ng
i
n t
he
i
r
a
t
t
a
c
ks
a
nd t
he
de
ve
l
opm
e
nt
of
i
nt
e
l
l
i
ge
nt
s
ys
t
e
m
s
w
i
ll b
e
a
c
ount
e
r
m
e
a
s
ur
e
i
n
t
he
di
s
c
ove
r
y o
f
phi
s
hi
ng
t
hr
e
a
t
s
.
A
l
s
o,
a
r
e
vi
e
w
o
f
a
nt
i
-
p
hi
s
hi
ng m
e
t
hods
i
n l
i
t
e
r
a
t
ur
e
w
as
s
ugge
s
t
e
d f
or
t
he
de
ve
l
opm
e
nt
of
a
m
or
e
r
ob
us
t
t
e
c
hni
que
.
A
va
i
l
a
bl
e
a
ppr
oa
c
he
s
i
n l
i
t
e
r
a
t
u
r
e
h
a
ve
be
e
n
not
e
d t
o e
i
t
he
r
c
om
pr
om
i
s
e
pr
e
c
i
s
i
on t
o i
m
p
r
ov
e
r
e
s
pons
e
t
i
m
e
or
i
m
pr
ove
pr
e
c
i
s
i
on a
t
t
he
e
x
pe
ns
e
of
r
es
p
o
n
s
e t
i
m
e [
20
]
.
S
o
f
tw
a
r
e
p
h
is
h
in
g
d
e
te
c
tio
n
mo
d
e
ls
g
e
n
e
r
a
lly
in
c
lu
d
e
th
e
b
la
c
k
/w
h
ite
lis
t,
h
e
u
r
is
tic
s
,
m
a
c
hi
ne
l
e
a
r
ni
ng,
vi
s
ua
l
s
i
m
i
l
a
r
i
t
y,
a
nd hybr
i
d a
pp
r
oa
c
he
s
.
S
onow
a
l
an
d
K
uppus
a
m
y [
2
1
]
pr
opos
e
d a
phi
s
hi
ng de
t
e
c
t
i
on m
ode
l
c
a
l
l
e
d
P
hi
D
m
a
w
hi
c
h
us
e
d
a
mu
ltif
ilte
r
a
p
p
r
oa
c
h.
T
he
pr
opos
e
d
m
ode
l
e
m
pl
oy
e
d
t
he
us
e
o
f
a
n a
udi
o
-
ba
s
e
d i
ndi
c
a
t
or
m
a
ki
ng i
t
a
c
c
e
s
s
i
bl
e
by pe
opl
e
w
ho a
r
e
vi
s
ua
l
l
y i
m
pa
i
r
e
d.
L
e
gi
t
i
m
a
t
e
da
t
a
w
a
s
got
t
e
n f
r
om
phi
s
hl
oa
d a
nd
phi
s
hi
ng da
t
a
f
r
om
phi
s
ht
a
nk f
or
t
he
i
m
pl
e
m
e
nt
a
t
i
on.
T
he
r
e
s
ul
t
f
r
o
m
t
he
e
xpe
r
i
m
e
nt
s
how
e
d t
ha
t
t
he
m
ode
l
w
a
s
s
uc
c
e
s
s
f
ul
s
i
nc
e
a
t
r
ue
pos
i
t
i
ve
r
a
t
e
of
90.
54
%
a
nd
94
.
18
%
t
r
ue
ne
ga
t
i
ve
r
a
t
e
w
a
s
r
e
c
or
de
d.
A
l
s
o,
a
n
a
c
c
ur
a
c
y of
92.
72%
,
a
f
a
l
s
e
pos
i
t
i
ve
a
nd
f
a
l
s
e
ne
ga
t
i
ve
r
a
t
e
of
5
.
82%
a
nd 9.
46
%
r
e
s
pe
c
t
i
ve
l
y w
e
r
e
a
c
hi
e
ve
d by t
he
m
ode
l
.
V
ol
ka
m
e
r
e
t a
l.
[2
2
]
p
r
es
en
t
ed
a w
al
k
-
t
hr
ough a
na
l
ys
i
s
of
r
e
a
s
ons
w
hy pe
opl
e
f
a
l
l
pr
e
y t
o
phi
s
hi
ng
a
nd s
ugge
s
t
e
d a
c
onc
e
pt
t
o c
i
r
c
um
ve
nt
t
he
pr
oc
e
s
s
:
T
he
to
o
ltip
-
pow
e
r
e
d phi
s
h e
m
a
il d
e
te
c
tio
n
(
T
or
pe
do)
.
T
or
pe
do s
how
ed
t
he
o
r
i
gi
na
l
uni
f
o
r
m
r
e
s
our
c
e
l
oc
a
t
or
(
U
RL
)
of
e
m
a
i
l
a
ddr
e
s
s
e
s
w
i
t
h t
he
dom
a
i
n
hi
ghl
i
ght
e
d
f
or
e
a
s
y r
e
c
ogni
t
i
on o
f
phi
s
hi
ng e
m
a
i
l
s
.
T
he
t
oo
l
w
a
s
e
va
l
ua
t
e
d i
n a
n e
m
a
i
l
e
nvi
r
onm
e
nt
s
i
nc
e
t
hi
s
c
a
n be
a
da
pt
e
d t
o s
ui
t
ot
he
r
m
e
s
s
a
gi
ng e
nvi
r
onm
e
nt
s
.
T
he
e
f
f
i
c
i
e
nc
y
of
T
o
r
pe
do w
a
s
t
e
s
t
e
d a
ga
i
ns
t
t
he
s
t
a
t
us
quo
s
t
a
t
us
ba
r
i
n T
hunde
r
bi
r
d a
nd
r
e
s
ul
t
s
s
how
ed
t
ha
t
T
or
pe
do de
t
e
c
t
e
d f
r
a
udul
e
nt
e
m
a
i
l
s
85.
17%
m
o
r
e
t
i
m
e
s
t
ha
n t
he
s
t
a
t
us
ba
r
U
R
L
w
hi
c
h ha
d
43
.
31%
.
M
o
g
h
imi
an
d
V
ar
j
an
i
[
2
3
]
pr
opos
e
d t
w
o
f
e
a
t
u
r
e
s
s
e
t
s
t
o
i
m
pr
ove
phi
s
h de
t
e
c
t
i
on i
n i
nt
e
r
ne
t
ba
nki
ng.
T
he
s
uppor
t
ve
c
t
or
m
a
c
hi
ne
(
S
V
M
)
w
a
s
us
e
d t
o c
l
a
s
s
i
f
y w
e
bpa
ge
s
w
i
t
h a
f
e
a
t
ur
e
ve
c
t
or
c
ons
i
s
t
i
ng of
17
f
e
a
t
ur
e
s
:
9 r
e
l
e
va
nt
f
e
a
t
ur
e
s
a
nd 8 s
ugge
s
t
e
d f
e
a
t
ur
e
s
.
T
he
r
e
s
ul
t
s
of
t
he
i
r
e
x
p
e
r
ime
n
t in
d
ic
a
te
d
a
t
r
ue
pos
i
t
i
ve
va
l
ue
o
f
99
.
14
%
a
nd
f
a
l
s
e
ne
ga
t
i
ve
,
0.
86
%
.
S
a
hi
ngoz
e
t
a
l.
[2
4
]
de
s
i
gne
d a
n
a
nt
i
-
phi
s
hi
ng s
ys
t
e
m
ba
s
e
d on m
a
c
hi
ne
l
e
a
r
n
i
ng t
ha
t
c
om
bi
ne
d
s
ev
en
cl
as
s
i
f
i
cat
i
o
n
al
g
o
r
i
t
h
m
s
w
i
t
h
n
at
u
r
al
l
an
g
u
ag
e p
r
o
ces
s
i
n
g
f
eat
u
r
es
.
