Co
m
pu
t
er
Science
a
nd
I
nfo
r
m
a
t
io
n T
ec
hn
o
lo
g
ies
Vo
l.
6
,
No
.
2
,
J
u
ly
2
0
2
5
,
p
p
.
20
2
~
2
1
3
I
SS
N:
2722
-
3
2
2
1
,
DOI
:
1
0
.
1
1
5
9
1
/csi
t
.
v6
i
2
.
p
p
2
0
2
-
2
1
3
202
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ia
e
s
p
r
ime.
co
m/in
d
ex
.
p
h
p
/csi
t
Att
a
ck
dete
ction i
n
inter
net
of t
hin
g
s
networks
w
ith
deep
lea
rning
using
de
ep t
ra
nsf
er learni
ng
method
Rik
i A
bd
illa
h H
a
s
a
nu
dd
in
1
,
M
uh
a
m
m
a
d Sub
a
li
2
1
D
e
p
a
r
t
me
n
t
E
l
e
c
t
r
i
c
a
l
E
n
g
i
n
e
e
r
i
n
g
,
F
a
c
u
l
t
y
o
f
T
e
c
h
n
o
l
o
g
y
a
n
d
E
n
g
i
n
e
e
r
i
n
g
,
G
u
n
a
d
a
r
ma
U
n
i
v
e
r
si
t
y
,
Ja
k
a
r
t
a
,
I
n
d
o
n
e
si
a
2
F
a
c
u
l
t
y
o
f
I
n
f
o
r
m
a
t
i
c
s
En
g
i
n
e
e
r
i
n
g
,
C
e
n
d
e
k
i
a
A
b
d
i
t
a
m
a
U
n
i
v
e
r
si
t
y
,
T
a
n
g
e
r
a
n
g
,
I
n
d
o
n
e
si
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Dec
15
,
2
0
2
4
R
ev
is
ed
Ap
r
22
,
2
0
2
5
Acc
ep
ted
Ap
r
29
,
2
0
2
5
Cy
b
e
rse
c
u
rit
y
b
e
c
o
m
e
s
a
c
r
u
c
ial
p
a
rt
with
i
n
th
e
in
f
o
rm
a
ti
o
n
m
a
n
a
g
e
m
e
n
t
fra
m
e
wo
rk
o
f
in
tern
e
t
o
f
th
i
n
g
s
(Io
T)
d
e
v
ice
n
e
two
r
k
s.
Th
e
l
a
rg
e
-
sc
a
le
d
istri
b
u
ti
o
n
o
f
Io
T
n
e
two
r
k
s
a
n
d
t
h
e
c
o
m
p
lex
it
y
o
f
c
o
m
m
u
n
ica
ti
o
n
p
ro
t
o
c
o
ls
u
se
d
a
re
c
o
n
tri
b
u
ti
n
g
fa
c
t
o
rs
t
o
t
h
e
wi
d
e
sp
re
a
d
v
u
ln
e
ra
b
il
i
ti
e
s
o
f
Io
T
d
e
v
ice
s.
Th
e
imp
lem
e
n
tatio
n
o
f
tran
sfe
r
lea
rn
in
g
m
o
d
e
ls
i
n
d
e
e
p
lea
rn
in
g
c
a
n
a
c
h
iev
e
o
p
ti
m
a
l
p
e
rf
o
rm
a
n
c
e
fa
ste
r
th
a
n
trad
it
io
n
a
l
m
a
c
h
in
e
lea
rn
in
g
m
o
d
e
ls,
a
s
th
e
y
lev
e
ra
g
e
k
n
o
w
led
g
e
fro
m
p
re
v
i
o
u
s
m
o
d
e
ls
th
a
t
a
lrea
d
y
u
n
d
e
rsta
n
d
th
e
se
fe
a
tu
re
s.
Ba
se
m
o
d
e
l
wa
s
b
u
il
t
u
si
n
g
t
h
e
1
-
d
ime
n
sio
n
c
o
n
v
o
lu
ti
o
n
a
l
n
e
u
ra
l
n
e
two
r
k
(1
D
-
CNN
)
m
e
th
o
d
,
u
sin
g
trai
n
in
g
a
n
d
tes
t
d
a
ta
fro
m
th
e
so
u
rc
e
d
o
m
a
in
d
a
tas
e
t.
M
o
d
e
l
1
wa
s
c
o
n
stru
c
ted
u
sin
g
th
e
sa
m
e
m
e
th
o
d
a
s
b
a
se
m
o
d
e
l.
