I
AE
S In
t
er
na
t
io
na
l J
o
urna
l o
f
Art
if
icia
l In
t
ellig
ence
(
I
J
-
AI)
Vo
l.
9
,
No
.
1
,
Ma
r
ch
2
0
2
0
,
p
p
.
146
~
154
I
SS
N:
2
2
5
2
-
8938
,
DOI
: 1
0
.
1
1
5
9
1
/i
j
ai.
v
9
.i
1
.
p
p
146
-
1
5
4
146
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
i
a
esco
r
e.
co
m
Deep
O
SN:
Bring
i
ng
deep
lea
rni
ng
a
s
m
a
licio
us de
tec
tion
sche
m
e in
onlin
e
so
cia
l net
w
o
rk
P
utr
a
Wa
nd
a
1
,
M
a
rselin
a
E
nd
a
h H
is
w
a
t
i
2
,
H
ua
ng
J
.
J
ie
3
1
,2
De
p
a
rtm
e
n
t
o
f
In
f
o
r
m
a
ti
c
s
,
Un
iv
e
rsit
y
o
f
Re
sp
a
ti
Yo
g
y
a
k
a
rta
,
In
d
o
n
e
sia
1,
3
De
p
a
rtm
e
n
t
o
f
Co
m
p
u
ter S
c
ien
c
e
&
Tec
h
n
o
lo
g
y
,
Ha
rb
in
Un
iv
e
rsity
o
f
S
c
ien
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
,
C
h
in
a
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Au
g
4
,
2
0
1
9
R
ev
i
s
ed
Oct
19
,
2
0
1
9
A
cc
ep
ted
Dec
2
6
,
2
0
1
9
M
a
n
u
a
l
a
n
a
ly
sis
f
o
r
m
a
li
c
io
u
s
p
r
e
d
ictio
n
i
n
On
li
n
e
S
o
c
ial
Ne
tw
o
r
k
s
(OSN)
is
ti
m
e
-
c
o
n
su
m
in
g
a
n
d
c
o
stly
.
W
i
th
g
ro
w
in
g
u
se
rs
w
it
h
in
th
e
e
n
v
iro
n
m
e
n
t,
it
b
e
c
o
m
e
s
o
n
e
o
f
th
e
m
a
in
o
b
sta
c
les
.
Cu
rre
n
tl
y
,
m
a
n
y
r
e
se
a
r
c
h
c
o
m
m
u
n
it
ies
h
a
v
e
p
ro
p
o
se
d
lea
rn
i
n
g
tec
h
n
i
q
u
e
s
t
o
a
u
to
m
a
te
se
c
u
rit
y
tas
k
s,
in
c
lu
d
in
g
a
n
o
m
a
lo
u
s
d
e
tec
ti
o
n
,
m
a
li
c
io
u
s
li
n
k
p
re
d
icti
o
n
,
a
n
d
in
tr
u
sio
n
d
e
tec
ti
o
n
in
OSN.
De
e
p
lea
rn
i
n
g
is
a
g
ro
w
in
g
a
lg
o
rit
h
m
th
a
t
g
a
in
s
a
b
ig
s
u
c
c
e
ss
in
c
o
m
p
u
ter
v
isio
n
p
ro
b
lem
s.
In
th
is
p
a
p
e
r,
w
e
p
ro
p
o
se
a
n
o
v
e
l
d
e
e
p
lea
rn
in
g
a
rc
h
it
e
c
tu
re
to
e
sta
b
li
sh
t
h
e
O
S
N
se
c
u
rit
y
te
c
h
n
iq
u
e
t
o
b
e
c
o
m
e
m
o
re
in
telli
g
e
n
t
f
o
r
d
e
tec
ti
n
g
m
a
li
c
io
u
s a
c
ti
v
it
ies
.
K
ey
w
o
r
d
s
:
Dee
p
L
ea
r
n
i
n
g
Ma
licio
u
s
Dete
c
tio
n
On
li
n
e
So
cial
Net
w
o
r
k
P
r
iv
ac
y
-
P
r
eser
v
i
n
g
Secu
r
it
y
Mo
d
el
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
:
P
u
tr
a
W
an
d
a,
Dep
ar
t
m
en
t o
f
I
n
f
o
r
m
atic
s
,
Un
i
v
er
s
it
y
o
f
R
esp
ati
Yo
g
y
a
k
ar
ta
,
I
n
d
o
n
esia
,
L
a
k
s
d
a
A
d
is
u
cip
to
St,
Sle
m
an
,
Yo
g
y
a
k
ar
ta,
I
n
d
o
n
esia.
E
m
ail:
w
p
w
a
w
a
n
@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
On
li
n
e
So
cial
Ne
t
w
o
r
k
is
a
p
o
p
u
lar
ap
p
licatio
n
f
o
r
s
h
ar
i
n
g
v
ar
io
u
s
d
ata
ite
m
s
i
n
cl
u
d
es
v
id
eo
s
,
p
h
o
to
s
,
an
d
m
ess
a
g
e
s
.
I
t
co
n
n
ec
ts
p
eo
p
le
ar
o
u
n
d
th
e
ea
r
th
.
Ho
w
e
v
er
,
an
an
o
m
a
lo
u
s
i
s
s
u
e
lik
e
f
a
k
e
ac
co
u
n
ts
b
ec
o
m
e
s
a
s
i
g
n
i
f
ica
n
t
co
n
ce
r
n
in
OSN
p
r
o
tectio
n
.
T
h
e
ir
r
eg
u
lar
ac
co
u
n
t
b
ec
o
m
es
o
n
e
o
f
t
h
e
m
ai
n
c
h
alle
n
g
e
s
in
c
u
r
r
en
t
O
SN
b
ec
au
s
e
g
r
o
w
i
n
g
u
s
er
s
o
n
t
h
e
OS
N
h
eig
h
ten
t
h
e
p
r
o
b
ab
ilit
y
lev
e
l
o
f
m
alicio
u
s
ac
t
iv
i
ties
.
Ho
w
e
v
er
,
it
r
e
m
ai
n
s
s
ig
n
i
f
i
ca
n
t
ch
a
llen
g
es
in
O
SNs
,
w
h
ic
h
h
a
v
e
a
lar
g
e
n
u
m
b
er
o
f
u
s
e
r
s
an
d
in
f
o
r
m
atio
n
w
it
h
a
d
y
n
a
m
ic
en
v
ir
o
n
m
e
n
t.
I
n
th
e
tr
an
s
m
is
s
io
n
p
r
o
ce
s
s
,
th
e
OSN
s
ab
le
to
r
u
n
eit
h
er
in
d
ep
en
d
en
tl
y
o
r
d
ep
en
d
en
tl
y
g
r
o
u
p
[
1
]
.
I
n
an
OS
N
s
y
s
te
m
,
t
h
e
p
r
i
m
ar
y
s
o
u
r
ce
o
f
t
h
e
t
h
r
ea
t
co
m
es
f
r
o
m
an
o
m
alo
u
s
ac
ti
v
it
ies.
I
n
f
ec
ted
u
s
e
r
ca
n
s
p
r
ea
d
th
e
f
a
k
e
o
r
d
ec
eiv
e
in
f
o
r
m
at
io
n
f
o
r
t
h
e
tar
g
et
u
s
er
.
T
h
e
lar
g
e
o
f
O
SN
is
a
n
attac
k
i
n
g
tar
g
e
t
f
o
r
S
y
b
il
b
y
h
ar
n
e
s
s
i
n
g
f
a
k
e
ac
co
u
n
t.
H
u
m
an
o
r
b
o
t
ca
n
cr
ea
te
a
f
ak
e
ac
co
u
n
t
to
i
m
p
er
s
o
n
ate
th
e
OSN
u
s
er
s
i
n
a
f
alse
n
a
m
e
a
n
d
p
o
s
t
m
i
s
lead
in
g
in
f
o
r
m
a
tio
n
o
n
th
eir
p
r
o
f
iles
.
Fo
r
ex
a
m
p
le,
i
n
d
iv
id
u
a
l
ac
co
u
n
t
s
an
d
b
o
t
ac
co
u
n
t
s
s
h
ar
e
a
s
i
m
i
lar
id
en
t
it
y
li
k
e
a
n
a
m
e
o
r
p
ictu
r
es
[
2
]
.
T
h
e
an
o
m
alo
u
s
i
n
tr
u
d
er
ca
n
ta
k
e
a
d
v
a
n
ta
g
e
o
f
th
e
d
r
a
w
b
ac
k
to
m
o
n
ito
r
th
e
t
ar
g
et,
an
d
th
e
s
u
s
p
icio
u
s
n
o
d
es
ca
n
p
o
s
e
a
S
y
b
i
l
th
r
ea
t
[
3
]
.
C
u
r
r
en
t
p
u
b
lic
OS
N
r
e
m
ain
s
s
h
o
r
tco
m
i
n
g
i
n
s
ev
er
a
l a
r
ea
s
,
i
n
clu
d
i
n
g
li
n
k
d
etec
tio
n
,
m
alicio
u
s
s
en
ti
m
e
n
t,
a
n
o
m
a
lo
u
s
ac
co
u
n
t
s
,
a
n
d
ac
tiv
itie
s
.
I
n
p
r
ac
tical
c
a
s
es,
t
h
e
an
o
m
al
y
is
s
u
e
s
ar
e
co
m
p
r
is
ed
o
f
d
if
f
er
en
t
t
y
p
es
o
f
th
r
ea
t
s
lik
e
v
er
tical
attac
k
to
th
e
OSNs
'
p
r
o
v
id
er
s
o
r
h
o
r
izo
n
tal
attac
k
ca
u
s
ed
to
OSN
m
e
m
b
er
s
[
4
]
.
Ma
n
y
p
ap
er
s
p
r
o
p
o
s
e
co
n
v
en
tio
n
al
tech
n
iq
u
es
to
d
ea
l
w
it
h
th
e
is
s
u
es,
in
cl
u
d
i
n
g
s
ta
tis
tic,
r
u
le
-
b
ased
,
an
d
clu
s
ter
i
n
g
.
Ho
w
e
v
er
,
th
e
cu
r
r
en
t
tech
n
iq
u
es
ar
e
lack
o
f
p
er
f
o
r
m
i
n
g
f
a
s
t
d
etec
tio
n
o
f
m
al
ici
o
u
s
ac
ti
v
itie
s
.
T
h
e
co
n
v
en
tio
n
al
m
et
h
o
d
r
e
m
ai
n
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
A
r
ti
f
I
n
tell
I
SS
N:
2252
-
8938
Dee
p
OS
N
:
B
r
in
g
in
g
d
ee
p
lea
r
n
in
g
a
s
ma
licio
u
s
d
etec
tio
n
s
ch
eme
in
o
n
lin
e
s
o
cia
l n
etw
o
r
k
(
P
u
tr
a
W
a
n
d
a
)
147
s
h
o
r
tco
m
i
n
g
to
d
etec
t
an
d
r
e
s
p
o
n
d
to
f
a
k
e
ac
co
u
n
ts
b
e
f
o
r
e
th
e
y
i
n
ter
ac
t
w
it
h
r
ea
l
u
s
e
r
s
.
I
t
en
ab
le
s
f
ak
e
a
cc
o
u
n
t
s
to
s
ta
y
in
t
h
e
n
et
w
o
r
k
f
o
r
estab
lis
h
in
g
m
alicio
u
s
co
n
n
ec
tio
n
s
an
d
ac
cu
m
u
late
t
h
e
b
u
lk
o
f
ac
tiv
i
t
y
d
ata
.
T
h
u
s
,
v
ar
io
u
s
s
tu
d
ie
s
p
r
o
p
o
s
e
cu
r
r
en
t te
ch
n
iq
u
e
s
to
i
m
p
r
o
v
e
OSN
s
ec
u
r
it
y
tec
h
n
iq
u
e
s
[
1
-
5
]
.
OSN
h
a
s
lar
g
e
-
s
ca
le
m
e
m
b
er
s
,
an
d
it
m
a
y
r
e
g
is
ter
t
h
o
u
s
a
n
d
s
o
f
n
e
w
u
s
er
s
e
v
er
y
d
a
y
.
U
n
l
u
ck
i
l
y
,
t
h
e
attac
k
er
s
tr
y
to
m
a
k
e
f
ak
e
ac
c
o
u
n
t
s
at
s
ca
le.
He
n
ce
,
th
e
OS
N
s
y
s
te
m
i
s
n
ec
ess
ar
y
to
cr
ea
t
e
a
clu
s
ter
-
le
v
el
a
n
d
r
o
b
u
s
t
d
etec
tio
n
m
o
d
el.
T
h
is
p
ap
er
p
r
o
p
o
s
es
a
lear
n
in
g
m
o
d
el
to
im
p
r
o
v
e
m
alicio
u
s
d
ete
ctio
n
in
t
h
e
O
SN
b
y
tr
ain
i
n
g
en
g
i
n
ee
r
ed
f
ea
t
u
r
es
o
f
ea
c
h
ac
co
u
n
t.
Se
v
er
al
s
tu
d
ies
s
u
g
g
est
a
co
m
m
o
n
a
n
o
m
alo
u
s
d
etec
t
io
n
tech
n
iq
u
e;
f
o
r
in
s
tan
ce
,
a
s
tu
d
y
p
r
ese
n
ts
a
f
a
k
e
ac
co
u
n
t
d
etec
tio
n
m
o
d
el
b
y
co
m
p
u
ti
n
g
t
h
e
s
i
m
ilar
it
y
o
f
th
e
u
s
er
'
s
f
r
ien
d
s
.
T
h
e
ap
p
r
o
ac
h
c
alcu
late
s
th
e
ad
j
ac
en
c
y
m
a
tr
ix
o
f
th
e
g
r
ap
h
.
T
o
m
ar
k
t
h
e
u
s
e
r
as
b
en
ig
n
o
r
f
a
k
e,
it
an
al
y
ze
s
th
e
f
r
ie
n
d
'
s
n
et
w
o
r
k
s
tr
u
ct
u
r
e
[
5
]
.
Fig
u
r
e
1
d
e
p
icts
th
e
ev
o
lu
tio
n
o
f
OSN
i
n
co
n
n
ec
ti
v
it
y
an
d
an
al
y
tical
tec
h
n
iq
u
es.