A
d
at
as
e
t
of
73
,
575
U
R
L
s
,
c
ons
i
s
t
i
ng of
36400 or
i
gi
na
l
l
y c
or
r
e
c
t
U
R
L
s
a
nd 37175 phi
s
hi
ng U
R
L
s
w
er
e
co
n
s
t
r
u
ct
ed
t
o
ev
al
u
at
e t
h
e
pe
r
f
or
m
a
nc
e
of
t
he
s
ys
t
e
m
.
R
e
s
ul
t
s
r
e
ve
a
l
e
d t
ha
t
t
he
pe
r
f
o
r
m
a
nc
e
of
t
he
pr
opos
e
d s
ys
t
e
m
w
a
s
i
nc
r
e
a
s
e
d by
2.
24%
a
nd 13
.
14%
f
or
n
at
u
r
al
l
a
ngua
ge
pr
oc
e
s
s
i
ng
(
N
LP
)
b
as
ed
f
eat
u
r
es
an
d
w
o
r
d
v
ect
o
r
s
r
es
p
ect
i
v
el
y
.
A
l
s
o,
t
he
R
a
ndom
f
or
e
s
t
a
l
gor
i
t
h
m
w
i
t
h
N
L
P
f
e
a
t
ur
e
s
pr
oduc
e
d t
he
hi
ghe
s
t
a
c
c
ur
a
c
y r
a
t
e
of
97
.
98
%
w
he
n
c
om
pa
r
e
d t
o t
he
s
e
ve
n ot
he
r
a
l
go
r
i
t
hm
s
,
N
a
i
ve
B
a
ye
s
a
l
gor
i
t
hm
,
k
-
ne
a
r
e
s
t
ne
i
ghbor
(n
=
3)
,
A
da
boos
t
,
s
e
q
u
e
n
tia
l min
ima
l o
p
ti
miz
a
tio
n
, K
-
st
a
r
,
a
nd
de
c
i
s
i
on t
r
e
e
s
.
T
he
us
e
of
pa
r
a
l
l
e
l
pr
oc
e
s
s
i
ng a
nd de
e
p
l
e
a
r
ni
ng
wa
s
s
u
g
g
es
t
ed
f
o
r
f
u
t
u
r
e r
es
ear
ch
.
Wi
t
h
t
h
e u
s
e
o
f
r
ei
n
f
o
r
cem
en
t
l
ear
n
i
n
g
,
S
m
a
d
i
et
a
l
.
[2
5
]
d
ev
el
o
p
ed
a
nove
l
a
ppr
oa
c
h i
n t
he
de
t
e
c
t
i
on of
phi
s
hi
ng a
t
t
a
c
ks
a
ga
i
ns
t
onl
i
ne
s
ys
t
e
m
s
.
T
he
i
r
pr
opos
e
d s
ys
t
e
m
c
a
l
l
e
d
p
h
is
h
in
g
e
ma
il d
e
te
c
tio
n
s
y
s
te
m
w
hi
c
h us
e
d
a
f
e
a
t
ur
e
e
va
l
ua
t
i
on a
nd r
e
duc
t
i
on a
l
gor
i
t
hm
a
dj
us
t
ed
r
e
gul
a
r
l
y
t
o r
e
f
l
e
c
t
c
ha
nge
s
i
n t
he
e
nvi
r
on
m
e
nt
th
a
t
is
,
e
xpl
or
e
d
ne
w
be
ha
vi
or
s
i
n
a n
ew
d
at
as
et
.
F
o
r
cl
as
s
i
f
i
cat
i
o
n
,
a
ne
ur
a
l
ne
t
w
or
k w
a
s
c
om
bi
ne
d w
i
t
h
r
e
i
nf
o
r
c
e
m
e
nt
l
e
a
r
ni
ng i
n
t
he
de
s
i
gne
d s
ys
t
e
m
a
nd us
e
d 50 f
e
a
t
ur
e
s
.
A
da
t
a
s
e
t
c
ont
a
i
ni
ng 7315 phi
s
hi
ng
e
m
a
i
l
s
,
4951 ha
m
e
m
a
i
l
s
f
o
r
t
r
a
i
ni
ng
a
nd 26722
phi
s
hi
ng U
R
L
’
s
w
a
s
us
e
d.
44%
of
t
h
e em
ai
l
s
we
r
e
us
e
d a
s
t
he
t
r
a
i
ni
ng
da
t
a
s
e
t
.
R
e
s
ul
t
s
of
t
he
pe
r
f
o
r
m
a
nc
e
e
va
l
ua
t
i
on
ove
r
50
i
nde
pe
nde
nt
r
uns
s
how
e
d a
n a
c
c
ur
a
c
y of
98.
63%
w
i
t
h a
t
r
ue
pos
i
t
i
ve
a
nd t
r
ue
ne
ga
t
i
ve
r
a
t
e
of
99.
0
7%
a
nd
1.
81%
.
A
cas
e
-
ba
s
e
d r
e
a
s
oni
ng phi
s
hi
ng de
t
e
c
t
i
on s
y
s
t
e
m
de
ve
l
ope
d by
A
but
a
i
r
a
nd B
e
l
ghi
t
h
[2
6
]
c
om
bi
ne
d bot
h onl
i
ne
a
nd of
f
l
i
ne
de
t
e
c
t
i
on o
f
phi
s
hi
ng a
t
t
a
c
ks
.
T
he
pr
opos
e
d s
ys
t
e
m
w
hi
c
h us
e
d
a r
el
at
i
v
el
y
s
m
al
l
d
at
as
et
(
5
7
2
cas
es
)
wa
s
ve
r
y a
da
pt
i
ve
a
nd
ab
l
e t
o
p
r
ed
i
ct
a zer
o
-
hour
phi
s
hi
ng a
t
t
a
c
k
e
a
s
i
l
y.
T
he
r
e
s
u
l
t
s
how
e
d t
he
pr
opos
e
d s
y
s
t
e
m
pr
oduc
e
d a
n a
c
c
ur
a
c
y of
95.
62%
.
H
a
di
e
t
a
l.
[2
7
]
e
xpe
r
i
m
e
nt
e
d o
n 11,
055
phi
s
hi
ng w
e
bs
i
t
e
s
us
i
ng a
W
E
K
A
s
of
t
w
a
r
e
e
nvi
r
onm
e
nt
.
T
he
y
pr
opos
e
d
a f
a
s
t
-
as
s
o
ci
at
i
v
e cl
as
s
i
f
i
cat
i
o
n
a
l
gor
i
t
hm
(
F
A
C
A
)
f
or
i
de
nt
i
f
yi
ng phi
s
hi
ng w
e
bs
i
t
e
s.
T
h
e
pr
opos
e
d
a
l
gor
i
t
hm
out
pe
r
f
or
m
ed
ot
he
r
as
s
o
ci
at
i
v
e
c
la
s
s
if
ic
a
tio
n
al
g
o
r
i
t
h
m
s
i
n
cl
as
s
i
f
i
cat
i
o
n
accu
r
acy
an
d
F
1
ev
al
u
at
i
o
n
m
eas
u
r
es
.
Z
ha
ng
e
t a
l.
[2
8
]
d
ev
el
o
p
ed
a
m
odi
f
i
e
d de
e
p ne
ur
a
l
ne
t
w
or
k m
ode
l
i
n
va
t
i
c
i
na
t
i
n
g phi
s
hi
ng a
t
t
a
c
ks
.
T
he
hybr
i
d de
e
p ne
u
r
a
l
ne
t
w
or
k m
ode
l
c
om
bi
ne
s
a
n a
ut
oe
nc
ode
r
w
i
t
h a
c
onvol
ut
i
ona
l
ne
ur
a
l
ne
t
w
or
k t
o be
a
bl
e
t
o
de
t
e
c
t
phi
s
h a
t
t
a
c
ks
i
n r
e
c
e
i
va
bl
e
time
.