Th
e
tes
t
a
n
d
trai
n
in
g
d
a
ta
u
se
d
fo
r
m
o
d
e
l
1
we
re
fro
m
th
e
targ
e
t
d
o
m
a
in
d
a
tas
e
t.
Th
is
m
o
d
e
l
su
c
c
e
ss
fu
ll
y
d
e
tec
ted
k
n
o
wn
a
tt
a
c
k
s
a
t
a
ra
t
e
o
f
9
9
.
3
5
2
%
,
b
u
t
d
id
n
o
t
p
e
rf
o
rm
we
ll
in
d
e
tec
ti
n
g
u
n
k
n
o
w
n
a
tt
a
c
k
s,
with
a
n
a
c
c
u
ra
c
y
o
f
8
4
.
6
4
5
%
.
M
o
d
e
l
2
is
a
n
e
n
h
a
n
c
e
m
e
n
t
o
f
m
o
d
e
l
1
,
i
n
c
o
rp
o
ra
ti
n
g
tra
n
sfe
r
lea
rn
in
g
fro
m
t
h
e
b
a
se
m
o
d
e
l.
I
ts
re
su
lt
s
sig
n
ifi
c
a
n
t
ly
imp
r
o
v
e
d
c
o
m
p
a
re
d
t
o
m
o
d
e
l
1
te
sti
n
g
.
M
o
d
e
l
2
h
a
s
a
n
a
c
c
u
ra
c
y
a
n
d
p
re
c
isio
n
ra
te
o
f
9
8
.
8
6
%
a
n
d
9
9
.
1
7
%
,
re
sp
e
c
ti
v
e
l
y
,
a
ll
o
win
g
it
to
d
e
tec
t
p
re
v
io
u
sly
u
n
k
n
o
wn
a
tt
a
c
k
s.
Ev
e
n
wit
h
a
slig
h
t
d
e
c
re
a
se
in
n
o
rm
a
l
d
e
tec
ti
o
n
,
m
o
st attac
k
s c
a
n
stil
l
b
e
d
e
tec
ted
.
K
ey
w
o
r
d
s
:
1D
-
C
NN
Attack
d
etec
tio
n
Dee
p
lear
n
in
g
I
o
T
n
etwo
r
k
T
r
an
s
f
er
lear
n
i
n
g
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
:
Mu
h
am
m
ad
Su
b
ali
Dep
ar
tm
en
t o
f
C
o
m
p
u
ter
Scie
n
ce
an
d
I
n
f
o
r
m
atio
n
T
ec
h
n
o
lo
g
y
,
C
en
d
ek
ia
A
b
d
itam
a
Un
iv
e
r
s
ity
I
s
lam
ic
R
ay
a
St.
,
Kela
p
a
Du
a,
T
an
g
er
a
n
g
,
B
an
ten
,
I
n
d
o
n
esia
E
m
ail:
s
u
b
ali@
u
ca
.
ac
.
id
1.
I
NT
RO
D
UCT
I
O
N
T
h
e
d
e
v
elo
p
m
e
n
t
o
f
in
f
o
r
m
atio
n
tech
n
o
lo
g
y
,
p
a
r
ticu
lar
l
y
in
ter
n
et
o
f
th
in
g
s
(
I
o
T
)
d
ev
ices,
is
p
r
o
g
r
ess
in
g
r
ap
id
ly
a
n
d
is
d
ir
e
ctly
p
r
o
p
o
r
tio
n
al
to
th
e
in
cr
ea
s
in
g
n
u
m
b
er
o
f
I
o
T
d
ev
ice
u
s
er
s
,
b
o
th
in
d
i
v
id
u
als
an
d
in
d
u
s
tr
ies.
T
h
e
r
o
le
o
f
I
o
T
d
ev
ices
h
as
i
n
f
lu
en
ce
d
all
a
r
ea
s
d
ir
ec
tly
r
elate
d
to
tech
n
o
lo
g
y
a
n
d
b
u
s
in
ess
,
en
h
an
cin
g
b
en
ef
its
f
o
r
b
o
th
in
d
iv
id
u
als
an
d
o
r
g
a
n
izatio
n
s
[
1
]
.