L
o
c
a
l
S
o
c
i
a
l
N
e
t
w
o
r
k
P
u
b
l
i
c
S
o
c
i
a
l
N
e
t
w
o
r
k
I
n
t
e
l
l
i
g
e
n
c
e
S
o
c
i
a
l
N
e
t
w
o
r
k
H
u
m
a
n
(
u
s
e
r
s
)
p
r
o
v
i
d
e
t
r
a
n
s
a
c
t
i
o
n
d
a
t
a
A
c
t
o
r
s
:
H
u
m
a
n
t
o
H
u
m
a
n
H
u
m
a
n
&
T
h
i
n
g
u
s
e
r
s
p
r
o
v
i
d
e
t
r
a
n
s
a
c
t
i
o
n
d
a
t
a
A
c
t
o
r
s
:
H
u
m
a
n
t
o
H
u
m
a
n
H
u
m
a
n
t
o
T
h
i
n
g
H
u
m
a
n
&
T
h
i
n
g
u
s
e
r
s
p
r
o
v
i
d
e
t
r
a
n
s
a
c
t
i
o
n
d
a
t
a
A
c
t
o
r
s
:
H
u
m
a
n
t
o
H
u
m
a
n
H
u
m
a
n
t
o
T
h
i
n
g
T
h
i
n
g
t
o
T
h
i
n
g
T
h
i
n
g
p
r
o
d
u
c
e
m
o
r
e
d
a
t
a
t
h
a
n
h
u
m
a
n
C
o
n
n
e
c
t
i
v
i
t
y
A
n
a
l
y
t
i
c
a
l
T
e
c
h
n
i
q
u
e
D
a
t
a
A
n
a
l
y
t
i
c
M
o
d
e
l
:
M
a
c
h
i
n
e
L
e
a
r
n
i
n
g
&
D
e
e
p
L
e
a
r
n
i
n
g
(
A
I
)
D
a
t
a
A
n
a
l
y
t
i
c
M
o
d
e
l
:
S
t
a
t
i
s
t
i
c
a
l
a
n
d
R
u
l
e
-
B
a
s
e
d
L
e
a
r
n
i
n
g
A
n
a
l
y
t
i
c
M
o
d
e
l
:
C
l
u
s
t
e
r
i
n
g
,
M
a
n
u
a
l
I
n
s
p
e
c
i
t
o
n
Fig
u
r
e
1
.
T
h
e
ev
o
lu
tio
n
ar
y
h
is
to
r
y
o
f
s
o
cial
n
et
w
o
r
k
tech
n
o
l
o
g
y
B
ased
o
n
Fig
u
r
e
1
,
th
e
n
e
x
t
g
en
er
atio
n
o
f
t
h
e
OSN
s
ec
u
r
it
y
m
o
d
el
is
lear
n
i
n
g
m
et
h
o
d
s
.
A
g
r
o
w
in
g
lear
n
in
g
a
lg
o
r
it
h
m
is
A
r
ti
f
icia
l
NNs
(
A
NNs),
al
s
o
ca
lled
N
Ns,
t
h
at
m
i
m
ics
t
h
e
h
u
m
a
n
b
r
ain
.
T
h
e
al
g
o
r
ith
m
f
o
cu
s
es
o
n
h
o
w
t
h
e
n
e
u
r
o
n
s
p
r
o
ce
s
s
in
f
o
r
m
a
tio
n
an
d
co
n
d
u
ct
co
m
p
u
ta
tio
n
s
[
6
]
.
Ma
n
y
a
s
p
ec
ts
h
a
v
e
i
m
p
le
m
en
ted
A
I
al
g
o
r
it
h
m
s
t
o
ad
d
r
ess
v
ar
io
u
s
p
r
o
b
le
m
s
[
7
-
9
]
.
I
n
th
e
n
e
u
r
al
n
e
t
w
o
r
k
,
t
h
e
n
e
t
w
o
r
k
ele
m
e
n
ts
u
s
u
all
y
ca
lled
n
eu
r
o
n
s
,
w
h
ic
h
ar
e
g
r
o
u
p
ed
in
d
if
f
er
en
t
la
y
er
s
.
I
n
th
e
o
p
er
atio
n
p
r
o
ce
s
s
,
ea
ch
la
y
er
o
f
n
e
u
r
o
n
s
tak
es
i
n
p
u
ts
f
r
o
m
t
h
e
p
r
ev
io
u
s
la
y
er
a
n
d
p
ass
e
s
it
s
o
u
tp
u
t
to
th
e
n
e
x
t
la
y
er
.
E
ac
h
n
e
u
r
o
n
h
as
p
ar
a
m
eter
s
,
in
cl
u
d
in
g
ac
tiv
a
tio
n
f
u
n
ctio
n
s
,
w
eig
h
t
s
,
an
d
b
iases
.
T
o
m
i
n
i
m
ize
er
r
o
r
(
lo
s
s
)
in
th
e
tr
ain
in
g
a
n
d
v
alid
ati
n
g
p
r
o
ce
s
s
,
it tu
n
e
s
an
d
o
p
ti
m
ize
s
th
e
s
etti
n
g
s
b
et
w
ee
n
th
e
p
r
ed
icted
an
d
p
r
o
v
id
ed
v
alu
es.
Dee
p
lear
n
in
g
(
DL
)
i
s
a
p
ar
t
o
f
m
ac
h
in
e
lear
n
i
n
g
w
h
ic
h
m
i
m
ics
t
h
e
b
io
lo
g
ical
n
eu
r
al
n
et
w
o
r
k
s
(
NNs)
in
t
h
e
h
u
m
a
n
b
r
ain
.
D
L
m
o
d
el
s
h
a
v
e
g
ai
n
ed
s
i
g
n
i
f
i
ca
n
t
i
n
ter
est
b
ec
a
u
s
e
o
f
s
u
cc
ess
f
u
l
l
y
ac
h
ie
v
e
a
n
asto
u
n
d
in
g
r
es
u
lt
i
n
co
m
p
u
t
er
v
is
io
n
p
r
o
b
le
m
s
,
i
n
cl
u
d
in
g
i
m
a
g
e
cla
s
s
i
f
ica
tio
n
,
o
b
j
ec
t
d
etec
tio
n
,
te
x
t
class
i
f
icatio
n
,
a
n
d
v
o
ice
r
ec
o
g
n
i
tio
n
.
T
h
e
D
L
ar
ch
itectu
r
e
is
g
r
o
w
i
n
g
w
it
h
t
h
e
in
cr
ea
s
ed
a
m
o
u
n
t
o
f
d
ata
av
ailab
le.
GP
U
co
m
p
u
ti
n
g
e
n
ab
les
u
s
to
ac
h
ie
v
e
s
s
i
g
n
i
f
ica
n
t
i
m
p
r
o
v
e
m
e
n
ts
i
n
co
m
p
u
ti
n
g
ca
p
ab
ilit
ies.
B
y
u
s
i
n
g
GP
Us,
it
ca
n
p
r
o
d
u
ce
f
ast
tr
ain
i
n
g
a
n
d
d
ep
lo
y
m
e
n
t
o
f
DL
m
o
d
els
p
o
s
s
ib
le.
I
t
is
a
p
r
o
m
is
i
n
g
tech
n
iq
u
e
to
i
m
p
r
o
v
e
p
r
ev
io
u
s
m
et
h
o
d
s
.
C
NN
i
s
a
D
L
ar
ch
itect
u
r
e
t
h
at
is
i
m
p
le
m
e
n
ted
i
n
m
o
s
t
co
m
p
u
ter
v
i
s
io
n
tas
k
s
,
in
cl
u
d
in
g
i
m
a
g
e
clas
s
i
f
icati
o
n
,
o
b
j
ec
t
r
ec
o
g
n
itio
n
,
an
d
n
atu
r
al
la
n
g
u
a
g
e
p
r
o
ce
s
s
in
g
.
T
h
e
s
u
cc
ess
f
u
l
p
er
f
o
r
m
a
n
ce
o
f
C
NNs
co
m
p
u
ter
v
is
io
n
p
r
o
b
le
m
s
is
o
n
e
o
f
t
h
e
p
r
i
m
ar
y
r
ea
s
o
n
s
w
h
y
D
L
r
esu
r
f
ac
ed
in
r
ec
en
t
y
ea
r
s
.
Di
f
f
er
en
t
f
r
o
m
t
h
e
C
NN
f
o
r
t
h
e
cla
s
s
i
f
icatio
n
p
r
o
b
lem
f
o
r
ti
m
e
s
er
ies
d
ata,
R
NN
i
s
p
ar
t
o
f
D
L
alg
o
r
ith
m
s
ap
p
lied
f
o
r
p
r
o
ce
s
s
in
g
s
eq
u
e
n
tial d
ata,
s
u
ch
a
s
ti
m
e
s
er
ies o
r
n
at
u
r
al
la
n
g
u
a
g
e
[
1
0
]
.
I
n
th
is
p
ap
er
,
w
e
f
o
cu
s
o
n
b
u
ild
in
g
a
s
ec
u
r
it
y
s
c
h
e
m
e
i
n
OSN
b
y
i
m
p
le
m
e
n
t
i
n
g
a
d
ee
p
lear
n
in
g
tech
n
iq
u
e.
A
t
th
e
f
ir
s
t
s
ta
g
e,
w
e
co
n
s
tr
u
ct
a
n
o
v
el
ap
p
r
o
ac
h
to
d
etec
t
m
alicio
u
s
ac
co
u
n
ts
b
y
tr
ai
n
i
n
g
t
h
e
OSN
f
ea
t
u
r
es
to
d
eter
m
i
n
e
w
h
et
h
er
th
e
ac
co
u
n
t
i
s
b
en
ig
n
o
r
m
a
lic
io
u
s
.
B
ased
o
n
o
u
r
r
ev
ie
w
,
i
t
i
s
th
e
o
n
e
ap
p
r
o
ac
h
th
at
ad
d
r
ess
e
s
t
h
e
a
n
o
m
alo
u
s
p
r
o
b
lem
b
y
an
al
y
zi
n
g
t
h
e
h
ig
h
-
lev
el
attr
ib
u
te
s
o
f
OSN.
No
t
ab
ly
,
w
e
p
r
ese
n
t
t
h
e
m
ai
n
co
n
tr
ib
u
t
io
n
s
o
f
th
e
t
h
esi
s
,
p
ar
ticu
lar
l
y
i
n
t
h
e
m
alic
io
u
s
ac
co
u
n
t c
las
s
i
f
icatio
n
a
s
f
o
llo
w
s
:
a.
W
e
in
tr
o
d
u
ce
a
d
ee
p
lear
n
in
g
ar
ch
itect
u
r
e
f
o
r
th
e
clas
s
if
y
i
n
g
ac
co
u
n
t
b
y
an
a
l
y
zi
n
g
t
h
e
OSN
f
ea
t
u
r
es
g
ath
er
ed
f
r
o
m
t
h
e
h
ig
h
er
-
lev
e
l
f
ea
tu
r
e
s
.
I
n
s
tead
o
f
u
s
in
g
a
co
n
v
e
n
tio
n
al
al
g
o
r
ith
m
s
u
c
h
as
a
s
tatis
tica
l,
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
I
n
t J
A
r
ti
f
I
n
tell
,
Vo
l.
9
,
No
.
1
,
Ma
r
ch
20
20
:
146
–
1
5
4
148
r
u
le
-
b
ased
,
o
r
s
tan
d
ar
d
m
ac
h
i
n
e
lear
n
in
g
a
lg
o
r
it
h
m
,
w
e
d
ev
elo
p
a
lear
n
i
n
g
tech
n
iq
u
e
w
it
h
s
o
m
e
g
e
n
er
ic
f
u
n
ctio
n
s
to
p
r
ed
ict
m
alicio
u
s
ac
co
u
n
ts
b
y
ca
lcu
la
tin
g
th
e
f
ea
tu
r
es.
W
e
p
r
o
v
id
e
th
e
ac
cu
r
ac
y
an
d
lo
s
s
g
r
ap
h
f
o
r
ea
ch
tec
h
n
iq
u
e
a
n
d
m
ea
s
u
r
e
th
e
e
v
al
u
atio
n
m
etr
ic
at
th
e
en
d
o
f
ea
c
h
ch
ap
ter
.
b.
W
e
co
n
s
tr
u
ct
t
h
e
s
u
p
er
v
i
s
e
d
lear
n
in
g
ar
c
h
itect
u
r
e
b
y
p
r
o
p
o
s
in
g
t
h
e
C
NN
m
o
d
el
to
in
cr
ea
s
e
class
i
f
icatio
n
ac
cu
r
ac
y
.
No
ta
b
ly
,
it
i
s
o
n
e
o
f
t
h
e
s
t
u
d
ies
t
h
at
f
o
c
u
s
o
n
e
s
tab
lis
h
i
n
g
a
n
OSN
s
ec
u
r
it
y
tech
n
iq
u
e
b
y
i
m
p
le
m
e
n
ti
n
g
a
lear
n
in
g
al
g
o
r
ith
m
,
esp
ec
iall
y
f
o
r
d
etec
tin
g
m
a
licio
u
s
ac
co
u
n
t
s
o
v
er
t
h
e
n
et
w
o
r
k
.
c.
W
e
test
t
h
e
p
r
o
p
o
s
ed
m
o
d
el
t
o
ac
h
iev
e
s
ta
te
-
of
-
th
e
-
ar
t
r
es
u
lts
,
esp
ec
iall
y
f
o
r
OS
N
m
alic
io
u
s
ac
co
u
n
t
class
i
f
icatio
n
.
T
h
is
s
tu
d
y
tr
ai
n
v
ar
io
u
s
o
f
OSN
f
ea
t
u
r
es
to
co
n
s
tr
u
ct
a
u
s
ef
u
l
m
o
d
el
w
it
h
a
b
en
ch
m
ar
k
d
ataset.
T
h
e
m
o
d
el
co
m
p
u
te
s
OSN
f
ea
t
u
r
es
d
atase
t
as
m
a
tr
i
x
i
n
p
u
t
s
a
n
d
e
m
p
lo
y
s
th
e
m
o
d
el
to
class
i
f
y
th
e
OSN
ac
co
u
n
t.
Mo
r
eo
v
er
,
b
y
es
tab
lis
h
in
g
t
h
e
g
e
n
er
ic
alg
o
r
ith
m
,
w
e
ca
n
m
ea
s
u
r
e
t
h
e
ac
cu
r
ac
y
o
f
m
alicio
u
s
ac
co
u
n
t d
etec
tio
n
c
o
m
p
ar
ed
w
it
h
o
th
er
ap
p
r
o
ac
h
es.
A
t
t
h
e
f
ir
s
t
s
ta
g
e,
w
e
co
n
d
u
ct
th
e
s
t
u
d
y
b
y
r
ev
ie
w
i
n
g
t
h
e
O
SN
p
ap
er
s
as
a
liter
atu
r
e
r
ev
iew
p
r
o
ce
s
s
to
m
a
k
e
w
ell
-
d
ef
i
n
ed
m
et
h
o
d
o
lo
g
y
an
d
to
h
in
d
er
b
ias
r
esu
l
ts
.
T
h
e
ex
p
er
im
e
n
t
p
er
f
o
r
m
s
th
e
f
o
llo
w
i
n
g
s
tep
s
d
u
r
in
g
t
h
e
p
lan
n
i
n
g
s
ta
g
e.