T
h
e
mo
d
e
l
wa
s
c
om
pa
r
e
d w
i
t
h t
he
S
V
M
,
d
e
c
i
s
i
on t
r
e
e
,
a
nd L
i
ne
a
r
S
V
C
a
l
gor
i
t
hm
on a
de
e
p
l
e
a
r
ni
ng
pl
a
t
f
or
m
,
T
e
ns
or
f
l
ow
.
T
he
r
e
s
u
lts
in
d
ic
a
te
d
t
h
at
t
h
e d
ev
el
o
p
ed
ar
ch
i
t
ect
u
r
e h
a
d
a
n a
c
c
ur
a
c
y of
97.
87%
,
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SSN
:
1693
-
6930
T
E
L
KOM
NI
KA
T
el
eco
m
m
u
n
C
om
put
E
l
C
ont
r
o
l
,
Vo
l
.
1
9
, N
o
.
5
,
O
ct
o
b
er
2021
:
17
07
-
17
14
1710
pr
e
c
i
s
i
on of
98.
69
,
a
nd a
r
e
c
a
l
l
of
97.
20
%
w
hi
c
h
wa
s
t
he
be
s
t
of
t
he
f
ou
r
m
ode
l
s
co
n
s
i
d
er
ed
.
S
imil
a
r
ly
,
[
2
9
]
de
pl
oye
d t
he
us
e
of
t
w
o
c
l
a
s
s
i
f
i
e
r
s
,
S
V
M
a
nd
de
c
i
s
i
on t
r
e
e
,
t
o
de
ve
l
op
a
l
ear
n
i
n
g
-
ba
s
e
d
a
ggr
e
ga
t
i
on a
na
l
ys
i
s
s
ys
t
e
m
i
n de
t
e
c
t
i
ng phi
s
hi
ng a
t
t
a
c
ks
on w
e
b p
a
ge
s
.
T
he
r
e
s
ul
t
s
s
how
e
d t
ha
t
t
hi
s
s
ys
t
e
m
e
nha
nc
e
d
t
h
e
pe
r
f
or
m
a
nc
e
of
e
xi
s
t
i
ng a
nt
i
-
phi
s
hi
ng m
e
t
hods
.
T
o s
a
f
e
gua
r
d s
e
ns
i
t
i
ve
i
nf
or
m
a
t
i
on of
us
e
r
s
,
i
t
i
s
c
r
uc
i
a
l
t
ha
t
an
ad
eq
u
at
e m
ean
s
o
f
i
de
nt
i
f
yi
ng a
nd a
ppr
e
he
ndi
ng
phi
s
hi
ng e
m
a
i
l
s
be
de
ve
l
ope
d.
I
n t
hi
s
s
t
udy,
t
he
m
a
xi
m
um
e
nt
r
opy
(
Ma
x
-
E
nt
)
c
l
a
s
s
i
f
i
c
a
t
i
on m
e
t
hod us
i
ng pa
r
s
i
m
oni
ous
,
bu
t
opt
i
m
a
l
f
e
a
t
u
r
e
s
w
as
imp
le
me
n
te
d
.
2.
R
ES
EA
R
C
H
M
ETH
O
D
I
n t
hi
s
s
e
c
t
i
on,
t
he
m
e
t
hodol
ogy us
e
d
i
nc
l
udi
ng
d
at
a
c
o
lle
c
tio
n
a
n
d
th
e
c
la
s
s
if
ic
a
tio
n
ta
s
k
is
de
s
c
r
i
be
d.
T
hi
s
i
nc
l
ude
s
t
he
da
t
a
c
ol
l
e
c
t
i
on a
nd pr
e
pa
r
a
t
i
on pr
oc
e
s
s
.
T
he
ma
x
imu
m e
n
tr
o
p
y
(M
E
)
m
ode
l
w
as
eq
u
al
l
y
d
ep
i
ct
ed
u
s
i
n
g
m
at
h
em
at
i
cal
m
o
d
el
s
.
2
.1
.
B
u
ild
in
g
co
rp
o
ra
w
i
t
h
p
ar
s
i
m
on
i
ou
s
f
e
at
u
r
e
s
T
he
da
t
a
s
e
t
us
e
d f
or
t
he
s
t
udy w
a
s
f
r
om
publ
i
c
l
y
a
va
i
l
a
bl
e
r
e
pos
i
t
or
i
e
s
by [
30]
(
f
or
phi
s
hi
ng e
-
ma
il
da
t
a
s
e
t
)
a
nd [
31]
(
f
or
ha
m
e
-
ma
il d
a
ta
s
e
t)
.
I
n
to
t
a
l,
w
e
w
o
r
k
e
d
w
ith
8
2
6
6
e
-
m
a
i
l
c
or
por
a
w
i
t
h 47
f
e
a
t
ur
e
s
w
h
i
ch
ar
e
c
om
m
onl
y us
e
d i
n l
i
t
e
r
a
t
ur
e
[
32
]
-
[
34]
.
O
f
t
he
8266 c
or
por
a
,
w
e
de
s
i
gna
t
e
d 6266 e
-
ma
ils
a
s
o
u
r
t
r
a
i
ni
ng da
t
a
,
l
e
a
vi
ng
us
w
i
t
h
2000
t
e
s
t
da
t
a
(
100
0 ha
m
s
a
nd 1000
phi
s
he
s
)
.
F
ur
t
he
r
m
or
e
,
w
e
c
a
r
r
i
e
d out
a
di
m
e
ns
i
ona
l
i
t
y r
e
duc
t
i
on o
f
t
he
f
e
a
t
ur
e
s
e
t
f
r
om
47 t
o
2
7 us
i
ng
r
e
gr
e
s
s
i
on.
T
hi
s
w
a
s
ba
s
e
d on
t
he
t
hi
nki
ng
th
a
t it is
p
o
s
s
ib
le
to
g
e
t th
e
p
a
r
s
imo
n
io
u
s
f
e
w
‘
p
r
in
c
ip
a
l’
f
e
a
tu
r
e
s
,
th
u
s
e
limin
a
tin
g
r
e
d
u
n
d
a
n
t f
e
a
tu
r
e
s
w
ith
o
u
t
m
uc
h i
nf
or
m
a
t
i
on
l
os
s
.
2
.
2
.
M
axi
m
u
m
en
t
ro
p
y
T
h
e
m
a
xi
m
um
e
nt
r
opy
(M
E
)
i
s a
pr
oba
bi
l
i
s
t
i
c
m
ode
l
,
ba
s
e
d on t
he
‘
pr
i
nc
i
pl
e
s
of
m
a
xi
m
um
e
nt
r
opy’
.
M
a
x
imu
m
e
nt
r
opy
h
as
a w
el
l
-
e
s
ta
b
lis
h
e
d
h
is
to
r
y
in
e
f
f
ic
ie
n
tly
s
o
lv
in
g
th
e
te
x
t c
la
s
s
if
ie
r
p
r
o
b
le
m.
A
ddi
t
i
ona
l
l
y,
m
a
xi
m
um
e
nt
r
opy
i
s
ad
ap
t
ab
l
e t
o
a l
ar
g
e f
eat
u
r
e s
et
an
d
i
t
s
p
er
f
o
r
m
an
ce i
s
n
o
t
af
f
ect
ed
b
y
t
h
e
f
eat
u
r
e
s
e
le
c
tio
n
me
th
o
d
[3
5
].