I
o
T
d
ev
ices
en
h
a
n
ce
u
s
er
ex
p
er
ien
ce
s
b
y
p
r
o
v
id
i
n
g
im
m
e
d
iate
d
ata
ac
c
ess
y
et
in
tr
o
d
u
ce
a
m
u
ltit
u
d
e
o
f
cy
b
e
r
s
ec
u
r
ity
v
u
ln
er
ab
ilit
i
es
ac
r
o
s
s
d
if
f
er
en
t
o
p
er
atio
n
al
la
y
er
s
.
C
y
b
er
-
atta
ck
s
o
n
th
ese
d
ev
i
ce
s
ca
n
b
e
ca
teg
o
r
ized
as
g
o
al
-
o
r
ien
ted
attac
k
s
(
tar
g
etin
g
o
b
jectiv
es
lik
e
u
n
au
th
o
r
ized
ac
ce
s
s
o
r
d
ata
ex
f
iltra
tio
n
)
,
p
er
f
o
r
m
a
n
ce
-
o
r
ie
n
ted
attac
k
s
(
s
u
ch
as
Do
S/DDo
S
ef
f
o
r
ts
th
at
d
eg
r
a
d
e
s
y
s
tem
av
ailab
ilit
y
)
,
an
d
lay
er
-
o
r
ien
t
ed
attac
k
s
(
ex
p
lo
itin
g
wea
k
n
ess
e
s
at
th
e
ed
g
e,
ac
ce
s
s
/m
id
d
lewa
r
e,
o
r
ap
p
licatio
n
lay
er
s
)
.
Ad
d
r
ess
in
g
th
ese
th
r
ea
ts
r
eq
u
ir
es
tailo
r
ed
s
ec
u
r
ity
s
tr
ateg
ies
—
f
r
o
m
f
o
r
tifie
d
clo
u
d
s
er
v
ices
an
d
r
esil
ien
t
5
G
n
et
wo
r
k
d
esig
n
s
to
p
r
o
tectio
n
s
th
at
ac
co
u
n
t
f
o
r
I
o
T
d
ev
ices’
h
eter
o
g
e
n
eo
u
s
n
atu
r
e
an
d
l
im
ited
co
m
p
u
tin
g
r
es
o
u
r
ce
s
.
[
2
]
.
T
h
e
ap
p
licatio
n
o
f
tr
an
s
f
er
lear
n
in
g
m
eth
o
d
s
in
d
ee
p
lear
n
in
g
h
a
s
b
ee
n
wid
ely
u
s
ed
to
tr
ain
m
o
d
els
an
d
is
ef
f
ec
tiv
e
in
id
en
tify
in
g
attac
k
s
.
T
r
an
s
f
er
lear
n
i
n
g
m
o
d
els
ca
n
ac
h
iev
e
o
p
tim
al
p
er
f
o
r
m
an
ce
m
o
r
e
q
u
ick
ly
t
h
an
tr
a
d
itio
n
a
l
m
ac
h
in
e
lear
n
i
n
g
Evaluation Warning : The document was created with Spire.PDF for Python.
C
o
m
p
u
t Sci
I
n
f
T
ec
h
n
o
l
I
SS
N:
2722
-
3
2
2
1
A
tta
ck
d
etec
tio
n
in
i
n
tern
et
o
f
th
in
g
s
n
etw
o
r
ks w
ith
d
ee
p
lea
r
n
in
g
u
s
in
g
…
(
R
iki
A
b
d
illa
h
H
a
s
a
n
u
d
d
in
)
203
m
o
d
els
b
ec
au
s
e
th
ese
m
o
d
els
lev
er
ag
e
k
n
o
wled
g
e
(
f
e
atu
r
es
an
d
weig
h
ts
)
f
r
o
m
p
r
ev
i
o
u
s
m
o
d
els
th
at
alr
ea
d
y
u
n
d
er
s
tan
d
th
ese
f
ea
tu
r
es,
m
a
k
in
g
it
f
aster
th
an
tr
ain
in
g
a
n
eu
r
al
n
etwo
r
k
f
r
o
m
s
cr
atch
.
A
d
d
itio
n
ally
,
tr
an
s
f
e
r
lear
n
in
g
is
m
o
r
e
co
m
p
u
tatio
n
a
lly
ef
f
icien
t a
n
d
h
elp
s
ac
h
iev
e
b
etter
r
esu
lts
u
s
in
g
a
s
m
all
d
at
aset
[
3
]
.