Fir
s
tl
y
,
w
e
u
n
d
er
g
o
t
h
e
liter
at
u
r
e
r
ev
ie
w
,
d
ef
i
n
e
t
h
e
r
esear
c
h
ap
p
r
o
ac
h
,
d
escr
ib
e
p
r
o
b
lem
f
o
r
m
u
la
tio
n
,
an
d
a
n
al
y
ze
th
e
e
x
p
er
i
m
e
n
t r
es
u
lt.
2.
RE
L
AT
E
D
WO
RK
S
A
lar
g
e
s
ca
le
OSN
co
n
s
is
t
o
f
m
illi
o
n
s
o
f
u
s
er
s
an
d
b
illi
o
n
s
o
f
tr
an
s
ac
tio
n
d
ata.
Gr
o
w
i
n
g
OSN
ca
n
in
cr
ea
s
e
t
h
e
p
o
p
u
lar
it
y
o
f
p
eo
p
le
an
d
s
o
cial
r
atin
g
s
.
A
p
r
ac
tical
ex
a
m
p
le,
OSN
u
s
er
s
ca
n
p
r
o
d
u
ce
p
o
p
u
lar
it
y
w
it
h
m
a
n
y
lik
e
s
,
f
o
llo
w
er
s
,
a
n
d
co
m
m
e
n
ts
.
Ho
w
e
v
er
,
it
is
to
o
ea
s
y
to
cr
ea
te
f
a
k
e
ac
co
u
n
t
s
,
o
r
p
eo
p
le
ca
n
b
u
y
it
o
n
lin
e
w
i
th
f
e
w
co
s
ts
.
Fo
r
ex
a
m
p
le,
it
i
s
ea
s
ier
to
b
u
y
T
w
it
ter
an
d
I
n
s
ta
g
r
a
m
f
o
l
lo
w
er
s
an
d
lik
e
s
o
n
th
e
in
ter
n
e
t
[
1
1
]
.
A
lar
g
e
n
u
m
b
e
r
o
f
m
alicio
u
s
ca
u
s
es
o
f
p
r
iv
ate
d
ata
leak
ag
e
an
d
b
ec
o
m
e
th
e
in
itial
s
tep
to
co
m
p
r
o
m
is
e
ex
i
s
ti
n
g
u
s
er
p
r
iv
ac
y
[
1
2
]
.
C
o
m
m
o
n
l
y
,
to
d
etec
t
an
o
m
alo
u
s
ac
co
u
n
ts
i
n
OSN,
th
e
m
e
th
o
d
s
an
al
y
ze
ac
t
iv
i
t
y
v
ar
iatio
n
s
.
U
s
u
all
y
,
t
h
e
u
s
er
s
'
ac
tiv
ities
k
e
ep
o
n
ch
an
g
in
g
i
n
a
p
er
io
d
.
S
u
d
d
en
ch
a
n
g
in
g
o
f
ac
ce
s
s
p
atter
n
f
o
r
th
e
in
f
o
r
m
a
t
io
n
an
d
b
eh
av
io
r
allo
w
s
t
h
e
s
e
r
v
er
to
ca
tch
th
e
s
u
s
p
icio
u
s
ac
co
u
n
t
u
p
.
On
t
h
e
o
th
er
h
an
d
,
d
iv
er
s
e
c
o
m
m
u
n
itie
s
p
r
o
p
o
s
e
a
lear
n
in
g
m
o
d
el
to
ad
d
r
ess
th
e
O
S
N
s
ec
u
r
it
y
p
r
o
b
lem
.
B
y
u
s
in
g
a
lear
n
i
n
g
t
ec
h
n
iq
u
e,
t
h
e
m
o
d
el
ca
n
tr
ain
th
e
f
ea
t
u
r
es
d
ata
i
n
a
p
er
io
d
w
ith
a
lar
g
e
s
ca
le
o
f
th
e
d
ataset.
Fo
r
in
s
ta
n
ce
,
an
in
v
esti
g
atio
n
r
ev
ea
ls
a
m
et
h
o
d
f
o
r
f
ak
e
n
o
d
es
d
etec
tio
n
b
y
co
m
b
in
i
n
g
SVM,
R
F,
an
d
A
d
aB
o
o
s
t
[
1
3
]
.
No
t
j
u
s
t
u
s
i
n
g
t
h
e
OSN
f
ea
t
u
r
es
d
at
a,
th
e
s
tu
d
y
o
f
co
u
n
ter
f
e
it
d
etec
tio
n
ca
n
u
tili
ze
d
y
n
a
m
ic
d
ata
s
u
c
h
a
s
b
eh
a
v
io
r
al
an
al
y
s
is
,
g
r
ap
h
t
h
eo
r
y
,
lear
n
i
n
g
al
g
o
r
ith
m
,
a
n
d
ap
p
licatio
n
d
esig
n
.
B
y
co
m
p
u
ti
n
g
th
e
f
ea
t
u
r
es,
th
e
y
co
n
s
tr
u
ct
v
ar
io
u
s
ap
p
r
o
ac
h
es
to
id
en
t
if
y
a
n
d
class
i
f
y
an
o
m
a
lies
.
T
o
h
in
d
er
th
e
s
u
s
p
icio
u
s
ac
tiv
ities
o
f
t
h
e
in
tr
u
d
er
,
a
s
tu
d
y
ex
p
lo
r
es
a
m
et
h
o
d
b
y
f
o
r
m
i
n
g
a
co
m
m
u
n
it
y
d
etec
tio
n
alg
o
r
ith
m
[
1
4
]
.
C
u
r
r
en
tl
y
,
s
ev
er
al
p
ap
er
s
p
r
o
p
o
s
e
d
ee
p
lear
n
in
g
ap
p
r
o
ac
h
es
to
ad
d
r
ess
s
o
cial
n
e
t
w
o
r
k
p
r
o
b
lem
.
T
h
ey
e
s
tab
li
s
h
th
e
D
L
f
o
r
v
ar
io
u
s
s
ec
u
r
it
y
tas
k
s
s
u
c
h
as
u
s
er
ac
ti
v
it
y
a
n
al
y
s
i
s
,
p
r
o
f
ile
d
etec
tio
n
,
li
n
k
p
r
ed
ictio
n
,
an
d
s
en
ti
m
en
t
an
al
y
s
i
s
.
Fo
r
in
s
ta
n
ce
,
a
p
ap
er
p
r
o
p
o
s
es
a
Dee
p
B
elief
Net
w
o
r
k
(
DB
N)
to
ca
lcu
late
lin
k
p
r
ed
ictio
n
in
s
ig
n
ed
OSN
.
An
o
th
er
s
t
u
d
y
e
x
p
l
o
r
e
m
alic
io
u
s
d
etec
tio
n
i
n
t
h
e
OSN.
T
h
e
m
o
d
el
d
etec
ts
th
e
m
alicio
u
s
ac
co
u
n
t
b
y
ad
o
p
tin
g
t
h
e
lo
n
g
s
h
o
r
t
-
ter
m
m
e
m
o
r
y
(
L
ST
M)
,
a
d
ee
p
lear
n
in
g
a
lg
o
r
ith
m
b
ased
o
n
R
NN
ar
ch
itect
u
r
e.
I
t
co
n
s
tr
u
c
ts
a
f
r
a
m
e
w
o
r
k
to
g
ai
n
t
h
e
b
in
ar
y
clas
s
i
f
icatio
n
w
i
th
1
8
0
m
illi
o
n
Mo
m
o
OSN
d
ataset
[
1
5
]
.
3.
P
RO
P
O
SE
D
M
O
DE
L
I
n
s
p
ir
ed
b
y
A
r
ti
f
icia
l
I
n
telli
g
en
ce
(
A
I
)
p
er
f
o
r
m
an
ce
i
n
v
ar
io
u
s
p
r
o
b
le
m
s
,
i
n
cl
u
d
in
g
in
telli
g
e
n
t
m
ac
h
in
e
s
,
esp
ec
iall
y
i
n
s
m
ar
t
co
m
p
u
ter
p
r
o
g
r
am
s
an
d
ap
p
licatio
n
s
i
n
s
cie
n
ce
an
d
en
g
in
ee
r
in
g
f
o
r
a
b
etter
h
u
m
a
n
li
f
e,
w
e
p
r
o
p
o
s
e
o
u
r
m
o
d
el
b
y
u
s
in
g
th
e
A
I
co
n
c
ep
t.
T
o
ex
p
r
ess
th
e
in
tell
ig
e
n
ce
ca
p
ab
ilit
y
,
t
h
e
r
esear
ch
er
ca
n
co
n
s
tr
u
c
t
s
o
m
e
m
o
d
els
s
u
c
h
as
b
u
ild
d
ec
is
io
n
m
a
k
i
n
g
a
n
d
lear
n
in
g
ta
s
k
s
.
W
ith
o
u
t
u
s
i
n
g
m
an
u
al
p
r
o
g
r
a
m
m
i
n
g
,
th
e
M
L
al
g
o
r
ith
m
allo
w
s
m
ac
h
in
e
s
(
co
m
p
u
ter
s
)
to
lea
r
n
a
n
d
g
et
d
ec
is
io
n
s
w
it
h
s
tatis
t
ical
m
e
th
o
d
s
[
1
6
]
.
I
t
co
n
s
tr
u
ct
s
a
m
o
d
el
to
co
n
d
u
ct
th
e
tr
ain
in
g
p
r
o
ce
s
s
b
y
f
ee
d
i
n
g
i
n
p
u
t
d
ata
an
d
m
ak
e
p
r
ed
ictio
n
s
o
n
n
e
w
u
n
s
ee
n
d
at
a.
Gen
er
all
y
,
th
er
e
ar
e
t
w
o
ca
te
g
o
r
ies
o
f
M
L
:
1
)
s
u
p
er
v
is
ed
le
ar
n
in
g
;
t
h
i
s
m
o
d
el
r
eq
u
ir
es
la
b
eled
d
ata
co
n
s
is
tin
g
o
f
p
air
s
o
f
i
n
p
u
t
a
n
d
th
e
ex
p
ec
ted
o
u
tp
u
t.
T
o
p
r
o
d
u
ce
in
tel
lig
e
n
t
d
ec
i
s
io
n
s
o
n
n
e
w
d
ata,
i
t
lear
n
s
m
ap
p
in
g
s
b
et
w
ee
n
t
h
e
d
ata
p
air
s
.
Sev
er
al
ap
p
licatio
n
s
ap
p
l
y
t
h
e
tec
h
n
iq
u
e,
s
u
c
h
as
i
m
a
g
e
class
i
f
icat
io
n
a
n
d
s
p
ee
ch
r
ec
o
g
n
i
t
io
n
.
2
)
u
n
s
u
p
er
v
is
ed
lear
n
i
n
g
,
t
h
is
m
o
d
el
p
r
ed
icts
th
e
p
atter
n
b
y
u
s
i
n
g
u
n
la
b
eled
d
ata
as
in
p
u
t
an
d
lear
n
s
t
h
e
r
ela
tio
n
s
h
ip
s
w
it
h
i
n
th
e
i
n
p
u
t
d
ata.
T
h
e
m
o
d
el
p
r
o
d
u
ce
s
d
ec
is
io
n
s
r
es
u
l
t
b
ased
o
n
lear
n
ed
p
atter
n
s
.
Ma
n
y
ap
p
licatio
n
p
r
o
p
o
s
es
a
tech
n
iq
u
e
lik
e
u
n
s
u
p
er
v
is
ed
lear
n
i
n
g
,
i
n
cl
u
d
in
g
d
ata
clu
s
ter
i
n
g
a
n
d
an
o
m
al
y
d
etec
tio
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
A
r
ti
f
I
n
tell
I
SS
N:
2252
-
8938
Dee
p
OS
N
:
B
r
in
g
in
g
d
ee
p
lea
r
n
in
g
a
s
ma
licio
u
s
d
etec
tio
n
s
ch
eme
in
o
n
lin
e
s
o
cia
l n
etw
o
r
k
(
P
u
tr
a
W
a
n
d
a
)
149
Sin
ce
a
d
ec
ad
e
ag
o
,
Ma
ch
in
e
L
ea
r
n
i
n
g
m
et
h
o
d
s
h
a
v
e
b
ee
n
u
til
ized
f
o
r
v
ar
io
u
s
s
ec
u
r
it
y
ta
s
k
s
in
cl
u
d
in
g
i
n
s
o
cial
n
et
w
o
r
k
s
.
I
n
s
tead
o
f
u
s
i
n
g
s
ta
n
d
ar
d
ML
al
g
o
r
ith
m
s
,
w
e
f
o
c
u
s
o
n
estab
li
s
h
i
n
g
D
L
tech
n
iq
u
es
a
s
th
e
p
r
o
tectio
n
s
tr
ateg
y
to
en
h
a
n
ce
OS
N
p
r
o
tectio
n
.
T
h
is
p
ap
er
d
escr
ib
es
h
o
w
D
L
ca
n
b
e
a
s
ec
u
r
it
y
s
tr
ate
g
y
to
ad
d
r
ess
m
alicio
u
s
d
etec
tio
n
b
y
u
s
i
n
g
a
l
ea
r
n
in
g
m
o
d
el.
O
n
e
cr
itical
is
s
u
e
O
SN
r
esear
ch
i
s
h
o
w
to
a
r
eliab
ilit
y
lear
n
i
n
g
m
et
h
o
d
s
w
it
h
a
m
o
r
e
in
telli
g
e
n
t
a
n
d
f
a
s
ter
p
er
io
d
.
I
n
th
i
s
p
a
p
er
,
w
e
also
p
r
ese
n
t
th
e
r
ec
e
n
t r
esear
c
h
ap
p
r
o
ac
h
es
in
t
h
e
O
SN a
n
d
h
o
w
to
b
r
i
n
g
th
e
D
L
al
g
o
r
ith
m
to
s
o
l
v
e
t
h
e
OSN
s
ec
u
r
it
y
w
it
h
a
lar
g
e
d
ataset.
I
n
th
e
f
i
n
al
ch
ap
ter
,
w
e
p
r
o
v
id
e
o
p
en
r
ese
ar
ch
o
p
p
o
r
tu
n
itie
s
f
o
r
co
m
m
u
n
i
ties
n
o
t
o
n
l
y
to
lev
er
ag
e
p
r
o
tectio
n
b
u
t a
l
s
o
to
o
p
tim
ize
t
h
e
lear
n
in
g
tec
h
n
iq
u
es
f
o
r
b
u
s
i
n
ess
p
u
r
p
o
s
es.
3
.
1
.
I
nte
llig
ence
a
rc
hite
ct
ur
e
a.
A
cto
r
s
—
I
n
th
e
OSN,
in
d
i
v
id
u
als
o
r
o
r
g
an
izatio
n
s
ar
e
r
ep
r
esen
ted
b
y
a
n
o
d
e.
OSN
e
n
v
ir
o
n
m
e
n
t
's ac
to
r
s
co
m
m
u
n
icate
i
n
th
e
OSN
s
y
s
t
e
m
.