M
a
xi
m
um
e
nt
r
opy
de
t
e
r
m
i
ne
s
pr
oba
bi
l
i
t
i
e
s
ba
s
e
d on
t
he
pr
i
nc
i
pl
e
of
m
a
ki
ng
min
ima
l a
s
s
u
mp
tio
n
s
a
s
f
o
llo
w
:
S
uppos
e
t
ha
t
w
e
ha
ve
a
s
e
t
of
f
e
a
t
u
r
e
s
,
a
s
e
t
o
f
f
u
nc
t
i
ons
1
,
…
,
(
b
y
w
h
ic
h
w
e
ma
y
d
e
te
r
min
e
th
e
c
ont
r
i
but
i
on of
e
a
c
h f
e
a
t
ur
e
t
o t
he
m
ode
l
)
a
nd a
s
e
t
of
c
ondi
t
i
ons
;
w
e
de
t
e
r
m
i
ne
t
he
pr
oba
bi
l
i
t
y di
s
t
r
i
but
i
on
th
a
t s
a
tis
f
ie
s
th
e
g
iv
e
n
c
o
n
d
itio
n
s
a
n
d
min
imiz
e
s
th
e
r
e
la
tiv
e
e
n
tr
o
p
y
(
d
i
v
er
g
en
ce
of
K
ul
l
ba
c
k
-
L
ei
b
l
er
)
(
|
|
0
)
,
w
ith
r
e
s
p
e
c
t to
th
e
d
is
tr
ib
u
t
i
on
0
.
T
he
c
ondi
t
i
ona
l
ma
x
imu
m e
n
tr
o
p
y
mo
d
e
l is
a
n
e
x
p
o
n
e
n
tia
l w
ith
th
e
f
o
r
m:
(
|
)
=
1
(
)
∏
(
,
)
=
1
w
h
er
e
(
|
)
de
not
e
s
t
he
pr
oba
bi
l
i
t
y
of
oc
c
ur
r
e
nc
e
of
ou
t
c
om
e
,
gi
ve
n c
ont
e
xt
w
ith
c
o
n
s
tr
a
in
t o
r
f
e
a
tu
r
e
f
unc
t
i
ons
(
|
)
.
M
E
m
ode
l
r
e
p
r
e
s
e
nt
s
e
vi
de
nc
e
w
i
t
h bi
na
r
y f
unc
t
i
o
ns
know
n a
s
c
ont
e
xt
ua
l
pr
e
di
c
a
t
e
s
i
n t
he
f
or
m
:
,
′
(
,
)
=
1
=
′
(
)
=
0
ℎ
i
s
t
he
c
ont
e
xt
ua
l
p
r
e
di
c
a
t
e
w
hi
c
h m
a
ps
a
pa
i
r
of
o
ut
c
om
e
o a
nd c
ont
e
xt
h
t
o
{t
r
ue
;
f
a
l
s
e
} [
35]
.
3.
R
ES
U
LTS
A
ND ANAL
YS
I
S
I
n t
hi
s
s
e
c
t
i
on,
w
e
r
e
po
r
t
a
nd e
va
l
ua
t
e
t
he
r
e
s
ul
t
s
of
t
he
m
a
xi
m
um
e
nt
r
opy
cl
as
s
i
f
i
cat
i
o
n
t
ech
n
i
q
u
es
vi
s
-
à
-
v
i
s
t
he
N
ai
v
e B
ay
es
(
B
as
el
i
n
e)
an
d
s
uppor
t
ve
c
t
or
m
a
c
hi
ne
(
S
V
M
)
.
We d
es
cr
i
b
e s
o
m
e b
en
ch
m
ar
k
m
e
t
r
i
c
s
us
e
d f
or
t
he
e
va
l
ua
t
i
on.
W
e
r
e
por
t
t
he
pe
r
f
or
m
a
nc
e
s
of
t
he
t
e
c
hni
que
s
us
i
ng t
a
bl
e
s
a
nd c
ha
r
t
s
.
3.
1.
P
erf
o
r
m
a
n
ce
m
et
ri
c
T
he
pe
r
f
or
m
a
nc
e
m
e
t
r
i
c
s
us
e
d t
o e
va
l
ua
t
e
our
w
or
k
w
er
e accu
r
acy
,
p
r
eci
s
i
o
n
,
r
ecal
l
,
an
d
er
r
o
r
r
at
e.
T
h
i
s
w
as
cal
cu
l
at
ed
b
as
ed
o
n
t
h
e co
r
r
ect
n
es
s
o
r
o
t
h
er
w
i
s
e o
f
t
h
e cl
a
s
s
i
f
i
ed
t
es
t
d
at
a d
ep
i
ct
ed
b
y
t
r
ue
pos
i
t
i
ve
,
t
r
ue
ne
ga
t
i
ve
,
f
a
l
s
e
pos
i
t
i
ve
,
a
nd
f
a
l
s
e
ne
ga
t
i
ve
.
T
r
ue
pos
i
t
i
ve
i
s
t
he
c
or
r
e
c
t
l
y
c
l
a
s
s
i
f
i
e
d
p
h
i
s
h
,
t
r
u
e n
eg
at
i
v
e i
s
t
h
e co
r
r
ect
l
y
cl
as
s
i
f
i
ed
h
am
,
f
al
s
e
-
pos
i
t
i
ve
de
pi
c
t
s
phi
s
h
e
s
w
r
ongl
y c
l
a
s
s
i
f
i
e
d a
s
ha
m
w
hi
l
e
f
a
l
s
e
ne
ga
t
i
ve
de
pi
c
t
s
ha
m
s
w
r
ongl
y c
l
a
s
s
i
f
i
e
d a
s
phi
s
he
s
.
I
n
t
he
c
ont
e
xt
o
f
our
s
t
udy,
w
e
de
f
i
ne
t
he
m
a
s
f
ol
l
ow
s
:
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
KA
T
el
eco
m
m
u
n
C
om
put
E
l
C
ont
r
o
l
A
m
ax
i
m
um
e
nt
r
opy
c
l
as
s
i
f
i
c
at
i
on s
c
he
m
e
f
or
phi
s
hi
ng de
t
e
c
t
i
on us
i
ng
…
(
E
m
m
anue
l
O
.
A
s
ani
)
1711
=
=
=
=
We d
ep
i
ct
t
h
e c
onf
us
i
on m
a
t
r
i
x a
nd p
r
e
s
e
nt
a
t
i
on of
r
e
s
ul
t
s
i
n t
a
bul
a
r
f
or
m
a
s
i
n
T
ab
l
e
1.
T
a
bl
e
1.
C
onf
us
i
on
ma
tr
ix
o
f
th
e
c
la
s
s
if
ic
a
tio
n
(
=
,
=
,
=
)
P
hi
s
he
s
Ha
m
P
hi
s
he
s
996
4
968
31
972
27
165
835
3
998
5
996
A
ccu
r
acy
:
T
h
i
s
i
s
t
h
e
p
er
cen
t
ag
e o
f
co
r
r
ect
l
y
cl
as
s
i
f
i
ed
ma
ils
(
h
a
m
s a
s w
e
l
l
a
s p
h
i
sh
e
s)
.
T
h
is
is
gi
ve
n a
s
:
=
+
P
r
e
c
is
io
n
: T
h
is
is
th
e
to
ta
l n
u
mb
e
r
o
f
tr
u
e
pos
i
t
i
ve
s
di
vi
de
d by
t
he
t
o
t
a
l
num
be
r
o
f
e
ma
ils
id
e
n
tif
ie
d
a
s
ha
m
s
.
T
hi
s
i
s
gi
ve
n a
s
;
=
+
R
ecal
l
:
T
h
i
s
i
s
t
h
e p
er
cen
t
ag
e o
f
co
r
r
ect
l
y
cl
as
s
i
f
i
ed
p
h
i
s
h
es
.
T
h
i
s
i
s
g
i
v
en
as
;
=
+
E
rro
r ra
t
e
,
th
is
is
g
i
v
en
as
;
=
1
−
3.
2
.