C
lu
s
ter
in
g
-
en
h
an
ce
d
tr
an
s
f
er
l
ea
r
n
in
g
a
p
p
r
o
ac
h
(
C
eHT
L
)
,
a
n
en
h
a
n
ce
d
a
p
p
r
o
ac
h
th
at
au
t
o
m
atica
lly
d
eter
m
in
es
th
e
r
elatio
n
s
h
ip
b
e
twee
n
n
ew
an
d
k
n
o
wn
attac
k
s
[
4
]
.
An
i
m
p
r
o
v
ed
co
n
v
o
l
u
tio
n
al
n
eu
r
al
n
etwo
r
k
(
I
C
NN)
ch
ar
ac
ter
izes
an
d
p
r
ep
r
o
ce
s
s
es
n
etw
o
r
k
tr
af
f
ic
d
ata,
ex
tr
ac
ts
ad
v
an
ce
d
f
ea
tu
r
es,
an
d
o
p
tim
izes
p
ar
am
eter
s
with
s
to
ch
asti
c
g
r
a
d
ien
t
d
escen
t
[
5
]
.
D
ee
p
tr
a
n
s
f
er
lear
n
in
g
m
eth
o
d
th
at
u
s
es
t
wo
au
to
e
n
co
d
e
r
s
to
alig
n
f
ea
tu
r
e
r
e
p
r
esen
tatio
n
s
a
n
d
ef
f
ec
tiv
ely
d
etec
t
I
o
T
attac
k
s
,
o
u
tp
er
f
o
r
m
in
g
o
th
e
r
ap
p
r
o
ac
h
es
o
n
n
in
e
I
o
T
d
atasets
[
6
]
. I
n
tr
u
s
io
n
d
etec
tio
n
m
o
d
el
th
at
u
s
es c
o
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
s
in
1
D,
2
D,
an
d
3
D,
alo
n
g
with
tr
an
s
f
er
lear
n
in
g
to
h
an
d
le
b
in
ar
y
an
d
m
u
lticlas
s
class
i
f
icatio
n
[
7
]
.
E
f
f
icien
t
n
etwo
r
k
-
b
ased
in
tr
u
s
io
n
d
etec
tio
n
s
y
s
tem
f
o
r
I
o
T
n
et
wo
r
k
s
th
at
co
m
b
in
es
a
d
ee
p
n
eu
r
al
n
etwo
r
k
(
D
NN)
with
m
u
tu
al
in
f
o
r
m
atio
n
(
MI
)
-
b
ased
f
ea
tu
r
e
s
elec
tio
n
to
d
etec
t
an
o
m
alies
an
d
ze
r
o
-
d
ay
cy
b
er
attac
k
s
.
[
8
]
.
T
r
a
n
s
f
er
lear
n
in
g
ap
p
r
o
ac
h
to
u
p
d
atin
g
in
tr
u
s
io
n
d
etec
tio
n
s
y
s
tem
s
(
I
DS)
th
at
h
av
e
b
ec
o
m
e
o
u
td
ated
d
u
e
to
th
ei
r
h
ea
v
y
r
elian
ce
o
n
in
itial
tr
ain
in
g
d
atasets
an
d
th
eir
in
a
b
ilit
y
to
d
etec
t
ch
an
g
es
in
attac
k
s
[
9
]
.
D
ee
p
tr
an
s
f
er
lear
n
i
n
g
f
r
am
ewo
r
k
u
s
in
g
weig
h
t
tr
an
s
f
er
r
in
g
an
d
n
eu
r
al
n
etwo
r
k
f
in
e
-
tu
n
in
g
f
o
r
e
n
d
-
to
-
e
n
d
lear
n
in
g
,
a
d
d
r
ess
in
g
co
n
ce
p
t
d
r
if
t
an
d
r
ed
u
cin
g
h
u
m
an
in
ter
v
en
tio
n
[
1
0
]
.
T
h
e
ex
p
e
r
im
en
ts
s
h
o
wed
th
at
th
e
in
d
iv
id
u
al
DNN,
R
NN,
o
r
C
NN
ap
p
r
o
ac
h
es
ar
e
b
ette
r
th
a
n
t
h
e
co
m
b
in
e
d
m
o
d
els
(
C
NN+
R
NN
an
d
C
NN+
L
STM
)
[
1
1
]
.
A
tr
an
s
f
er
lear
n
in
g
f
r
am
ewo
r
k
u
s
in
g
an
o
p
tim
a
l
s
o
u
r
ce
d
o
m
ai
n
d
ataset
im
p
r
o
v
es
n
etwo
r
k
-
lev
el
in
tr
u
s
io
n
d
etec
tio
n
[
1
2
]
.