T
h
e
OSN
co
m
m
u
n
it
y
h
as
n
o
d
es
an
d
r
elatio
n
s
b
et
w
ee
n
t
w
o
n
o
d
es
t
h
at
r
ep
r
esen
t
d
ata
ex
c
h
an
g
e
f
o
r
o
b
tain
in
g
lo
ca
l
a
n
d
g
lo
b
al
n
et
w
o
r
k
h
ier
ar
ch
ie
s
.
T
h
e
OSN
g
r
ap
h
w
i
th
G
=
(
N
,
L
)
h
as
p
ar
a
m
eter
N
r
ep
r
esen
ts
n
o
d
es
(
v
er
tices)
in
t
h
e
OS
N
g
r
ap
h
N
=
{
1
,
2
,
3
…
}
,
an
d
p
ar
am
eter
L
r
ep
r
esen
t
s
s
e
t
o
f
l
in
k
s
(
ed
g
es)
i
n
a
s
et
L
=
{
1
,
2
,
3
…
}
.
T
h
e
g
r
a
p
h
d
ef
i
n
es
n
o
d
es
(
n
)
an
d
lin
k
(
l
)
to
co
n
n
ec
t th
e
n
o
d
es.
b.
I
n
telli
g
en
t
C
o
m
p
o
n
e
n
t
—
T
h
i
s
p
ar
t
is
r
esp
o
n
s
ib
le
f
o
r
m
a
n
a
g
in
g
a
n
d
o
r
ch
e
s
tr
ati
n
g
t
h
e
e
n
tire
co
m
m
u
n
icati
o
n
a
m
o
n
g
t
h
e
ac
to
r
s
.
W
e
estab
lis
h
th
is
I
n
te
lli
g
en
t
C
o
m
p
o
n
en
t
as
th
e
m
ai
n
s
u
b
to
p
r
o
v
id
e
s
er
v
ice
an
d
ap
p
licatio
n
m
a
n
ag
e
m
en
t.
I
t
ca
n
g
at
h
er
,
h
a
n
d
le,
an
d
co
n
d
u
ct
d
ata
an
al
y
tic
s
.
T
h
e
co
m
p
o
n
e
n
t
ca
n
h
ar
v
est
a
b
etter
r
ec
o
m
m
en
d
atio
n
,
s
er
v
ice
d
is
co
v
er
y
,
ac
c
u
r
ate
d
a
ta
an
al
y
s
i
s
,
an
d
co
n
tex
t
m
an
a
g
e
m
e
n
t o
f
t
h
e
OSN
e
n
v
ir
o
n
m
e
n
t.
c.
I
n
ter
f
ac
e
—
T
o
en
ab
le
ad
m
i
n
i
s
tr
ato
r
s
to
m
o
n
ito
r
u
s
er
in
ter
a
ctio
n
s
i
n
OSN.
I
t
r
eq
u
ir
es
an
in
ter
f
ac
e
t
h
at
allo
w
s
th
e
i
n
p
u
t o
f
d
ata
an
d
q
u
er
ies to
p
r
o
v
id
e
th
e
r
eq
u
ested
o
u
tp
u
t.
I
n
th
e
I
n
tel
lig
e
n
t
C
o
m
p
o
n
e
n
t,
it
r
eq
u
ir
es
th
e
lear
n
i
n
g
alg
o
r
i
th
m
to
ca
lcu
late
t
h
e
lab
eled
f
ea
tu
r
es
as
in
p
u
t
m
o
d
el
s
to
p
r
ed
ict
an
o
u
tco
m
e.
T
h
ese
f
ea
t
u
r
es
d
escr
i
b
e
in
f
o
r
m
atio
n
ab
o
u
t
s
o
m
e
o
f
th
e
OS
N
ac
co
u
n
t
s
u
c
h
as
n
a
m
e,
g
en
d
er
,
n
u
m
b
e
r
o
f
f
r
ien
d
s
,
n
u
m
b
er
o
f
l
i
n
k
s
,
an
d
s
o
f
o
r
th
.
Fi
g
u
r
e
2
d
is
p
lay
s
t
h
e
in
telli
g
en
ce
OSN
f
r
a
m
e
w
o
r
k
to
o
p
ti
m
ize
b
u
s
i
n
es
s
an
d
s
ec
u
r
it
y
p
u
r
p
o
s
es.
O
S
N
A
c
t
o
r
F
a
k
e
A
c
t
o
r
D
a
t
a
b
a
s
e
C
h
a
o
s
D
a
t
a
D
a
t
a
M
a
n
a
g
e
m
e
n
t
D
a
t
a
V
i
s
u
a
l
i
z
a
t
i
o
n
L
e
a
r
n
i
n
g
A
n
a
l
y
t
i
c
s
T
r
a
i
n
e
d
C
l
a
s
s
i
f
i
e
r
N
e
w
D
a
t
a
P
r
e
d
i
c
t
i
o
n
R
e
s
u
l
t
O
b
j
e
c
t
i
v
e
:
S
e
c
u
r
i
t
y
P
r
e
d
i
c
t
i
o
n
B
u
s
i
n
e
s
s
A
n
a
l
y
t
i
c
s
M
a
r
k
e
t
i
n
g
B
e
h
a
v
i
o
r
,
e
t
c
Fig
u
r
e
2
.
I
n
tellig
e
n
ce
OS
N
w
i
th
d
ee
p
lear
n
in
g
clas
s
i
f
ier
T
h
e
OSN
d
ataset
co
n
s
is
ts
o
f
s
ev
er
al
u
s
er
in
f
o
r
m
atio
n
s
u
c
h
as
id
en
tit
y
an
d
r
elatio
n
.
User
id
en
tit
y
r
ef
er
s
to
a
r
ea
l
n
at
u
r
al
m
e
m
b
er
.
I
t
is
a
u
n
iq
u
e
ac
co
u
n
t
i
n
t
h
e
O
SN.
co
n
s
is
ts
o
f
u
s
er
r
ep
r
esen
tatio
n
in
cl
u
d
in
g
u
s
er
n
a
m
e,
ag
e,
an
d
o
th
er
attr
ib
u
t
es.
r
ep
r
esen
ts
th
e
u
s
er
's
s
o
cial
co
n
n
ec
tio
n
s
in
th
e
OSN.
in
cl
u
d
es
h
o
w
m
a
n
y
f
r
ie
n
d
s
i
n
th
eir
n
et
w
o
r
k
.
OSN
d
ef
i
n
e
s
as
∁
an
d
is
r
e
p
r
esen
ted
b
y
∁
(
∪
,
)
w
h
er
e
∪
=
{
1
,
2
,
…
}
is
th
e
s
et
o
f
u
s
er
id
en
tit
y
,
an
d
th
e
s
et
o
f
th
e
OSN
lin
k
is
d
es
cr
ib
ed
as
⊆
∪
×
∪
.
I
n
th
e
lear
n
in
g
an
al
y
s
is
m
o
d
el,
it is
cr
u
cia
l to
tr
an
s
f
o
r
m
ce
r
tai
n
f
ea
t
u
r
es i
n
a
n
u
m
er
ical
i
n
te
g
er
in
to
t
h
e
b
in
a
r
y
v
ec
to
r
.
3
.
2
.
P
r
o
po
s
ed
CNN
A
co
m
m
o
n
tech
n
iq
u
e
to
b
u
i
ld
th
e
OSN
s
ec
u
r
it
y
m
o
d
el
i
s
m
ac
h
i
n
e
lear
n
i
n
g
.
Ho
w
e
v
e
r
,
m
a
n
u
al
f
ea
t
u
r
e
en
g
i
n
ee
r
i
n
g
i
s
a
co
s
tl
y
an
d
lab
o
r
io
u
s
tas
k
.
I
n
s
te
ad
o
f
u
s
in
g
t
h
e
co
n
v
e
n
tio
n
a
l
ar
ch
itectu
r
e,
w
e
in
tr
o
d
u
ce
a
n
o
v
e
l
C
NN
ar
c
h
itect
u
r
e
to
d
ea
l
w
it
h
m
a
lic
io
u
s
ac
ti
v
itie
s
in
O
SN.
T
h
is
ap
p
r
o
ac
h
ad
o
p
ts
s
u
p
er
v
i
s
ed
lear
n
in
g
to
d
etec
t
ab
n
o
r
m
al
ev
e
n
t
s
o
f
t
h
e
O
S
N
ac
co
u
n
t
b
y
a
n
al
y
z
in
g
v
ar
i
o
u
s
at
tr
ib
u
tes
w
it
h
i
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
I
n
t J
A
r
ti
f
I
n
tell
,
Vo
l.
9
,
No
.
1
,
Ma
r
ch
20
20
:
146
–
1
5
4
150
th
e
m
.
T
h
is
m
o
d
el
u
til
izes O
S
N
attr
ib
u
tes,
w
h
ic
h
ar
e
d
escr
ib
ed
in
t
h
e
ab
o
v
e
s
ec
t
io
n
.
Fi
g
u
r
e
3
d
ep
icts
th
e
C
NN
to
p
o
lo
g
y
to
tr
ai
n
th
e
cla
s
s
i
f
ier
b
y
u
s
i
n
g
en
g
i
n
ee
r
ed
f
ea
t
u
r
es.
Fig
u
r
e
3
.
P
r
o
p
o
s
ed
C
NN
to
p
o
l
o
g
y
f
o
r
tr
ain
in
g
t
h
e
OSN
d
ata
s
et
T
h
e
p
r
o
p
o
s
ed
C
NN
e
m
p
lo
y
s
Dee
p
Neu
r
al
Net
w
o
r
k
w
it
h
s
e
v
er
al
h
id
d
en
la
y
er
s
to
tr
ai
n
a
n
d
test
th
e
m
o
d
el.
I
t
also
u
tili
ze
s
a
g
r
ad
ien
t
d
escen
t
to
m
i
n
i
m
ize
th
e
o
b
jectiv
e
f
u
n
ctio
n
w
it
h
th
e
m
o
d
e
l's
p
ar
am
e
ter
s
.
T
h
e
m
o
d
el
u
p
d
ates
th
e
s
ett
in
g
s
i
n
th
e
o
p
p
o
s
ite
d
ir
ec
tio
n
o
f
th
e
g
r
ad
ien
t
o
f
th
e
o
b
j
ec
tiv
e
f
u
n
cti
o
n
.
Dif
f
er
en
t
f
r
o
m
th
e
r
eg
u
lar
p
o
o
lin
g
la
y
er
,
t
h
is
s
tu
d
y
ad
o
p
ts
a
p
o
o
lin
g
la
y
er
t
o
o
p
tim
ize
a
n
d
ac
ce
ler
ate
tr
ain
in
g
ti
m
e
in
a
n
e
u
r
al
n
et
w
o
r
k
.
B
y
u
s
in
g
t
h
e
p
r
o
p
o
s
ed
C
NN
to
p
o
lo
g
y
,
w
e
ca
lc
u
lat
e
th
e
ac
c
u
r
a
c
y
a
n
d
lo
s
s
o
f
t
h
e
tr
ain
i
n
g
a
n
d
test
in
g
p
r
o
ce
s
s
to
ac
h
iev
e
t
h
e
b
est
r
es
u
lt
w
it
h
t
h
e
d
iv
er
s
e
in
p
u
t
v
ec
t
o
r
.
T
h
is
s
t
u
d
y
ca
lc
u
lates
th
e
1
D
d
ataset
o
f
OS
N,
s
o
th
e
tu
n
i
n
g
an
ap
p
r
o
p
r
iate
h
y
p
er
p
ar
a
m
eter
is
b
e
n
ef
ic
ial
.
W
e
estab
lis
h
a
s
u
p
er
v
is
ed
lear
n
in
g
m
o
d
el
b
y
de
f
in
i
n
g
ca
lcu
latio
n
o
v
er
NN
as f
o
llo
w
s
:
I
n
p
u
t
f
ea
tu
r
es
(
)
∈
Ou
tp
u
ts
(
)
∈
(
.
.
,
{
0
,
1
}
,
{
1
,
.
.
.
,
}
)
Mo
d
el
p
ar
am
eter
s
∈
ℝ
H
y
p
o
th
es
is
f
u
n
ctio
n
ℎ
∶
∶
ℝ
→
ℝ
L
o
s
s
f
u
n
c
tio
n
ℓ
∶
ℝ
×
→
ℝ
+
∶
I
n
th
i
s
s
t
u
d
y
,
w
e
ca
lc
u
late
t
h
e
o
p
tim
izat
io
n
p
r
o
b
le
m
as
f
o
llo
w
s
:
∑
ℓ
(
ℎ
∶
(
(
)
)
,
(
)
)
=
1
(
1
)
I
n
p
ap
er
,
w
e
p
r
o
v
id
e
h
y
p
o
t
h
es
is
f
u
n
c
tio
n
ℎ
∶
∶
ℝ
→
ℝ
in
n
e
u
r
al
n
et
w
o
r
k
p
r
o
ce
s
s
in
g
.
O
n
a
C
NN
w
e
n
ee
d
to
ca
lcu
late
f
o
r
w
ar
d
p
ass
an
d
b
ac
k
w
ar
d
p
ass
to
m
ea
s
u
r
e
th
e
g
r
ad
ien
t
o
f
lo
s
s
f
u
n
ctio
n
i
n
th
e
m
o
d
el.
T
h
e
s
tu
d
y
ca
lcu
late
th
e
f
o
r
w
ar
d
p
ass
to
co
n
v
o
lv
e
in
p
u
t
m
atr
i
x
w
it
h
f
ilter
to
p
r
o
d
u
ce
co
v
o
lu
tio
n
o
u
tp
u
t
∶
as
f
o
llo
w
s
:
∶
ℝ
→
ℝ
(
2
)
∶
(
)
=
+
T
h
e
C
NN
co
n
s
i
s
ts
o
f
th
e
f
ilt
er
s
an
d
b
ias
ter
m
as
th
e
p
a
r
a
m
eter
s
o
f
t
h
e
co
n
v
o
lu
t
io
n
a
l
la
y
er
d
u
r
in
g
tr
ain
i
n
g
.
I
t i
s
a
s
u
p
er
v
i
s
ed
lear
n
in
g
m
o
d
el
t
h
at
h
as t
h
e
i
n
p
u
t,
r
ep
r
esen
tatio
n
,
a
n
d
m
etr
ic
s
to
co
m
p
u
te
te
n
s
o
r
s
in
th
e
h
id
d
en
la
y
er
.
C
N
N
h
as
m
a
n
y
id
en
t
ical
n
e
u
r
o
n
s
a
m
o
n
g
t
h
e
lay
er
s
to
r
u
n
lar
g
e
m
o
d
el
s
’
c
o
m
p
u
tatio
n
w
i
th
a
little
n
u
m
b
er
o
f
p
ar
a
m
eter
s
.