P
erf
o
r
m
a
n
ce m
ea
s
u
re a
n
d
d
i
s
cu
s
s
i
o
n
We m
eas
u
r
ed
t
h
e p
er
f
o
r
m
an
ce o
f
o
u
r
w
o
r
k
w
i
t
h
r
ed
u
ced
f
eat
u
r
e
s
p
ace o
f
2
7
,
r
el
at
i
v
e t
o
N
aï
v
e
B
ay
es
(
b
as
el
i
n
e)
an
d
S
V
M
w
h
i
ch
h
as
4
7
f
eat
u
r
es
.
We p
r
es
en
t
t
h
e
c
onf
us
i
on
m
a
t
r
i
x
i
n
T
a
bl
e
1
a
nd t
he
p
er
f
o
r
m
an
ce m
eas
u
r
e
o
f
t
h
e
3
-
cl
as
s
i
f
i
cat
i
o
n
s
ch
em
e i
n
T
ab
l
e
2
.
W
e
pr
e
s
e
nt
a
pl
ot
of
t
r
ue
va
l
ue
a
g
a
i
ns
t
t
he
va
l
ue
s
pr
e
di
c
t
e
d by our
c
l
a
s
s
i
f
i
c
a
t
i
on s
c
he
m
e
w
hi
c
h i
s
de
pi
c
t
e
d i
n t
he
c
onf
us
i
on m
a
t
r
i
x
i
n F
i
gur
e
3.
T
h
e
p
e
rfo
rm
a
nc
e
m
e
t
r
i
c
s
,
t
ha
t
i
s
t
he
num
be
r
of
a
t
r
ue
pos
i
t
i
ve
,
t
r
ue
ne
ga
t
i
ve
,
f
a
l
s
e
pos
i
t
i
ve
a
nd f
a
l
s
e
ne
ga
t
i
ve
i
s
pr
e
s
e
nt
e
d gr
a
phi
c
a
l
l
y
i
n
F
i
g
ur
e
4
.
T
h
i
s
wa
s
us
e
d t
o
e
va
l
ua
t
e
t
he
pe
r
f
or
m
a
nc
e
of
t
he
c
l
a
s
s
i
f
i
c
a
t
i
on m
ode
l
s
w
hi
c
h i
s
pr
e
s
e
nt
e
d gr
a
phi
c
a
l
l
y i
n
F
i
g
ur
e
5
.
T
ab
l
e 2
.
P
er
f
o
r
m
an
ce m
eas
u
r
e o
f
3
cl
as
s
i
f
i
cat
i
o
n
s
ch
em
es
P
ar
am
et
er
M
a
xi
m
um
E
nt
r
opy
SV
M
N
aï
v
e B
ay
e
s
T
r
ue
pos
i
t
i
ve
(
)
996
968
972
T
r
ue
ne
ga
t
i
ve
(
)
835
998
996
F
a
ls
e
p
o
s
itiv
e
(
)
165
3
5
F
al
s
e n
eg
at
i
v
e (
)
4
31
27
A
ccu
r
acy
0.9155
0.983
0.984
P
r
e
c
is
io
n
0.99523242
0.969873664
0.973607038
R
ecal
l
0.835
0.997002997
0.995004995
E
rro
r ra
t
e
0.0845
0.017
0.016
N
o of
F
e
a
t
ur
e
s
27 (
di
m
e
ns
i
ona
l
i
t
y r
e
duc
t
i
on f
r
om
47 t
o 27 t
hr
ough
r
e
gr
e
s
s
i
on)
47
47
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SSN
:
1693
-
6930
T
E
L
KOM
NI
KA
T
el
eco
m
m
u
n
C
om
put
E
l
C
ont
r
o
l
,
Vo
l
.
1
9
, N
o
.
5
,
O
ct
o
b
er
2021
:
17
07
-
17
14
1712
F
ig
ur
e
3
.
P
l
ot
of
t
r
ue
va
l
ue
a
ga
i
ns
t
pr
e
di
c
t
e
d va
l
ue
F
ig
ur
e
4
.
P
er
f
o
r
m
an
ce
me
tr
ic
F
ig
ur
e
5
.
P
er
f
o
r
m
an
ce ev
al
u
at
i
o
n
o
f
m
a
xi
m
um
e
nt
r
opy
,
S
V
M
an
d
N
aï
v
e B
ay
es
0
200
400
600
800
1000
1200
1400
1600
1800
2000
P
hi
s
h
Ha
m
ot
o
ue
a
ues
aga
s
t
ed
c
t
ed
a
ues
M
ai
l
i
ndi
c
es
M
ai
l
C
l
as
s
es
T
r
ue v
al
ues
P
r
edi
c
t
ed v
al
ues
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
KA
T
el
eco
m
m
u
n
C
om
put
E
l
C
ont
r
o
l
A
m
ax
i
m
um
e
nt
r
opy
c
l
as
s
i
f
i
c
at
i
on s
c
he
m
e
f
or
phi
s
hi
ng de
t
e
c
t
i
on us
i
ng
…
(
E
m
m
anue
l
O
.
A
s
ani
)
1713
4.
CO
NCL
US
I
O
N
T
h
e
m
a
xi
m
um
e
nt
r
opy
w
i
t
h pa
r
s
i
m
oni
ous
f
e
a
t
u
r
e
s
pa
c
e
c
l
e
a
r
l
y out
pe
r
f
or
m
e
d bot
h t
he
N
aï
v
e B
ay
es
a
nd S
V
M
i
n t
e
r
m
s
of
pr
e
c
i
s
i
on.
N
a
ï
ve
B
a
ye
s
,
how
e
ve
r
,
ha
d
t
h
e h
i
g
h
es
t
accu
r
acy
r
at
e i
n
o
u
r
ex
p
er
i
m
en
t
.
F
ut
ur
e
w
o
r
k
m
a
y
i
nc
l
ude
hyb
r
i
di
z
i
ng
m
a
xi
m
um
e
nt
r
opy
w
hi
c
h
ha
s
t
he
hi
ghe
s
t
pr
e
c
i
s
i
on a
nd
N
a
v
e
B
a
ye
s
w
h
i
ch
h
as
t
h
e h
i
g
h
es
t
accu
r
acy
.
We n
o
t
e t
h
a
t
m
a
xi
m
um
e
nt
r
opy
i
s
g
en
er
at
i
v
e w
h
i
l
e N
aï
v
e
B
ay
es
i
s
d
is
c
r
imin
a
tiv
e
; it w
il
l b
e
in
te
r
e
s
tin
g
to
s
e
e
th
e
r
e
s
u
lt o
f
th
e
h
y
b
r
id
o
f
b
o
th
.
A
C
K
N
O
W
LED
G
EM
EN
TS
W
e
gr
a
t
e
f
ul
l
y a
c
know
l
e
dge
t
he
s
uppor
t
of
t
he
I
n
f
o
r
ma
tio
n
S
y
s
te
ms
M
o
d
e
llin
g
,
S
imu
la
tio
n
&
S
ecu
r
i
t
y
R
es
ear
c
h G
r
oup (
i
S
yM
S
)
-
w
w
w
.
i
s
ym
s
.
t
e
c
h a
nd t
he
S
oc
i
e
t
y f
or
M
ul
t
i
di
s
c
i
pl
i
na
r
y &
A
dva
nc
e
d
R
es
ear
ch
T
ech
n
i
q
u
es
(
S
M
A
R
T
)
f
o
r
acces
s
t
o
r
es
ear
ch
r
ep
o
s
i
t
o
r
i
es
,
l
i
t
er
a
r
y
m
at
er
i
al
s
an
d
u
s
ef
u
l
i
n
s
i
g
h
t
s
f
r
o
m
pa
r
t
ne
r
s
.
W
e
e
qua
l
l
y
a
ppr
e
c
i
a
t
e
t
he
f
undi
ng p
r
o
vi
de
d by t
he
L
an
d
m
ar
k
U
n
i
v
er
s
i
t
y
C
en
t
er
f
o
r
R
es
ear
c
h
I
nnova
t
i
on a
nd D
e
ve
l
opm
e
nt
(
L
U
C
R
I
D
)
.