F
ed
er
ated
tr
an
s
f
er
lear
n
in
g
(
FTL
)
f
r
a
m
ewo
r
k
f
o
r
I
I
o
T
n
etwo
r
k
i
n
tr
u
s
io
n
d
etec
tio
n
u
s
es
a
n
eu
r
al
n
etwo
r
k
th
at
d
is
tr
ib
u
tes
I
o
T
d
ata
p
r
o
ce
s
s
in
g
b
etwe
en
clien
t
an
d
s
er
v
er
d
ev
ices
[
1
3
]
.
D
e
e
p
tr
a
n
s
f
er
lea
r
n
in
g
ap
p
r
o
ac
h
f
o
r
r
o
llin
g
b
ea
r
in
g
f
a
u
lt d
iag
n
o
s
is
u
s
in
g
a
1D
-
C
NN
e
x
tr
ac
ts
f
ea
tu
r
es f
r
o
m
v
ib
r
atio
n
s
ig
n
als an
d
u
s
es C
OR
r
elatio
n
AL
ig
n
m
en
t
(
C
OR
AL
)
to
m
in
im
ize
th
e
m
ar
g
in
al
d
is
tr
ib
u
tio
n
d
is
cr
ep
an
cy
b
etwe
en
s
o
u
r
ce
an
d
tar
g
e
t
d
o
m
ain
s
[
1
4
]
.
T
h
e
s
u
itab
ilit
y
o
f
d
ee
p
lear
n
i
n
g
f
o
r
an
o
m
al
y
-
b
ased
I
DS
b
y
d
e
v
elo
p
in
g
m
o
d
els
u
s
in
g
v
ar
io
u
s
d
ee
p
n
eu
r
al
n
etwo
r
k
ar
c
h
itectu
r
es
in
clu
d
i
n
g
C
NNs,
AE
s
,
a
n
d
R
NNs
[
1
5
]
.
T
r
an
s
f
er
lear
n
in
g
–
b
ased
I
DS
f
o
r
clo
u
d
-
b
ased
I
o
T
en
v
ir
o
n
m
e
n
ts
,
ad
d
r
ess
in
g
t
h
e
in
cr
ea
s
ed
s
ec
u
r
ity
r
is
k
s
in
h
e
r
en
t
in
ce
n
tr
ali
ze
d
d
ata
p
r
o
ce
s
s
in
g
[
1
6
]
.
U
n
if
ied
i
n
d
o
o
r
–
o
u
td
o
o
r
lo
ca
lizatio
n
s
o
lu
tio
n
f
o
r
I
o
T
d
ev
ices
in
s
m
ar
t
cities
u
s
in
g
an
e
n
co
d
e
r
-
b
ased
tr
an
s
f
er
lear
n
in
g
s
ch
em
e
,
th
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
b
u
ild
s
a
s
in
g
le
d
ee
p
lear
n
in
g
m
o
d
el
th
at
ad
ap
ts
ac
r
o
s
s
b
o
th
in
d
o
o
r
a
n
d
o
u
t
d
o
o
r
s
ettin
g
s
,
r
ed
u
cin
g
co
m
p
lex
ity
an
d
co
s
ts
[
1
7
]
.
S
eq
u
e
n
tial
in
tr
u
s
io
n
d
et
ec
tio
n
s
y
s
tem
th
at
lev
er
ag
es
d
ee
p
lear
n
in
g
tech
n
iq
u
es
s
p
ec
if
ically
,
T
ex
t
-
C
NN
an
d
GR
U
to
ex
tr
ac
t
f
ea
t
u
r
es
f
r
o
m
th
e
n
etwo
r
k
lay
er
an
d
t
h
e
a
p
p
licatio
n
lay
er
,
tr
ea
tin
g
s
eq
u
en
tial
d
ata
li
k
e
a
lan
g
u
ag
e
m
o
d
el.
[
1
8
]
.
I
n
r
ela
tio
n
to
th
e
s
tu
d
ies,
we
p
r
o
p
o
s
ed
m
o
d
el
1D
-
C
NN
ar
ch
itectu
r
e
in
U
b
u
n
tu
en
v
ir
o
n
m
en
t.