T
h
e
la
y
er
r
ec
eiv
e
s
a
s
i
n
g
le
i
n
p
u
t
(
th
e
f
ea
tu
r
e
m
ap
s
)
a
n
d
co
m
p
u
te
s
t
h
e
f
ea
t
u
r
e
m
ap
s
a
s
i
ts
o
u
tp
u
t
b
y
co
n
v
o
lv
in
g
f
ilter
s
ac
r
o
s
s
t
h
e
f
ea
t
u
r
e
m
ap
s
.
T
h
e
p
ar
a
m
eter
s
o
f
th
e
co
n
v
o
l
u
tio
n
la
y
er
ca
lled
f
ilter
s
a
n
d
b
ac
k
-
pr
o
p
ag
atio
n
m
o
d
el
u
s
ed
to
lear
n
d
u
r
i
n
g
tr
ai
n
i
n
g
.
I
n
th
e
b
ac
k
w
ar
d
p
ass
,
w
e
ca
lc
u
late
t
h
e
v
ec
to
r
-
v
al
u
ed
f
u
n
c
tio
n
∶
ℝ
→
ℝ
w
it
h
th
e
J
ac
o
b
ian
m
atr
i
x
.
(
2
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
A
r
ti
f
I
n
tell
I
SS
N:
2252
-
8938
Dee
p
OS
N
:
B
r
in
g
in
g
d
ee
p
lea
r
n
in
g
a
s
ma
licio
u
s
d
etec
tio
n
s
ch
eme
in
o
n
lin
e
s
o
cia
l n
etw
o
r
k
(
P
u
tr
a
W
a
n
d
a
)
151
T
o
d
ea
l
w
ith
a
co
m
m
o
n
is
s
u
e
i
n
C
NN
tr
ai
n
in
g
,
w
e
ad
o
p
t
a
r
eg
u
lizer
to
d
ea
l
w
it
h
o
v
er
f
itti
n
g
p
r
o
b
lem
s
i
n
t
h
e
tr
ai
n
in
g
p
r
o
ce
s
s
.
I
t
p
r
o
v
id
es
a
w
a
y
o
f
ap
p
r
o
x
i
m
atel
y
co
m
b
i
n
in
g
e
x
p
o
n
en
ti
all
y
m
an
y
d
i
f
f
er
en
t
n
et
w
o
r
k
ar
ch
itect
u
r
es.
A
p
p
l
y
i
n
g
d
r
o
p
o
u
t
is
to
o
b
tain
i
n
g
a
t
h
in
n
ed
n
et
w
o
r
k
i
n
t
h
e
tr
ai
n
i
n
g
p
r
o
ce
s
s
.
T
h
e
s
t
u
d
y
u
tili
ze
s
t
h
e
Dr
o
p
o
u
t
r
e
g
u
lar
izatio
n
d
u
r
i
n
g
t
h
e
tr
ai
n
i
n
g
s
a
m
p
l
e
(
b
ac
k
w
ar
d
p
ass
)
,
n
o
t
i
n
t
h
e
p
r
ed
ictio
n
s
p
r
o
ce
s
s
(
f
o
r
w
ar
d
p
ass
)
.
Dr
o
p
o
u
t
is
a
m
o
d
er
n
an
d
ex
ce
llen
t
r
eg
u
lar
i
ze
r
th
at
is
ea
s
y
to
i
m
p
le
m
e
n
t
an
d
co
m
p
atib
le
w
it
h
m
an
y
tr
ai
n
i
n
g
al
g
o
r
ith
m
s
a
n
d
m
o
d
el
s
.
I
n
th
e
e
x
p
er
i
m
e
n
t,
i
n
s
tead
o
f
d
o
in
g
it
r
a
n
d
o
m
l
y
,
w
e
test
ed
s
o
m
e
t
y
p
e
s
o
f
d
r
o
p
o
u
t
w
ith
d
i
f
f
er
e
n
t
Dr
o
p
o
u
t
v
alu
e.
B
ased
o
n
th
e
te
s
ti
n
g
o
f
th
e
a
m
o
u
n
t,
it
g
i
v
es
t
h
e
co
n
tr
ib
u
tio
n
o
f
t
h
e
n
eu
r
o
n
to
t
h
e
o
u
tp
u
t.
On
t
h
e
f
u
n
c
tio
n
,
co
n
s
id
er
a
n
et
w
o
r
k
h
as
L
h
id
d
en
la
y
er
s
.
W
e
u
s
e
th
e
Dr
o
p
o
u
t
f
u
n
ctio
n
to
ca
lcu
lat
e
th
e
tr
ain
in
g
p
r
o
ce
s
s
f
o
r
p
ar
a
m
eter
s
lik
e
i
n
p
u
t
an
d
B
er
n
o
u
lli
p
r
o
b
ab
ilit
y
p
.
I
n
th
e
s
ta
n
d
ar
d
f
ee
d
f
o
r
w
ar
d
as
f
o
r
∈
{
0
,
⋯
,
−
0
}
th
er
e
ar
e
d
en
o
tes
th
e
in
p
u
ts
v
ec
to
r
in
to
la
y
er
l
,
d
en
o
tes
th
e
o
u
tp
u
t
s
v
ec
to
r
f
r
o
m
la
y
er
l
,
r
ep
r
e
s
en
t
s
an
i
n
d
ep
en
d
en
t
v
ec
to
r
o
f
B
er
n
o
u
lli
r
a
n
d
o
m
v
ar
iab
les
w
it
h
p
r
o
b
a
b
ilit
y
p
o
f
b
ein
g
1
an
d
̌
r
ep
r
esen
ts
t
h
in
n
e
d
o
u
tp
u
ts
w
h
ich
ca
lc
u
lated
b
y
̌
=
∗
.
T
h
u
s
,
w
e
ca
lcu
late
ca
lcu
la
te
th
e
Dr
o
p
o
u
t r
eg
u
laize
r
as
f
o
llo
w
s
:
(
+
1
)
=
(
+
1
)
+
(
+
1
)
(
+
1
)
=
(
(
+
1
)
)
(
3
)
I
n
f
ee
d
f
o
r
w
ar
d
o
p
er
atio
n
,
w
e
ca
lcu
late
t
h
e
r
eg
u
lizer
w
it
h
th
e
f
o
llo
w
in
g
f
o
r
m
u
la:
~
B
e
r
n
oul
l
i
pr
ob
a
b
il
ity
̌
=
∗
(
+
1
)
=
(
+
1
)
̌
+
(
+
1
)
(
4
)
(
+
1
)
=
(
(
+
1
)
)
T
o
ca
lcu
late
th
e
n
ex
t
la
y
er
,
th
e
r
eg
u
lizer
u
ti
lizes
t
h
e
th
i
n
n
ed
o
u
tp
u
ts
(
+
1
)
as
n
e
w
i
n
p
u
t.
T
h
e
p
r
o
ce
s
s
is
ap
p
lied
to
ea
ch
la
y
er
in
th
e
h
id
d
en
la
y
er
.
T
h
is
am
o
u
n
t
is
to
co
n
s
tr
u
ct
to
a
s
u
b
-
n
e
t
w
o
r
k
s
a
m
p
lin
g
f
r
o
m
a
b
ig
g
er
n
e
t
w
o
r
k
.
I
n
th
e
tr
ain
i
n
g
ti
m
e,
it
b
ac
k
-
p
r
o
p
a
g
ates
th
e
d
er
iv
ati
v
es
o
f
t
h
e
l
o
s
s
f
u
n
ctio
n
v
ia
t
h
e
s
u
b
-
n
et
w
o
r
k
.
3
.
3
.
Da
t
a
s
et
OSN
h
a
s
v
ast
d
ata
i
n
t
h
e
e
n
v
ir
o
n
m
e
n
t,
i
n
cl
u
d
in
g
li
n
k
i
n
f
o
r
m
atio
n
,
u
s
er
p
r
o
f
ile
s
,
an
d
ac
tiv
it
y
b
eh
av
io
r
.
T
h
e
OS
N
u
s
er
p
r
o
f
i
l
e
r
ep
r
esen
ts
h
is
to
r
ical
in
f
o
r
m
a
tio
n
o
f
u
s
e
r
ac
tiv
i
ties
i
n
t
h
e
O
SN e
n
v
ir
o
n
m
e
n
t
to
ca
p
tu
r
e
t
h
e
r
eg
u
lar
u
s
er
's
e
x
p
ec
ted
b
eh
av
io
r
.
At
t
h
e
i
n
it
ial
s
tag
e,
w
e
g
at
h
er
t
h
e
d
ata
f
o
r
m
h
i
g
h
p
r
o
f
ile
OSN
in
f
o
r
m
atio
n
an
d
u
s
e
t
h
e
i
n
f
o
r
m
atio
n
to
co
n
s
tr
u
ct
t
h
e
e
x
tr
ac
ted
f
ea
t
u
r
es.
T
h
e
m
o
d
els
ca
p
t
u
r
e
a
c
h
ar
ac
ter
is
tic
f
ea
t
u
r
e
o
f
a
n
O
SN
u
s
er
l
ik
e
t
h
e
u
s
er
n
a
m
e,
e
-
m
ail
ad
d
r
ess
,
th
e
m
es
s
ag
e
s
tr
ea
m
,
lo
ca
tio
n
,
o
r
n
ati
v
e
la
n
g
u
a
g
e
th
at
w
a
s
u
s
ed
to
tal
k
a
m
o
n
g
u
s
er
s
.
T
o
s
u
m
it
u
p
,
th
e
m
o
d
el
tak
es
th
e
d
ata
an
d
e
x
tr
ac
ts
a
co
u
p
le
o
f
f
ea
tu
r
e
s
,
an
d
tr
ain
s
t
h
e
f
ea
tu
r
e
s
b
y
u
s
in
g
a
s
u
p
e
r
v
is
ed
lear
n
i
n
g
m
o
d
el.
I
n
th
i
s
p
ap
er
,
w
e
g
at
h
er
ed
an
ex
ten
s
i
v
e
d
atab
ase
o
f
lab
ele
d
OSN
d
ataset
f
r
o
m
a
w
eb
s
er
v
ice
th
at
p
r
o
v
id
es
a
co
r
p
u
s
w
it
h
a
m
ali
cio
u
s
o
r
b
en
ig
n
lab
el.
W
e
co
ll
ec
t
th
e
d
ataset
f
r
o
m
Vir
u
s
T
o
ta
l
an
d
P
h
is
h
T
an
k
a
s
a
p
r
o
v
id
er
th
at
p
r
o
v
id
es
m
an
y
d
ataset
s
f
o
r
O
SN
s
ec
u
r
it
y
s
tu
d
y
.
Vir
u
s
T
o
tal
s
er
v
ice
u
s
ed
to
v
alid
ate
a
U
R
L
,
w
h
et
h
er
it
is
m
alic
io
u
s
o
r
b
en
ig
n
.
W
e
also
co
llected
ab
o
u
t
3
0
,
0
0
0
m
alicio
u
s
ac
co
u
n
ts
f
r
o
m
P
h
i
s
h
T
an
k
.
W
e
co
llected
ab
o
u
t 1
1
0
,
4
6
5
s
a
m
p
les,
w
h
ic
h
co
n
s
is
t o
f
5
5
,
3
3
8
b
e
n
ig
n
a
n
d
5
5
,
1
2
7
m
alicio
u
s
lab
els.
I
n
d
ata
p
r
e
-
p
r
o
ce
s
s
in
g
,
w
e
r
e
q
u
ir
e
to
co
n
v
er
t
i
n
te
g
er
o
r
s
tr
in
g
d
ata
t
y
p
e
i
n
to
a
n
u
m
er
ica
l
ar
r
ay
o
r
m
atr
i
x
.
I
t
r
eq
u
ir
e
s
a
m
e
th
o
d
to
co
n
v
er
t
t
h
e
f
ea
t
u
r
es
i
n
to
a
v
e
cto
r
b
ec
au
s
e
t
h
e
in
p
u
t
m
a
y
b
e
in
te
g
er
s
o
r
s
tr
in
g
s
.
T
y
p
icall
y
,
a
s
t
u
d
y
ca
n
b
u
ild
t
h
e
tr
ain
in
g
a
n
d
test
i
n
g
d
atase
t
in
co
l
u
m
n
s
an
d
r
o
w
s
.
T
h
e
m
et
h
o
d
co
n
s
tr
u
c
ts
a
b
in
ar
y
co
lu
m
n
f
o
r
ea
ch
ca
te
g
o
r
y
a
n
d
p
r
o
d
u
ce
s
a
d
en
s
e
ar
r
a
y
.
T
h
e
s
p
ar
s
e
m
atr
ix
i
n
ea
c
h
co
lu
m
n
r
ep
r
esen
t
s
o
n
e
p
o
s
s
ib
le
v
alu
e
o
f
o
n
e
f
e
atu
r
e.
Data
p
r
e
-
p
r
o
ce
s
s
i
n
g
p
r
o
ce
s
s
is
a
c
r
u
cial
s
ta
g
e
to
co
n
s
tr
u
ct
a
n
ac
c
u
r
ate
lean
i
n
g
m
o
d
el
b
y
u
s
in
g
t
h
e
D
L
al
g
o
r
ith
m
.
4.
RE
SU
L
T
S
A
ND
AN
AL
Y
SI
S
I
n
t
h
is
e
x
p
er
i
m
e
n
t
w
it
h
Dee
p
L
ea
r
n
i
n
g
ar
ch
itect
u
r
e,
w
e
tes
t
OSN
s
o
cial
li
n
k
d
ataset
to
m
ea
s
u
r
e
t
h
e
ac
cu
r
ac
y
an
d
th
e
p
er
f
o
r
m
a
n
ce
ti
m
e.
B
y
t
u
n
i
n
g
d
iv
er
s
e
h
y
p
e
r
p
ar
am
eter
s
,
w
e
o
b
tain
th
e
h
i
g
h
e
s
t
p
er
f
o
r
m
an
c
e
o
f
th
e
n
et
w
o
r
k
.
I
n
t
h
e
tr
ai
n
i
n
g
an
d
tes
tin
g
p
r
o
ce
s
s
,
w
e
t
u
n
e
ep
o
ch
=
5
0
0
,
b
atch
s
ize
=
5
0
an
d
w
it
h
lear
n
in
g
r
ate
(
lr
=
0
.
0
1
)
b
y
f
ee
d
in
g
5
0
0
0
0
Face
b
o
o
k
an
d
5
0
0
0
0
Fli
ck
r
s
a
m
p
les
.
T
ab
le
1
d
escr
ib
es
th
e
p
r
o
p
o
s
ed
C
NN
p
er
f
o
r
m
a
n
ce
,
esp
ec
iall
y
i
n
tr
ain
in
g
a
n
d
test
i
n
g
lo
s
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
I
n
t J
A
r
ti
f
I
n
tell
,
Vo
l.
9
,
No
.
1
,
Ma
r
ch
20
20
:
146
–
1
5
4
152
T
ab
le
1
.