R
EF
ER
EN
C
ES
[1
]
E.
O.
Asa
ni,
A.
Om
o
to
sh
o,
O.
B
.
L
onge
,
J.
O.
Om
on
i
gho a
nd B
.
G
ba
dm
o
si,
“
A R
e
a
l
-
ti
m
e
Ge
s
tu
r
e
En
gi
n
e
e
r
e
d
C
AP
TC
H
A,
”
I
n
te
r
na
ti
on
al J
o
ur
na
l o
f Me
c
h
an
ic
a
l E
ng
i
ne
e
r
in
g an
d T
e
c
hn
ol
og
y
vo
l
9,
n
o
12,
p
p.
61
8
-
62
9
,
20
18.
[2
]
S
.
S
.
Na
ir
,
M
.
M
.
S
he
r
in a
nd T.
S
a
n
tha
,
"
De
d
up
lic
a
ti
on
Ena
ble
d S
e
c
ur
e
E
-
m
a
il
S
e
r
ve
r
o
n
C
l
ou
d
En
vir
onm
e
nt
u
s
in
g
Vir
tua
l Da
ta
O
pt
im
iz
e
r
,
"
202
0 6t
h I
nt
e
rn
at
io
na
l C
o
nfe
r
e
nc
e
on A
dv
a
nc
e
d C
om
pu
ti
ng a
nd C
om
mu
nic
at
io
n Sy
ste
m
s
(ICA
CCS
)
,
20
20
,
pp.
2
70
-
27
5,
do
i: 1
0.
1
10
9/I
C
AC
C
S
48
705.
20
20.
9
07
44
63.
[3
]
J.
Ho
ng,
“
The
s
ta
te
of
p
hi
sh
in
g a
tta
c
ks,
”
C
om
mu
n
ic
a
ti
on
s o
f t
he
AC
M,
vo
l 5
5,
n
o 1,
pp.
74
-
81,
20
12,
doi
:
1
0.
11
45
/2
06
31
76.
20
63
19
7
.
[4
]
S
.
A.
R
obi
la
a
n
d J.
W
.
R
a
guc
c
i,
"
Do
n’
t
be
a
P
h
is
h:
S
te
p
s
in
Use
r
Ed
uc
a
t
io
n,
”
AC
M SI
GC
SE Bu
lle
ti
n
,
v
ol
.
38
,
no
.
3,
pp 23
7
-
2
41,
S
e
p
te
m
be
r
20
06,
do
i:
10.
11
45
/1
14
01
23.
1
14
01
87
.
[5
]
E.
O.
Asa
ni a
nd
A.
A.
A
de
g
un “
M
a
xim
um
P
hi
sh B
a
it
: To
wa
r
d
s F
e
a
t
ur
e
B
a
se
d De
te
c
t
io
n of
P
hi
sh
in
g u
s
in
g
M
a
xim
um
E
ntr
op
y C
la
ss
if
ic
a
ti
on
Te
c
h
ni
que
,
”
I
n p
roc
e
e
din
gs
o
f I
n
te
r
na
ti
on
al C
on
fe
re
nc
e
o
n Sc
ie
nc
e
,
T
e
c
hn
ol
o
gy
,
Ed
uc
a
ti
on,
A
rt
s,
Ma
na
ge
me
nt a
nd
Soc
ia
l Sc
ie
nc
e
s
iST
E
AM
S
,
20
14.
[6
]
G.
Xia
ng,
“
To
wa
r
d a
P
h
is
h F
r
e
e
Wor
ld
: A F
e
a
t
ur
e
-
ty
pe
-
a
wa
r
e
C
a
sc
a
de
d L
e
a
r
n
in
g F
r
a
m
e
wor
k f
or
P
h
is
h
De
te
c
ti
on,
”
Doc
tor
a
l Di
sse
r
ta
t
io
n,
L
a
ngua
ge
Te
c
hn
ol
og
ie
s I
ns
ti
tu
te
S
c
ho
ol of
C
om
p
ute
r
S
c
ie
nc
e
C
a
r
ne
g
ie
M
e
ll
o
n
Uni
ve
r
si
ty,
P
it
ts
bur
gh,
20
13
.
[7
]
R
.
I
sla
m
a
nd
J.
Aba
wa
jy
,
“
A
m
ul
ti
-
t
ie
r
p
hi
sh
in
g de
t
e
c
ti
on a
n
d f
i
lte
r
in
g a
p
pr
oa
c
h,
”
J
ou
rn
al
of N
e
tw
o
rk
a
nd
C
omp
ut
e
r A
pp
lic
at
io
ns
,
vo
l
.
36
,
no
.
1,
p
p.
32
4
-
3
35,
2
01
3,
doi
:
10.
10
16
/j.
j
nc
a
.
2
01
2.
0
5.
00
9
.
[8
]
R.
W
.
L
uc
ky,
“
C
lic
k
ph
ob
ia
[
R
e
f
le
c
t
io
ns]
,
”
I
EEE Spe
c
tr
um
,
vo
l 48 no,
p.
25,
201
1
,
doi
:
1
0.
11
09
/M
S
P
EC
.
20
11.
5
67
63
77
.
[9
]
A.
K.
Ja
in a
nd B
.
B
.
G
upta
“
P
hi
sh
in
g De
te
c
ti
on
: Ana
l
ysi
s of
Vi
sua
l S
im
i
la
r
i
ty B
a
se
d Appr
oa
c
he
s
”
Se
c
uri
ty
an
d
C
omm
un
ic
at
io
n N
e
tw
o
rk
s,
vo
l
.
20
17
,
pp.
1
-
20,
20
17
,
do
i:
1
0.
11
55
/2
01
7/
54
21
04
6.
[
10]
M
.
Ala
ut
hm
a
n,
A.
A
lm
om
a
ni,
M
.
Alwe
sha
hm
,
W
.
Al
om
ou
sh a
nd K.
A
lie
ya
n
,
“
M
a
c
hi
n
e
Le
a
r
n
in
g
f
o
r
P
hi
shi
ng
De
te
c
ti
on a
n
d M
iti
ga
t
io
n
,
”
B.
B
.
G
up
ta
,
&
Q.
Z
.
S
h
e
ng,
(
Ed
s.
)
.
Mac
h
ine
L
e
a
rn
in
g fo
r C
omp
ut
e
r an
d C
y
be
r
Se
c
u
ri
ty
:
Pri
nc
ip
le
,
Alg
or
it
hm
s,
and P
rac
tic
e
s
(
1st e
d.
)
,
CRC P
res
s
,
p
p 26
-
47,
20
19
,
doi
:
1
0.
12
01
/9
78
04
29
50
40
44
.
[
11]
N.
M
a
le
ki
a
n
d
A.
A.
G
hor
ba
n
i,
“
G
e
ne
r
a
tin
g
P
hi
sh
in
g Em
a
i
ls
Us
in
g G
r
a
ph
Da
ta
ba
s
e
,
”
I
n: He
n
g
S
.
H.
,
L
ope
z
J.
(ed
s
)
I
nf
or
ma
ti
on
Se
c
ur
ity
Pr
ac
t
ic
e
an
d Ex
pe
r
ie
nc
e
.
I
S
PEC
20
19.
L
e
c
tu
re
N
o
te
s
i
n C
o
mp
ute
r
Sc
ie
nc
e
,
v
ol.
11
879
,
201
9.
S
pr
i
ng
e
r
,
C
ha
m
.
do
i: 1
0.
1
00
7/
97
8
-
3
-
0
30
-
34
33
9
-
2
_25
.
[
12]
A.
Ha
m
id,
I
.
R
a
hm
i a
nd
A.
Je
m
a
l,
“
P
r
of
i
li
ng P
h
is
hi
ng
Em
a
il B
a
se
d on C
l
us
te
r
i
ng A
ppr
oa
c
h
,
”
I
n p
roc
e
e
di
ng
s o
f
12t
h
I
E
EE I
nte
rn
at
io
na
l C
on
fe
re
nc
e
o
n T
ru
st,
S
e
c
u
ri
ty
and
Pr
iv
ac
y
i
n
C
o
mp
ut
in
g
a
nd
C
om
mu
ni
c
at
io
ns
,
”
20
13,
pp.
6
28
-
6
35
,
d
oi
: 1
0.
1
10
9/
Tr
u
stC
om
.
20
13.
76.
[
13]
A
W
P
G
,
“
P
his
hi
ng Ac
t
iv
it
y
Tr
e
nd
s
R
e
por
t,
”
2n
d Q
ua
r
te
r
201
8.
[
14]
P
his
hla
b
, “
2
01
8 P
hi
sh
in
g Tr
e
nd
s a
nd I
nte
ll
ige
nc
e
R
e
por
t:
Ha
c
k
in
g t
he
Hum
a
n
,
”
20
18
.
[
O
nl
ine
]
.
Ava
ila
ble
a
t
:
htt
ps
:/
/i
nf
o.
ph
is
hla
bs.
c
om
/
hu
bf
s
/2
01
8%
20P
TI
%
20R
e
po
r
t/P
h
is
hL
a
bs
%
20
Tr
e
n
d%
20R
e
por
t_
20
18
-
di
gi
ta
l.
p
df
.
Ac
c
e
s
se
d 2
2/
02
/2
01
9.
[
15]
I
.
Qa
ba
je
h,
F
.
Tha
b
ta
ha
a
n
d F
.
C
hic
la
na
,
“
A r
e
c
e
nt r
e
v
ie
w of
c
o
nve
nt
io
na
l v
s.
a
utom
a
te
d c
y
be
r
se
c
u
r
it
y
a
n
ti
-
ph
is
h
in
g te
c
h
ni
que
s
,”
C
om
pu
te
r Sc
ie
nc
e
Re
v
ie
w
,
vo
l.
29,
pp.
44
-
55
,
20
18
,
d
oi
:
1
0.
1
01
6/
j.
c
osr
e
v.
2
01
8.
05.
00
3
.
[
16]
A.
Ale
r
o
ud a
nd L
.
Z
hou
,
“
P
hish
in
g
e
nv
ir
o
nm
e
n
ts,
te
c
hni
qu
e
s,
a
nd
c
ou
nte
r
m
e
a
sur
e
s:
A s
ur
ve
y
,
”
C
omp
ute
rs an
d
Se
c
u
ri
ty
,
vo
l.
68,
pp
.
1
60
-
19
6
,
2
01
7,
do
i:
10
.
1
01
6/
j.
c
ose
.
201
7.
04.
00
6
.
[
17]
D.
G
oe
l a
nd
A.
K.
Ja
i
n,
“
M
o
bi
le
ph
is
hi
ng a
tta
c
ks
a
nd
de
f
e
nc
e
m
e
c
ha
ni
sm
s
: S
ta
te
of
a
r
t
a
n
d o
pe
n
r
e
se
a
r
c
h
c
h
a
l
l
e
n
g
e
s
,”
C
o
mp
ute
rs
a
nd Se
c
u
ri
ty
,
vo
l.
73,
pp
.
5
19
-
54
4
,
20
18,
d
oi
:
1
0.
10
16
/j.
c
ose
.
20
17.
1
2.
0
06
[
18]
K.
S
um
a
nt
hi a
nd R
.
Da
m
oda
r
a
m
,
“
S
ur
ve
y a
nd A
na
l
ys
is
on P
his
hi
ng De
te
c
ti
on Te
c
hn
iq
ue
s,
”
I
nt
e
rn
at
io
na
l
J
o
urn
al
of A
dv
a
nc
e
d
Re
se
arc
h in C
om
pu
te
r Sc
ie
nc
e
,
vo
l.
9
,
no.
1
,
201
8.
[
19]
K.
L
.
C
hie
w,
K.
S
.
C
.
Yo
ng a
nd
C
.
L
.
Ta
n,
“
A sur
v
e
y
of
ph
is
hi
ng
a
tta
c
ks
: the
ir
t
ype
s,
ve
c
tor
s
a
n
d te
c
h
ni
c
a
l
a
ppr
oa
c
he
s
,”
Ex
pe
rt Sy
ste
ms w
it
h Ap
pl
ic
a
ti
on
s
,
vo
l.
10
6
,
pp.
1
-
20
,
20
18
,
d
oi
: 10.
10
16
/j.
e
s
wa
.
2
01
8.
03.
05
0
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SSN
:
1693
-
6930
T
E
L
KOM
NI
KA
T
el
eco
m
m
u
n
C
om
put
E
l
C
ont
r
o
l
,
Vo
l
.
1
9
, N
o
.
5
,
O
ct
o
b
er
2021
:
17
07
-
17
14
1714
[
20]
A.
Alm
om
a
ni,
B
.
B
.
G
upta
,
S
.
Ata
wne
h,
A.
M
e
ule
n
be
r
g a
nd E.
A
lm
om
a
ni,
“
A s
ur
ve
y of
ph
is
hi
ng e
m
a
i
l f
i
lte
r
in
g
te
c
h
ni
que
s
,”
I
EE
E c
o
m
mu
nic
at
io
ns s
urv
e
y
s &
tut
or
ia
ls
,
v
ol.
15
,
no.
4
,
pp
.
207
0
-
20
90
,
2
01
3,
doi
: 1
0.
11
09
/S
UR
V.
2
01
3.
03
07
13.
00
02
0.
[
21]
G
.
S
onowa
l a
n
d K.
S
.
K
up
pu
sa
m
y,
“
P
hi
DM
A
-
A p
his
hi
ng d
e
te
c
ti
on m
ode
l w
it
h m
u
lt
i
-
f
i
lte
r
a
ppr
oa
c
h
,
”
J
o
ur
na
l of
K
ing Sa
ud
Un
iv
e
r
si
ty
-
C
o
mp
ute
r a
nd I
nf
or
ma
ti
o
n Sc
ie
nc
e
s
,
vo
l.
32,
no.
1,
pp.
99
-
11
2
,
201
7
,
doi
:
1
0.
10
16
/j.
jk
suc
i.
20
17.
07.
0
05
.
[
22]
M
.
Volka
m
e
r
,
K.
R
e
na
ud,
B
.
R
e
inhe
im
e
r
a
nd A.
K
u
nz
,
“
Use
r
e
xp
e
r
ie
nc
e
s of
T
OR
P
E
DO
: TO
ol
ti
p
-
po
w
e
R
e
d
P
his
hi
ng Em
a
i
l
De
te
c
t
i
On
,”
C
o
mp
ute
rs &
Se
c
ur
ity
,
v
ol.
71,
pp.
10
0
-
1
13
,
20
17,
d
oi
:
1
0.
10
16
/j.
c
ose
.
2
01
7.
0
2.
00
4
.
[
23]
M
.
M
ogh
im
i a
nd
A.
Y.
Va
r
ja
n
i,
“
Ne
w r
u
le
-
ba
se
d ph
is
hin
g de
te
c
ti
on
m
e
th
od
,
”
Ex
pe
r
t Sy
ste
ms w
it
h A
pp
li
c
a
ti
ons
,
vol.
53
,
pp.
23
1
-
24
2
,
20
16
,
d
oi
: 1
0.
1
01
6/
j.
e
swa
.
20
16.
0
1.
028
.
[
24]
O.
K.
S
a
hing
oz
,
E.
B
ube
r
,
O.
De
m
ir
a
nd B
.