P
er
f
o
r
m
a
n
ce
o
f
p
r
o
p
o
s
ed
C
NN
w
it
h
d
if
f
er
en
t p
ar
a
m
eter
s
etti
n
g
O
p
t
i
mi
z
e
r
T
r
a
i
n
i
n
g
L
o
ss
T
e
st
i
n
g
L
o
ss
A
d
a
m
0
.
5
0
5
3
0
.
5
0
5
9
S
G
D
,
m
=
0
.
5
0
.
5
0
5
8
0
.
5
0
6
0
A
d
a
g
r
a
d
0
.
5
0
4
7
0
.
5
0
4
9
R
M
S
P
r
o
p
0
.
5
1
8
6
0
.
5
2
1
3
A
d
a
m
0
.
5
0
4
2
0
.
5
0
5
0
S
G
D
,
m
=
0
.
5
0
.
5
1
5
8
0
.
5
1
6
2
A
d
a
g
r
a
d
0
.
5
0
7
2
0
.
5
0
7
5
R
M
S
P
r
o
p
0
.
5
0
8
2
0
.
5
0
8
9
W
e
test
g
r
ad
ien
t
d
escen
t
a
n
d
tu
n
e
o
p
ti
m
izatio
n
alg
o
r
it
h
m
s
w
it
h
d
if
f
er
en
t
h
y
p
er
p
ar
am
eter
s
.
T
h
e
s
tu
d
y
s
ets
t
h
e
lear
n
in
g
r
ate
η
(
lr
=
0
.
0
1
)
to
d
eter
m
in
e
t
h
e
s
tr
id
e
s
ize
to
r
ea
ch
a
(
lo
ca
l)
m
i
n
i
m
u
m
.
I
n
t
h
e
tr
ain
i
n
g
p
r
o
ce
s
s
,
w
e
g
ai
n
th
e
b
est
v
alu
e
o
f
o
p
ti
m
izatio
n
w
h
en
tr
ain
i
n
g
th
e
m
o
d
el
w
it
h
Ad
am
o
p
ti
m
izer
.
W
e
f
ee
d
A
d
a
m
o
p
ti
m
izer
i
n
to
th
e
g
r
ap
h
w
it
h
d
if
f
er
en
t
a
m
o
u
n
t
s
o
f
h
id
d
en
la
y
er
s
.
T
h
e
r
esu
l
t
s
h
o
w
s
a
d
y
n
a
m
ic
o
p
tim
izer
ca
n
ac
h
ie
v
e
a
s
m
all
lo
s
s
in
a
lar
g
e
n
u
m
b
er
o
f
h
id
d
en
la
y
er
s
.
An
o
th
er
g
r
ad
ien
t
d
esce
n
t
l
ik
e
SGD
w
i
th
m
o
m
e
n
t
u
m
also
p
r
o
d
u
ce
s
a
co
m
p
e
titi
v
e
r
es
u
lt.
W
e
test
t
h
e
SGD
w
it
h
m
o
m
e
n
t
u
m
(
m
=
0
.
5
)
ca
n
p
r
o
v
id
e
a
b
etter
r
es
u
lt
i
f
w
e
t
u
n
e
in
lar
g
er
ep
o
ch
an
d
b
ig
r
eg
u
lizer
.
Ho
w
e
v
er
,
th
e
ex
p
er
i
m
e
n
t
r
ev
ea
ls
th
at
ad
d
in
g
m
o
r
e
h
id
d
en
la
y
er
s
f
o
r
SGD
ca
n
n
o
t
r
ed
u
ce
th
e
lo
s
s
.
Us
in
g
a
lar
g
e
n
u
m
b
er
o
f
h
id
d
en
la
y
er
ca
u
s
e
o
v
er
h
ea
d
co
m
p
u
tat
io
n
in
a
p
ar
ticu
lar
d
ev
ice,
esp
ec
iall
y
i
n
C
P
U
p
r
o
ce
s
s
in
g
.
W
e
f
in
d
t
h
at
b
y
a
d
d
in
g
m
o
r
e
la
y
er
s
an
d
n
e
u
r
o
n
n
u
m
b
er
s
o
f
C
NN
co
u
ld
n
o
t
en
g
a
g
e
i
n
i
m
p
r
o
v
i
n
g
th
e
p
r
ed
ictiv
e
ca
p
ab
ilit
y
.
B
ec
au
s
e
o
f
t
h
e
l
i
m
itatio
n
o
f
n
e
u
r
o
n
n
u
m
b
er
,
it
e
n
lar
g
e
s
co
m
p
u
tin
g
r
eso
u
r
ce
s
.
T
h
e
ex
p
er
i
m
e
n
t
r
esu
l
ts
s
h
o
w
t
h
e
class
i
f
icatio
n
m
o
d
el
ca
n
ac
h
i
ev
e
h
i
g
h
ac
c
u
r
ac
y
w
it
h
a
tin
y
lo
s
s
.
No
tab
l
y
,
th
e
m
o
d
el
d
o
es n
o
t o
n
l
y
p
r
o
d
u
ce
h
ig
h
ac
c
u
r
ac
y
,
b
u
t t
h
e
g
r
ap
h
a
ls
o
r
u
n
s
in
b
etter
p
er
f
o
r
m
a
n
ce
.
5.
F
UT
UR
E
RE
SE
A
RCH
DIR
E
CT
I
O
N
C
u
r
r
en
tl
y
,
th
e
p
o
p
u
lar
O
SN
p
r
o
v
id
er
s
h
av
e
d
e
v
elo
p
ed
v
a
r
io
u
s
tec
h
n
o
lo
g
ies
w
i
th
th
e
b
en
ch
m
ar
k
d
ataset
to
en
h
a
n
ce
t
h
e
s
ec
u
r
i
t
y
le
v
el
i
n
t
h
eir
s
y
s
te
m
.
Face
b
o
o
k
,
L
in
k
ed
I
n
,
a
n
d
T
w
itter
h
av
e
s
tar
ted
u
s
i
n
g
ar
tif
icial
i
n
telli
g
e
n
ce
s
o
lu
tio
n
s
to
elev
ate
d
ee
p
lear
n
in
g
ca
p
ab
ilit
ies.
T
o
f
lag
p
o
s
ts
au
to
m
atica
ll
y
,
Face
b
o
o
k
ad
o
p
ts
A
I
to
d
etec
t
s
u
icid
al
th
o
u
g
h
ts
.
L
i
n
k
ed
I
n
p
r
ed
icts
t
h
e
h
i
g
h
est
m
atc
h
f
o
r
t
h
e
u
s
e
r
s
'
r
o
le.
T
h
eir
A
I
alg
o
r
ith
m
ca
n
p
r
ed
ict
th
e
m
o
s
t
s
i
m
ilar
p
eo
p
le
s
ee
k
i
n
g
n
e
w
j
o
b
s
o
r
co
n
n
ec
tio
n
s
.
T
w
itt
er
is
u
s
i
n
g
a
n
e
u
r
al
s
y
s
te
m
to
cr
o
p
s
a
p
ictu
r
e
u
s
in
g
f
ac
e
d
etec
tio
n
o
r
m
a
k
i
n
g
a
th
u
m
b
n
ail
f
r
o
m
th
e
w
h
o
le
p
h
o
t
o
[
1
6
].
5
.
1
.
Senti
m
ent
a
na
ly
s
is
P
u
b
lic
OSN
p
r
o
v
id
es
a
lar
g
e
-
s
ca
le
s
e
n
ti
m
en
t
d
ataset
to
s
u
p
p
o
r
t
th
e
r
esear
ch
co
m
m
u
n
i
t
y
.
Fo
r
in
s
ta
n
ce
,
Am
az
o
n
th
at
co
n
tai
n
s
a
r
ic
h
d
ataset
w
it
h
a
v
ast
n
u
m
b
er
o
f
c
u
s
to
m
er
r
ev
ie
ws
an
d
a
b
i
g
s
ize
o
f
tr
an
s
ac
tio
n
s
.
Sev
er
al
p
ap
er
s
h
av
e
p
r
o
p
o
s
ed
d
etec
tin
g
ten
s
io
n
m
o
d
el
s
to
e
x
p
lo
r
e
th
e
s
p
ik
es
f
ea
tu
r
e
i
n
OSN
ten
s
io
n
a
n
d
m
ea
s
u
r
es
th
e
le
v
el
o
f
d
eter
io
r
ated
i
n
t
h
e
r
ela
tio
n
s
h
ip
b
et
w
ee
n
in
d
i
v
id
u
a
ls
o
r
g
r
o
u
p
s
[
1
7
]
.
An
o
th
er
p
ap
er
p
r
o
p
o
s
es
an
en
s
e
m
b
le
class
i
f
ier
to
d
ea
l
w
it
h
s
e
n
ti
m
en
t
an
al
y
s
i
s
[
1
8
]
.
T
h
e
cu
r
r
en
t
s
tu
d
y
o
f
s
e
n
ti
m
en
t
an
al
y
s
is
ad
o
p
ts
a
d
ee
p
lear
n
in
g
d
ea
l
w
it
h
t
h
e
s
e
n
ti
m
e
n
t
class
i
f
icat
io
n
[
1
9
-
2
0
]
b
y
c
alcu
lati
n
g
W
ea
k
l
y
Su
p
er
v
i
s
ed
Mu
lt
i
m
o
d
al
(
W
S
M
-
D
L
)
[
2
1
]
.
Ho
w
e
v
er
,
it
r
eq
u
ir
es
a
n
e
w
w
a
y
to
co
n
s
tr
u
ct
a
b
etter
u
s
a
g
e
o
f
n
o
is
y
lab
els
as
ac
ti
v
i
t
y
b
eh
a
v
io
r
o
r
lo
g
s
.
Se
n
ti
m
en
t
a
n
al
y
s
is
i
n
c
lu
d
in
g
s
e
m
an
t
ic
e
v
alu
a
tio
n
with
DL
al
g
o
r
ith
m
s
b
ec
o
m
e
s
an
ac
ti
v
e
r
esear
ch
ar
ea
in
f
u
r
th
er
.
5
.
2
.
L
ink
pre
dict
io
n
No
w
ad
a
y
s
,
O
SN
s
ca
m
i
s
s
u
es
ar
e
ar
is
in
g
an
d
g
r
o
w
in
g
.
T
h
e
f
o
r
m
o
f
th
e
s
ca
m
ac
tiv
ities
co
u
ld
co
er
ce
th
e
O
SN
u
s
er
s
to
s
en
d
t
h
eir
s
en
s
it
iv
e
in
f
o
r
m
at
io
n
.
T
h
e
m
o
d
el
o
f
li
n
k
p
r
ed
ictio
n
ca
n
s
o
l
v
e
t
h
e
li
n
k
o
f
s
ca
m
w
it
h
a
s
o
p
h
is
tica
ted
tech
n
iq
u
e.
T
h
e
lin
k
p
r
ed
ictio
n
p
r
o
b
le
m
i
n
OSN
w
it
h
th
e
D
L
a
p
p
r
o
ac
h
is
h
o
t
ar
ea
r
esear
ch
.
I
n
t
h
is
r
esear
c
h
,
s
o
m
e
s
t
u
d
ies
u
til
ize
th
e
R
B
M
a
r
ch
itect
u
r
e
to
ca
lcu
late
t
h
e
u
n
k
n
o
w
n
o
r
d
y
n
a
m
ic
lin
k
s
b
et
w
ee
n
u
s
er
s
in
a
n
OS
N.
I
n
lin
k
p
r
ed
ictio
n
,
R
B
M
ca
lcu
late
s
th
e
u
n
k
n
o
w
n
o
r
d
y
n
a
m
ic
lin
k
s
b
et
w
ee
n
u
s
er
s
i
n
a
s
o
cial
n
et
w
o
r
k
en
v
ir
o
n
m
en
t.
I
t
is
d
if
f
er
en
t
f
r
o
m
th
e
ta
s
k
s
o
f
t
h
e
s
e
m
an
t
ic
class
i
f
icatio
n
cr
is
i
s
r
esp
o
n
s
e
p
r
o
b
lem
[
2
2
]
.
An
o
th
er
ex
p
er
i
m
en
t
h
as
f
o
cu
s
ed
o
n
a
d
y
n
a
m
ic
s
ett
in
g
,
p
r
ed
ictin
g
t
h
e
co
n
s
tr
u
ctio
n
an
d
d
estru
ctio
n
o
f
lin
k
s
in
n
et
w
o
r
k
s
.
T
h
e
lin
k
an
o
m
al
y
p
r
o
b
lem
is
w
id
el
y
u
s
ed
as
a
s
ec
u
r
it
y
p
ar
a
m
eter
in
O
SN.
A
s
tu
d
y
p
r
o
p
o
s
es
a
li
n
k
an
o
m
al
y
m
o
d
el
to
d
is
co
v
er
i
n
g
e
m
er
g
in
g
to
p
ics
i
n
s
o
cial
s
tr
ea
m
s
b
y
o
f
f
er
in
g
Seq
u
en
t
iall
y
Di
s
co
u
n
ti
n
g
No
r
m
alize
d
Ma
x
i
m
u
m
L
i
k
eli
h
o
o
d
(
SDNM
L
)
an
d
Klein
b
er
g
’
s
b
u
r
s
t
m
o
d
el
[
2
3
]
.
T
h
e
f
u
r
th
er
p
ar
ad
ig
m
s
o
f
t
h
e
OSN
s
ec
u
r
it
y
an
d
p
r
iv
ac
y
s
h
o
u
ld
in
clu
d
e
r
eliab
le
an
d
ef
f
i
cien
t
a
u
t
h
e
n
ticatio
n
v
ia
s
m
ar
t,
tr
u
s
t,
an
o
n
y
m
it
y
,
d
etec
t
tr
ac
ea
b
ilit
y
,
a
n
d
s
u
p
p
o
r
t
lin
k
p
r
ed
ictio
n
s
.
T
h
e
co
n
v
en
t
io
n
al
s
tati
s
tica
l
tech
n
iq
u
e
ca
n
n
o
t
d
etec
t
an
o
m
alie
s
lik
e
ze
r
o
-
d
a
y
attac
k
s
o
r
u
n
ex
p
lo
ited
v
u
ln
er
ab
ilit
ies
.
T
h
e
f
u
r
t
h
er
o
p
en
r
esear
ch
to
p
ics in
t
h
is
ar
ea
ar
e
r
ec
o
m
m
e
n
d
atio
n
s
er
v
ice,
n
et
w
o
r
k
m
alicio
u
s
ac
ti
v
itie
s
,
an
d
lin
k
class
if
icatio
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
A
r
ti
f
I
n
tell
I
SS
N:
2252
-
8938
Dee
p
OS
N
:
B
r
in
g
in
g
d
ee
p
lea
r
n
in
g
a
s
ma
licio
u
s
d
etec
tio
n
s
ch
eme
in
o
n
lin
e
s
o
cia
l n
etw
o
r
k
(
P
u
tr
a
W
a
n
d
a
)
153
5
.
3
.