Dir
i,
“
M
a
c
hin
e
le
a
r
ni
ng ba
se
d ph
is
hi
ng de
te
c
t
io
n f
r
om
UR
L
s
,
”
E
x
p
e
rt
Sy
s
te
m
s w
it
h A
pp
lic
at
io
ns
,
v
ol.
1
17,
pp
.
3
45
-
35
7
,
20
19
,
doi
:
1
0.
10
16
/j.
e
swa
.
20
18.
0
9.
0
29
.
[
25]
S
.
S
m
a
di,
N.
Asla
m
a
n
d
L
.
Z
ha
n
g
,
“
De
te
c
ti
on of
on
li
ne
ph
is
hi
ng
e
m
a
i
l u
si
ng
d
yna
m
ic
e
v
ol
vi
ng
ne
ur
a
l ne
tw
or
k
ba
se
d on r
e
inf
or
c
e
m
e
n
t le
a
r
ni
ng
,”
De
c
i
si
on Sup
po
rt Sy
s
te
m
s
,
v
ol.
1
07,
pp
.
88
-
10
2
,
201
8
,
doi
:
1
0.
10
16
/j.
ds
s.
20
18.
01.
0
01
.
[
26]
H.
Y.
A.
Ab
uta
ir
a
n
d A.
B
e
l
gh
it
h,
“
Us
in
g
C
as
e
-
B
a
se
d R
e
a
s
on
in
g f
or
P
hi
sh
in
g
De
te
c
ti
on
,
”
Pr
oc
e
d
ia C
om
p
ute
r
Sc
ie
nc
e
,
vo
l.
10
9,
pp.
28
1
-
2
88
,
20
17
,
do
i
:
10.
10
16
/j.
pr
o
c
s.
20
17.
05.
3
52
.
[
27]
W
.
Ha
d
i,
F
.
Abur
ub a
n
d S
.
Al
ha
wa
r
i,
“
A ne
w f
a
st a
sso
c
ia
ti
ve
c
la
s
sif
ic
a
t
io
n a
l
gor
it
hm
f
or
de
te
c
t
in
g p
hi
sh
i
ng
we
b
si
te
s
,”
A
pp
lie
d So
ft C
om
pu
ti
ng J
o
ur
na
l
,
vo
l.
48,
pp
.
729
-
73
4
,
20
16
,
d
oi
:
1
0.
1
01
6/
j.
a
soc
.
20
16.
0
8.
0
05
.
[
28]
X.
Z
ha
ng,
D.
S
hi,
H.
Z
ha
n
g,
W
.
L
iu a
nd R
.
L
i,
“
Ef
f
ic
ie
nt De
te
c
ti
on of
P
h
is
hi
ng At
ta
c
k
s
w
it
h
H
ybr
id Ne
u
r
a
l
Ne
t
wor
ks
,”
20
18 I
E
EE 18
th I
n
te
r
na
ti
on
al C
o
nfe
re
nc
e
o
n C
omm
un
ic
a
ti
on T
e
c
h
no
lo
gy
(
I
C
C
T
)
,
2018,
pp.
84
4
-
84
8
,
doi
: 1
0.
11
09
/I
C
C
T.
20
18.
8
60
00
18.
[
29]
J
. M
a
o
,
e
t al.
,
“
De
te
c
ti
ng P
h
is
hi
ng W
e
bs
ite
s v
ia
Ag
g
r
e
ga
t
io
n Ana
ly
si
s of
P
a
ge
L
a
yo
ut
s
,
”
P
ro
c
e
di
a
C
om
put
e
r
Sc
ie
nc
e
,
vo
l.
12
9,
pp.
22
4
-
2
30
,
20
18
,
do
i:
10.
10
16
/j.
pr
o
c
s.
20
18.
03.
0
53.
[
30]
J.
Na
z
a
r
i
os,
“
P
h
is
hi
ng C
or
pu
s
,
”
20
18
ur
l
: ht
tp
s:
//m
on
ke
y.
or
g
/~
jo
se
/
ph
is
hi
ng
/ Ac
c
e
sse
d 26
/0
1/
20
19.
[
31]
S
pa
m
As
sa
s
in
,
“
P
ub
lic
C
or
p
us,
”
2
01
8.
ur
l
: h
tt
ps
:/
/s
pa
m
a
ssa
s
si
n.
a
pa
c
he
.
or
g/
ol
d/
pu
bl
ic
c
or
pu
s/.
Ac
c
e
sse
d 2
6/
01
/2
01
9
ur
l
: ht
tp
s:
//
doc
s
.
a
p
wg.
or
g/r
e
por
ts
/a
p
wg
_tr
e
nd
s_r
e
por
t_
q
2_2
01
8.
p
df
.
Ac
c
e
s
se
d 22
/0
2/
20
19.
[
32]
M
.
Khon
ji,
A.
Jone
s a
nd Y.
I
r
a
qi,
“
A S
tu
dy of
F
e
a
t
ur
e
S
ubse
t E
va
l
ua
t
or
s a
n
d F
e
a
tur
e
S
u
bse
t S
e
a
r
c
hin
g M
e
th
od
s
f
or
P
hi
sh
in
g
C
la
s
sifi
c
a
ti
on
,”
I
n p
roc
e
e
di
ng
s of
the
8
th
Ann
ua
l C
ol
la
bo
ra
ti
on,
E
le
c
t
ro
nic
me
ss
ag
in
g,
An
ti
-
A
bu
se
and
Sp
am C
o
nf
e
re
nc
e
(
C
EA
S '
1
1
)
,
AC
M
Ne
w Y
or
k,
US
A,
20
11,
p
p.
1
35
-
14
4
,
d
oi
: 10.
11
45
/2
03
03
76.
2
03
03
92.
[
33]
I.
R.
A.
Ha
m
id,
J.
A
ba
wa
jy a
n
d T.
H.
Kim
,
“
U
si
ng F
e
a
tur
e
S
e
le
c
ti
on a
nd C
la
s
sif
ic
a
t
io
n S
c
he
m
e
f
or
Au
tom
a
t
in
g
P
his
hi
ng Em
a
i
l
De
te
c
t
io
n,
”
St
ud
ie
s in I
nf
or
ma
tic
s an
d C
o
nt
ro
l
,
vo
l.
22
,
no.
1
,
61
-
70
,
20
13
,
doi
:
10.
2
48
46
/v
22
i1
y2
01
30
7.
[
34]
N.
Va
i
sh
na
w a
nd S
.
R
.
T
a
n
d
a
n
,
“
De
ve
lo
pm
e
n
t of
A
n
ti
-
P
h
is
hi
ng M
ode
l f
or
C
la
s
sif
ic
a
t
io
n of
P
h
is
hi
ng
E
-
m
ai
l
,
”
I
nte
rn
at
io
na
l
J
ou
rn
al o
f
Adv
anc
e
d
Re
se
arc
h
in C
o
mp
ute
r
an
d C
om
mu
nic
at
io
n En
gi
ne
e
r
in
g,
vol.
4
,
no.
6
,
pp.
39
-
45
,
20
15,
do
i:
10.
17
14
8/I
JAR
C
C
E.
20
15.
4
61
0 3
9
.
[
35]
L.
Zh
a
n
g
a
nd Y
.
Tia
n
-
sh
un
,
“
F
il
te
r
i
ng
ju
nk m
a
il w
i
th a
m
a
x
im
um
e
ntr
op
y m
o
de
l
,
”
In
P
ro
c
e
e
di
ng
of
20
th
I
nte
rn
at
io
na
l C
on
fe
re
nc
e
o
n C
om
pu
te
r
Pr
oc
e
s
si
ng o
f O
r
ie
n
ta
l L
an
gu
age
s (
I
C
C
P
OL
,
200
3)
,
20
03
,
pp.
4
46
-
4
53
.
Evaluation Warning : The document was created with Spire.PDF for Python.