Ano
m
a
lo
u
s
predict
io
n
Ma
licio
u
s
d
etec
t
io
n
i
s
k
in
d
s
o
f
a
p
r
ac
tical
m
et
h
o
d
to
b
u
ild
s
ec
u
r
it
y
p
o
licies
[
2
4
]
.
T
h
u
s
,
th
e
lear
n
i
n
g
ap
p
r
o
ac
h
is
a
h
elp
f
u
l
tech
n
iq
u
e
f
o
r
co
n
d
u
ctin
g
m
al
icio
u
s
p
r
ed
ictio
n
s
.
I
n
a
m
o
b
ile
d
ev
ice,
a
s
tu
d
y
p
r
o
p
o
s
es
a
DL
al
g
o
r
ith
m
to
d
etec
t
m
alici
o
u
s
ac
ti
v
itie
s
b
y
co
n
s
tr
u
c
tin
g
a
m
u
ltis
tag
e
a
n
d
elast
ic
d
etec
t
io
n
f
r
a
m
e
w
o
r
k
[
2
5
]
.
R
ec
en
t
p
ap
er
s
ad
o
p
t
b
eh
av
io
r
tech
n
iq
u
e
f
o
r
m
alic
io
u
s
d
et
ec
tio
n
,
C
OM
P
A
,
an
ap
p
r
o
ac
h
p
r
o
p
o
s
e
a
w
a
y
to
d
etec
t
co
m
p
r
o
m
i
s
ed
ac
co
u
n
ts
b
eh
av
io
r
al
p
r
o
f
ile
an
d
an
o
m
al
y
d
etec
tio
n
[
2
6
]
,
an
d
a
s
tu
d
y
u
tili
ze
s
h
y
b
r
id
f
ea
t
u
r
es
to
d
etec
t
s
u
s
p
icio
u
s
ac
tiv
itie
s
[
2
7
]
.
T
h
e
n
ex
t
w
av
e
o
f
OSN
i
s
b
u
ild
i
n
g
OSN
a
d
ap
tiv
e
s
ec
u
r
it
y
b
y
u
tili
zi
n
g
lear
n
i
n
g
m
et
h
o
d
s
li
k
e
C
NN
an
d
R
NN
r
ath
er
t
h
an
u
s
e
a
tr
ad
itio
n
al
r
u
le
-
b
ased
o
r
s
t
atis
tical
a
n
al
y
s
i
s
.
6.
CO
NCLU
SI
O
N
T
h
e
cu
r
r
en
t
m
et
h
o
d
o
f
OSN
p
r
o
tectio
n
r
em
ai
n
s
s
h
o
r
tco
m
i
n
g
i
n
m
a
s
s
i
v
e
d
ata
an
al
y
s
i
s
an
d
m
an
u
a
l
f
ea
t
u
r
e
en
g
in
ee
r
i
n
g
.
Mo
r
eo
v
er
,
s
tan
d
ar
d
f
ea
tu
r
es
e
n
g
in
ee
r
i
n
g
i
s
ti
m
e
-
co
n
s
u
m
i
n
g
a
n
d
co
s
tl
y
.
T
o
d
ea
l
w
i
th
t
h
e
is
s
u
e,
m
o
d
er
n
OSN
n
ee
d
s
to
estab
lis
h
a
s
ca
lab
le
d
etec
tio
n
m
o
d
el
to
d
ea
l
w
i
th
m
o
d
er
n
OSN.
I
n
s
tead
o
f
u
s
i
n
g
a
co
n
v
e
n
tio
n
al
tech
n
iq
u
e,
w
e
b
r
in
g
t
h
e
D
L
ar
ch
itect
u
r
e
as
th
e
a
n
s
w
er
to
t
h
e
OSN
m
al
icio
u
s
p
r
o
b
le
m
b
y
an
a
l
y
z
in
g
d
iv
er
s
e
f
ea
t
u
r
es to
t
r
ain
th
e
cla
s
s
i
f
ier
.
I
n
th
i
s
p
ap
er
,
w
e
te
s
t
th
e
d
ee
p
lear
n
in
g
ap
p
r
o
ac
h
to
tr
ain
th
e
O
SN
d
ataset
f
o
r
m
alicio
u
s
class
i
f
icatio
n
.
W
e
g
ath
er
d
i
v
e
r
s
e
d
atasets
f
r
o
m
co
m
m
u
n
itie
s
o
r
o
r
g
an
iza
tio
n
s
to
co
n
s
tr
u
c
t
a
g
o
o
d
m
o
d
el
f
o
r
OSN
p
r
o
tect
io
n
,
in
cl
u
d
in
g
m
alicio
u
s
d
etec
tio
n
.
I
n
t
h
is
e
x
p
er
i
m
en
t,
w
e
tr
ain
t
h
e
d
ataset
b
y
e
s
tab
lis
h
i
n
g
th
e
s
u
p
er
v
i
s
ed
lear
n
i
n
g
clas
s
i
f
ier
ca
n
lear
n
u
s
er
ac
t
iv
i
t
y
p
att
er
n
s
ac
c
u
r
atel
y
.
B
ased
o
n
t
h
e
test
i
n
g
,
t
h
e
D
L
alg
o
r
ith
m
g
ai
n
s
a
n
ac
c
u
r
ate
r
esu
lt
to
cr
ea
te
in
telli
g
e
n
ce
p
r
o
tectio
n
in
OSN.
I
t
a
ls
o
ca
n
b
e
i
m
p
le
m
e
n
ted
to
o
p
tim
ize
lear
n
in
g
tec
h
n
iq
u
e
s
f
o
r
b
u
s
in
e
s
s
p
u
r
p
o
s
es.
T
h
e
d
ee
p
lear
n
in
g
is
a
r
ea
s
o
n
ab
l
e
an
s
w
er
to
tack
le
s
ca
lab
ilit
y
p
r
o
b
le
m
s
s
u
ch
a
s
g
r
o
w
in
g
u
s
er
s
a
n
d
d
ata
in
O
SN
.
T
h
e
n
ex
t
r
esear
ch
o
f
th
e
OS
N
s
ec
u
r
it
y
m
e
t
h
o
d
is
h
o
w
to
b
u
ild
in
telli
g
en
ce
a
n
d
r
ea
l
-
ti
m
e
p
r
o
tectio
n
s
c
h
e
m
e
th
an
a
s
tatis
tical
o
r
r
u
le
-
b
ased
m
o
d
el.
ACK
NO
WL
E
D
G
E
M
E
NT
S
T
h
is
p
ap
er
is
co
n
d
u
cted
i
n
t
h
e
I
n
s
t
itu
te
o
f
R
e
s
ea
r
ch
in
I
n
f
o
r
m
at
io
n
P
r
o
ce
s
s
i
n
g
L
ab
o
r
ato
r
y
,
Har
b
in
Un
i
v
er
s
it
y
o
f
Scie
n
ce
a
n
d
T
ec
h
n
o
lo
g
y
u
n
d
er
C
SC
Sc
h
o
lar
s
h
ip
.
RE
F
E
R
E
NC
E
S
[1
]
W
a
n
d
a
,
P
.
,
&
Jie
,
H.J.
Eff
icie
n
t
Da
ta
S
e
c
u
rit
y
f
o
r
M
o
b
il
e
In
sta
n
t
M
e
ss
e
n
g
e
r.
T
EL
KOM
NIKA
(
T
e
le
c
o
mm
u
n
ica
t
io
n
,
Co
mp
u
t
in
g
,
El
e
c
tro
n
ics
a
n
d
Co
n
t
ro
l)
J
o
u
r
n
a
l
,
V
o
l
.
1
6
(
3
).
2
0
1
8
.
[2
]
P
.
W
a
n
d
a
,
S
e
lo
a
n
d
B
.
S
.
Ha
n
t
o
n
o
"
Ef
fi
c
ien
t
me
ss
a
g
e
se
c
u
rity
b
a
se
d
Hy
p
e
rEll
ip
ti
c
C
u
rv
e
Cry
p
to
sy
ste
m
(
HECC)
fo
r
M
o
b
il
e
I
n
sta
n
t
M
e
ss
e
n
g
e
r,"
2
0
1
4
T
h
e
1
st
In
tern
a
ti
o
n
a
l
Co
n
f
e
re
n
c
e
o
n
I
n
f
o
rm
a
ti
o
n
T
e
c
h
n
o
l
o
g
y
,
Co
m
p
u
ter,
a
n
d
El
e
c
tri
c
a
l
En
g
in
e
e
rin
g
,
S
e
ma
r
a
n
g
,
p
p
.
2
4
5
-
2
4
9
.
2
0
1
4
.
[3
]
W
.
P
u
tra,
S
e
lo
a
n
d
B.
S
.
Ha
n
to
n
o
"
M
o
d
e
l
o
f
s
e
c
u
re
P
2
P
m
o
b
il
e
in
sta
n
t
m
e
ss
a
g
in
g
b
a
se
d
o
n
v
irt
u
a
l
n
e
tw
o
rk
,
"
2
0
1
4
In
ter
n
a
t
io
n
a
l
C
o
n
fer
e
n
c
e
o
n
I
n
fo
r
ma
ti
o
n
T
e
c
h
n
o
l
o
g
y
S
y
ste
ms
a
n
d
I
n
n
o
v
a
ti
o
n
(
ICIT
S
I)
,
Ba
n
d
u
n
g
,
p
p
.
8
1
-
8
5
.
2
0
1
4
[4
]
Am
a
n
y
A
.
Na
e
m
,
Ne
v
e
e
n
I.
G
h
a
li
.
Op
ti
m
izin
g
c
o
m
m
u
n
it
y
d
e
te
c
ti
o
n
in
so
c
ial
n
e
tw
o
rk
s
u
sin
g
a
n
tl
io
n
a
n
d
K
-
m
e
d
ian
,
Bu
ll
e
ti
n
o
f
El
e
c
trica
l
E
n
g
in
e
e
rin
g
a
n
d
In
f
o
rm
a
ti
c
s (
BE
EI)
,
Vo
l.
8
,
N
o
.
4
,
De
c
e
m
b
e
r
2
0
1
9
[5
]
M
.
M
o
h
a
m
m
a
d
re
z
a
e
i,
M
E
S
h
iri
,
A
.
M
Ra
h
m
a
n
i.
"
Id
e
n
ti
fy
in
g
F
a
k
e
Ac
c
o
u
n
ts
o
n
S
o
c
ial
Ne
tw
o
rk
s
Ba
se
d
o
n
G
ra
p
h
A
n
a
l
y
si
s a
n
d
Clas
sif
ica
ti
o
n
A
lg
o
r
it
h
m
s"
,
Hin
d
a
w
i
S
e
c
u
rit
y
a
n
d
Co
m
m
u
n
ica
ti
o
n
Ne
tw
o
rk
s,
2
0
1
8
[6
]
M
.
v
a
n
G
e
rv
e
n
a
n
d
S
.
M
.
B
o
h
te.
“
E
d
it
o
rial:
A
rti
f
icia
l
n
e
u
ra
l
n
e
tw
o
rk
s
a
s
m
o
d
e
ls
o
f
n
e
u
ra
l
in
f
o
rm
a
ti
o
n
p
ro
c
e
ss
in
g
,
”
F
ro
n
t.
C
o
m
p
u
t.
Ne
u
r
o
sc
i.
,
v
o
l.
1
1
,
n
o
.
1
1
4
,
De
c
.
2
0
1
7
.
[7
]
N.
F
.
F
a
d
z
a
il
,
S
.
M
a
t
Zali.
F
a
u
lt
d
e
tec
ti
o
n
a
n
d
c
las
sif
ic
a
ti
o
n
in
w
in
d
tu
rb
in
e
b
y
u
sin
g
a
rti
f
icia
l
n
e
u
ra
l
n
e
tw
o
rk
.
In
ter
n
a
t
io
n
a
l
J
o
u
rn
a
l
o
f
P
o
we
r E
lec
tro
n
ics
a
n
d
Dr
ive
S
y
ste
ms
(
IJ
P
EDS
).
V
o
l.
1
6
,
N
o
.
1
,
2
0
1
9
.
[8
]
P
a
ri
n
it
h
R
Iy
e
r,
S
h
ru
th
e
e
sh
Ra
m
a
n
Iy
e
r,
Ra
g
h
a
v
e
n
d
ra
n
Ra
m
e
sh
,
A
n
a
la
M
.
R.
,
K.N.
S
u
b
ra
m
a
n
y
a
,
A
d
a
p
ti
v
e
re
a
l
ti
m
e
tra
ff
ic
p
re
d
ictio
n
u
sin
g
d
e
e
p
n
e
u
ra
l
n
e
tw
o
rk
s
,
IAE
S
I
n
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
Arti
fi
c
ia
l
In
tell
ig
e
n
c
e
(
IJ
-
AI)
.
,
V
o
l
.
8
,
No
.
2
.
2
0
1
9
.
[9
]
W
a
n
d
a
,
P
u
tra
a
n
d
H
u
a
n
g
J.
Jie
.
“
URL
De
e
p
:
Co
n
ti
n
u
o
u
s
P
re
d
i
c
ti
o
n
o
f
M
a
li
c
io
u
s
URL
w
it
h
D
y
n
a
m
ic
De
e
p
L
e
a
rn
in
g
in
S
o
c
ial
Ne
tw
o
rk
s.”
I.
J
.
Ne
two
rk
S
e
c
u
rity
2
1
p
p
:
9
7
1
-
9
7
8
,
2
0
1
9
.
[1
0
]
S
k
y
m
in
d
,
“
Re
c
u
rre
n
t
n
e
two
rk
s.”
A
c
c
e
ss
e
d
o
n
:
No
v
.
2
0
,
2
0
1
8
.
[
On
li
n
e
]
.
Re
tri
e
v
e
d
:
h
tt
p
:
//
sk
y
m
i
n
d
.
a
i/
w
ik
i/
re
c
u
rre
n
t
-
n
e
tw
o
rk
-
rn
n
[1
1
]
B.
Hu
d
s
o
n
,
B.
R.
Vo
ter.
"
P
ro
f
il
e
c
h
a
ra
c
teristics
o
f
f
a
k
e
t
w
it
ter ac
c
o
u
n
ts"
,
Big
Da
ta &
S
o
c
iety
.
2
0
1
6
.
[1
2
]
N.
Kö
k
c
i
y
a
n
,
P
.
Yo
l
u
m
.
"
P
ro
G
u
a
rd
:
A
se
m
a
n
ti
c
a
p
p
ro
a
c
h
to
d
e
te
c
t
p
riv
a
c
y
v
io
latio
n
s
in
o
n
l
in
e
so
c
ial
n
e
tw
o
rk
s
"
,
IEE
E
T
ra
n
s
.
K
n
o
wl
.
Da
t
a
E
n
g
.
,
v
o
l.
2
8
,
n
o
.
1
0
,
p
p
.
2
7
2
4
-
2
7
3
7
,
Oc
t.
2
0
1
6
.
[1
3
]
E.
V
a
n
De
r
W
a
lt
a
n
d
J.
El
o
ff
,
"
Us
in
g
M
a
c
h
in
e
Lea
rn
in
g
to
De
t
e
c
t
F
a
k
e
Id
e
n
ti
ti
e
s:
Bo
ts
v
s
Hu
m
a
n
s
,
"
in
IEE
E
Acc
e
s
s,
v
o
l.
6
,
p
p
.
6
5
4
0
-
6
5
4
9
.
2
0
1
8
.
[1
4
]
L
iu
,
B.
-
H.,
Hs
u
Y.
-
P
.
,
a
n
d
Ke
W
.
-
C.
.
"
V
iru
s
i
n
f
e
c
ti
o
n
c
o
n
tr
o
l
i
n
o
n
li
n
e
s
o
c
ial
n
e
tw
o
rk
s
b
a
se
d
o
n
p
ro
b
a
b
il
isti
c
c
o
m
m
u
n
it
ies
.
"
In
t.
J.
Co
m
m
u
n
.
S
y
st.
(2
0
1
4
):
2
7
:4
4
8
1
–
4
4
9
1
.
2
0
1
4
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
I
n
t J
A
r
ti
f
I
n
tell
,
Vo
l.
9
,
No
.
1
,
Ma
r
ch
20
20
:
146
–
1
5
4
154
[1
5
]
J.
W
a
n
g
,
X
.
He
,
Q.
G
o
n
g
,
Y.
Ch
e
n
,
T
.
Wan
g
a
n
d
X
.
W
a
n
g
.
"
De
e
p
L
e
a
rn
in
g
-
B
a
se
d
M
a
li
c
io
u
s A
c
c
o
u
n
t
De
tec
ti
o
n
i
n
th
e
M
o
mo
S
o
c
i
a
l
Ne
t
wo
rk
,
"
2
0
1
8
2
7
t
h
I
n
tern
a
ti
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
ter
Co
m
m
u
n
ica
ti
o
n
a
n
d
Ne
tw
o
rk
s
(ICCCN),
Ha
n
g
z
h
o
u
,
p
p
.
1
-
2
.
2
0
1
8
.
[1
6
]
A
lb
e
rt
S
m
it
h
"
W
h
y
th
e
F
u
tu
re
o
f
S
o
c
ial
M
e
d
ia
W
il
l
De
p
e
n
d
o
n
A
rti
f
icia
l
In
telli
g
e
n
c
e
"
,
A
p
ril
1
3
,
2
0
1
8
,
Re
tri
e
v
e
d
h
tt
p
s:/
/www
.
s
m
a
rtd
a
tac
o
ll
e
c
ti
v
e
.
c
o
m
/
f
u
tu
re
-
so
c
ial
-
m
e
d
ia
-
d
e
p
e
n
d
-
a
rti
f
icia
l
-
in
telli
g
e
n
c
e
/
[1
7
]
P
.
Bu
rn
a
p
,
O.
F
.
Ra
n
a
,
N.
A
v
is,
M
.
W
il
li
a
m
s,
W
.
Ho
u
sle
y
,
A
.
Ed
w
a
rd
s,
J.
M
o
rg
a
n
,
a
n
d
L
.
S
lo
a
n
.
De
tec
ti
n
g
ten
sio
n
in
o
n
li
n
e
c
o
m
m
u
n
it
ies
w
it
h
c
o
m
p
u
tati
o
n
a
l
tw
it
ter
a
n
a
l
y
sis.
T
e
c
h
n
o
lo
g
ica
l
F
o
re
c
a
stin
g
a
n
d
S
o
c
ial
Ch
a
n
g
e
,
9
5
:9
6
–
1
0
8
,
Ju
n
e
.
2
0
1
5
[1
8
]
S
a
v
it
a
S
a
n
g
a
m
,
S
u
b
h
a
sh
S
h
in
d
e
.
S
e
n
ti
m
e
n
t
c
las
sif
ica
ti
o
n
o
f
so
c
ial
m
e
d
ia
re
v
i
e
w
s
u
sin
g
a
n
e
n
se
m
b
le
c
las
si
f
ier.
In
d
o
n
e
sia
n
J
o
u
rn
a
l
o
f
El
e
c
trica
l
En
g
i
n
e
e
rin
g
a
n
d
C
o
mp
u
ter
S
c
ien
c
e
(
I
J
EE
CS
).
V
o
l
.
1
6
,
N
o
.
1
.
2
0
1
9
[1
9
]
W
a
n
d
a
,
P
.
,
&
Jin
-
ji
e
,
H.
M
o
d
e
l
o
f
S
e
n
ti
m
e
n
t
A
n
a
l
y
sis
w
it
h
De
e
p
L
e
a
rn
in
g
in
S
o
c
ial
Ne
t
w
o
rk
En
v
iro
n
m
e
n
t.
2
0
1
9
IEE
E
2
n
d
In
ter
n
a
ti
o
n
a
l
C
o
n
fer
e
n
c
e
o
n
El
e
c
tro
n
ic
I
n
f
o
rm
a
ti
o
n
a
n
d
C
o
mm
u
n
ica
ti
o
n
T
e
c
h
n
o
lo
g
y
(
ICEICT
)
,
6
2
5
-
6
3
0
,
2
0
1
9
.
[2
0
]
P
u
tra.
W
,
a
n
d
H.
Jin
J.
De
e
p
S
e
n
ti
m
e
n
t:
F
in
d
in
g
M
a
li
c
io
u
s
S
e
n
ti
m
e
n
t
in
On
li
n
e
S
o
c
ial
Ne
tw
o
rk
b
a
se
d
o
n
Dy
n
a
m
i
c
De
e
p
L
e
a
rn
in
g
.
IAE
NG In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
Co
m
p
u
ter
S
c
ie
n
c
e
,
Vo
l.
4
6
,
N
o
.
4
,
p
p
:
6
1
6
-
6
2
7
,
2
0
1
9
[2
1
]
F
.
Ch
e
n
,
R.
Ji,
J.
S
u
,
D.
Ca
o
,
a
n
d
Y.
Ga
o
.
"
P
re
d
ictin
g
M
icro
b
lo
g
S
e
n
ti
m
e
n
ts
v
ia
W
e
a
k
l
y
S
u
p
e
rv
ise
d
M
u
lt
im
o
d
a
l
De
e
p
L
e
a
rn
in
g
,
"
in
IEE
E
T
ra
n
sa
c
ti
o
n
s
o
n
M
u
lt
ime
d
i
a
,
v
o
l
.
2
0
,
n
o
.
4
,
p
p
.
9
9
7
-
1
0
0
7
.
2
0
1
8
[2
2
]
Da
t
T
i
e
n
Ng
u
y
e
n
,
S
h
a
f
iq
R.
Jo
t
y
,
M
u
h
a
m
m
a
d
I
m
ra
n
,
Ha
ss
a
n
S
a
jj
a
d
,
a
n
d
P
ra
se
n
ji
t
M
i
tra.
2
0
1
6
.
A
p
p
li
c
a
ti
o
n
s
o
f
o
n
li
n
e
d
e
e
p
lea
rn
i
n
g
f
o
r
c
risis
re
sp
o
n
se
u
si
n
g
so
c
ial
m
e
d
ia
in
f
o
rm
a
ti
o
n
.
Co
RRa
b
s/1
6
1
0
.
0
1
0
3
0
(
2
0
1
6
).
Re
tri
e
v
e
d
f
ro
m
h
tt
p
:/
/arx
iv
.
o
rg
/ab
s/1
6
1
0
.
0
1
0
30
.
[2
3
]
T
.
T
a
k
a
h
a
sh
i,
R.
T
o
m
io
k
a
,
a
n
d
K.
Ya
m
a
n
ish
i.
Disc
o
v
e
rin
g
e
m
e
r
g
in
g
to
p
ics
in
so
c
ial
stre
a
m
s
v
ia
li
n
k
a
n
o
m
a
l
y
d
e
tec
ti
o
n
.
IEE
E
T
ra
n
sa
c
ti
o
n
s o
n
Kn
o
wled
g
e
a
n
d
D
a
ta
En
g
i
n
e
e
rin
g
,
2
6
(
1
):1
2
0
–
1
3
0
,
De
c
e
m
b
e
r
2
0
1
4
.
[2
4
]
W
a
n
d
a
,
P
.
,
Hu
a
n
g
J.
Jie
.
A
S
u
rv
e
y
o
f
In
tru
sio
n
De
te
c
ti
o
n
S
y
s
tem
,
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
I
n
fo
rm
a
t
ics
a
n
d
Co
mp
u
t
a
ti
o
n
(
IJ
ICOM
)
,
V
o
l.
1
,
No
.
1
,
A
u
g
u
st,
2
0
1
9
[2
5
]
B.
F
e
n
g
,
Q.
F
u
,
M
.
Do
n
g
,
D.
G
u
o
a
n
d
Q.
L
i.
"
M
u
lt
istag
e
a
n
d
El
a
stic
S
p
a
m
De
t
e
c
ti
o
n
in
M
o
b
il
e
S
o
c
ial
Ne
t
w
o
rk
s
th
ro
u
g
h
De
e
p
L
e
a
rn
in
g
,
"
i
n
IEE
E
Ne
two
rk
,
v
o
l
.
3
2
,
n
o
.
4
,
p
p
.
1
5
-
2
1
.
2
0
1
8
[2
6
]
M
.
Eg
e
le,
G
.
S
tri
n
g
h
in
i
,
C.
Kru
e
g
e
l,
a
n
d
G
.
V
ig
n
a
.
T
o
w
a
rd
s
d
e
te
c
ti
n
g
c
o
m
p
ro
m
is
e
d
a
c
c
o
u
n
ts
o
n
so
c
ial
n
e
tw
o
rk
s.
IEE
E
T
ra
n
sa
c
ti
o
n
s
o
n
De
p
e
n
d
a
b
l
e
a
n
d
S
e
c
u
re
Co
m
p
u
ti
n
g
,
1
4
(4
):4
4
7
–
4
6
0
,
J
u
ly
.
2
0
1
7
.
[2
7
]
Ku
ra
p
a
ti
S
u
b
b
a
Re
d
d
y
,
E.
S
rin
i
v
a
sa
R
e
d
d
y
.
In
teg
r
a
ted
a
p
p
ro
a
c
h
to
d
e
tec
t
sp
a
m
in
so
c
ial
m
e
d
ia
n
e
tw
o
rk
s
u
sin
g
h
y
b
rid
f
e
a
tu
re
s
.
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
El
e
c
trica
l
a
n
d
C
o
mp
u
ter
En
g
i
n
e
e
rin
g
(
IJ
ECE
),
V
o
l.
9
,
No
.
1
.
2
0
1
9
.
B
I
O
G
RAP
H
I
E
S
O
F
AUTH
O
RS
P
u
tra
W
a
n
d
a
,
re
c
e
iv
e
d
B.
En
g
i
n
In
f
o
rm
a
ti
c
En
g
in
e
e
rin
g
in
2
0
1
1
.
He
g
ra
d
u
a
ted
w
it
h
M
.
E
n
g
.
De
g
re
e
s
in
In
f
o
rm
a
ti
o
n
T
e
c
h
n
o
lo
g
y
f
ro
m
Ga
d
jah
M
a
d
a
Un
iv
e
rsity
2
0
1
5
.
S
i
n
c
e
A
u
g
u
st
2
0
1
6
,
h
e
is
w
it
h
th
e
S
c
h
o
o
l
o
f
Co
m
p
u
ter
S
c
ien
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
f
ro
m
Ha
rb
in
Un
iv
e
rsity
o
f
S
c
ien
c
e
a
n
d
T
e
c
h
n
o
lo
g
y
a
s
a
P
h
.
D.
c
a
n
d
i
d
a
te.
He
h
a
s
p
u
b
li
sh
e
d
se
v
e
ra
l
p
a
p
e
rs
in
th
e
n
e
tw
o
rk
a
n
d
in
f
o
rm
a
ti
o
n
se
c
u
rit
y
b
o
th
i
n
c
o
n
f
e
re
n
c
e
s
a
n
d
jo
u
rn
a
ls
i
n
d
e
x
e
d
b
y
S
c
o
p
u
s
a
n
d
EI
C
o
m
p
e
n
d
e
x
.
h
tt
p
s:/
/o
rc
id
.
o
rg
/0
0
0
0
-
0
0
0
3
-
0
1
3
0
-
3
1
9
6
M
a
rse
li
n
a
En
d
a
h
H,
re
c
e
iv
e
d
B.
En
g
i
n
I
n
f
o
rm
a
ti
c
En
g
in
e
e
rin
g
.
S
h
e
g
ra
d
u
a
ted
f
ro
m
M
.
Cs.
De
g
re
e
s
in
Co
m
p
u
ter
S
c
ien
c
e
f
r
o
m
G
a
d
jah
M
a
d
a
Un
iv
e
rsit
y
,
In
d
o
n
e
sia
.
S
h
e
is
a
lec
tu
re
r
in
th
e
De
p
a
rtme
n
t
o
f
In
f
o
r
m
a
ti
c
s,
th
e
Un
iv
e
rsit
y
o
f
Re
sp
a
ti
Yo
g
y
a
k
a
rta.
S
h
e
h
a
s
p
u
b
l
ish
e
d
se
v
e
ra
l
p
a
p
e
rs i
n
in
f
o
rm
a
ti
o
n
tec
h
n
o
lo
g
y
a
re
a
s b
o
th
i
n
c
o
n
f
e
re
n
c
e
s a
n
d
jo
u
rn
a
ls i
n
d
e
x
e
d
b
y
S
c
o
p
u
s.
Hu
a
n
g
J.
Jie
is
a
P
ro
f
e
ss
o
r
a
n
d
P
h
.
D.
S
u
p
e
rv
iso
r
i
n
t
h
e
S
c
h
o
o
l
o
f
S
c
ien
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
,
Ha
rb
in
Un
iv
e
rsit
y
o
f
S
c
ien
c
e
a
n
d
T
e
c
h
n
o
lo
g
y
,
Ch
in
a
.
His
c
u
rre
n
t
re
se
a
rc
h
in
tere
sts
in
c
lu
d
e
De
e
p
L
e
a
rn
in
g
,
Ro
b
o
ti
c
a
n
d
A
u
to
m
a
t
io
n
P
u
b
li
sh
m
a
n
y
a
rti
c
les
in
t
h
e
ro
b
o
ti
c
sy
ste
m
a
n
d
p
a
tt
e
rn
re
c
o
g
n
it
io
n
in
d
e
x
e
d
b
y
S
CI
a
n
d
E
I
Co
m
p
e
n
d
e
x
.
h
tt
p
s://
o
rc
i
d
.
o
rg
/0
0
0
0
-
0
0
0
2
-
2
1
0
7
-
2
6
9
0
Evaluation Warning : The document was created with Spire.PDF for Python.