T
h
e
m
ain
p
u
r
p
o
s
e
o
f
th
is
r
esear
ch
is
to
p
r
o
d
u
ce
a
m
o
d
el
u
s
in
g
d
ee
p
tr
an
s
f
er
lear
n
in
g
m
eth
o
d
s
th
at
is
tr
ain
ed
with
a
s
o
u
r
ce
d
o
m
ain
d
ataset,
wh
er
e
th
e
m
o
d
el
ca
n
d
etec
t
b
o
th
k
n
o
wn
an
d
u
n
k
n
o
wn
attac
k
s
in
th
e
t
ar
g
et
d
o
m
ain
d
ataset
with
s
m
all
d
ataset.
I
n
o
th
er
r
esear
ch
,
two
-
d
im
e
n
s
io
n
al
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
(
C
NN2
D)
ar
ch
itectu
r
e
is
u
s
ed
to
b
u
ild
m
o
d
el
in
W
in
d
o
ws en
v
ir
o
n
m
e
n
t
[
1
9
]
.
2.
M
E
T
H
O
D
2
.
1
.
B
a
c
k
g
ro
un
d t
heo
ry
T
r
an
s
f
er
lear
n
in
g
is
an
im
p
o
r
tan
t
to
o
l
in
m
ac
h
i
n
e
lear
n
in
g
to
ad
d
r
ess
th
e
f
u
n
d
am
e
n
tal
p
r
o
b
l
em
o
f
in
s
u
f
f
icien
t
tr
ain
in
g
d
ata.
I
t
attem
p
ts
to
tr
an
s
f
er
k
n
o
wled
g
e
f
r
o
m
t
h
e
s
o
u
r
ce
d
o
m
ain
to
th
e
tar
g
et
d
o
m
ain
b
y
r
elax
in
g
th
e
ass
u
m
p
tio
n
th
at
tr
ain
in
g
an
d
test
in
g
d
ata
m
u
s
t
b
e
in
teg
r
ated
,
id
en
tically
d
is
tr
ib
u
ted
(
I
I
D)
.
T
h
is
will
lead
to
s
ig
n
if
ican
t
p
o
s
i
tiv
e
ef
f
ec
ts
f
o
r
m
a
n
y
d
o
m
a
in
s
th
at
ar
e
d
if
f
icu
lt
to
im
p
r
o
v
e
d
u
e
t
o
a
lack
o
f
tr
ain
in
g
d
ata.
T
h
e
d
e
f
in
itio
n
o
f
tr
an
s
f
er
lear
n
in
g
ca
n
b
e
d
escr
ib
ed
as
f
o
llo
ws:
a
lear
n
i
n
g
ta
s
k
is
ass
ig
n
ed
to
T
t
b
ased
o
n
Dt
an
d
ca
n
r
ec
ei
v
e
ass
is
tan
ce
f
r
o
m
Ds
f
o
r
th
e
lea
r
n
in
g
t
ask
T
s
.
T
r
an
s
f
er
lear
n
i
n
g
aim
s
to
e
n
h
an
ce
th
e
p
er
f
o
r
m
an
ce
o
f
th
e
p
r
ed
ict
iv
e
f
u
n
ctio
n
f
T
(
.
)
f
o
r
th
e
lear
n
in
g
task
T
t
b
y
id
e
n
tify
in
g
an
d
tr
an
s
f
er
r
in
g
laten
t
k
n
o
wled
g
e
f
r
o
m
Ds
an
d
T
s
wh
er
e
Ds=Dt
an
d
/o
r
T
s
=T
t.
I
n
m
an
y
ca
s
es,
th
e
s
ize
o
f
D
s
is
lar
g
er
th
an
Dt,
N
s
>>
Nt
[
2
0
]
.
T
h
e
co
n
ce
p
t
o
f
I
o
T
was
cr
ea
ted
b
y
a
m
em
b
e
r
o
f
th
e
r
ad
io
f
r
eq
u
e
n
cy
id
e
n
tific
atio
n
(
R
FID
)
d
ev
elo
p
m
e
n
t
co
m
m
u
n
ity
in
1
9
9
9
.
Gen
er
ally
,
I
o
T
is
d
ef
in
e
d
as
a
n
etwo
r
k
o
f
p
h
y
s
ical
o
b
jects.
T
h
e
in
ter
n
et
is
n
o
t
o
n
l
y
a
n
etwo
r
k
o
f
c
o
m
p
u
t
er
s
b
u
t
h
as
ev
o
lv
ed
i
n
to
a
n
et
wo
r
k
o
f
d
e
v
ices
o
f
all
k
i
n
d
s
a
n
d
s
izes,
in
clu
d
in
g
v
eh
icles,
s
m
ar
tp
h
o
n
es,
h
o
u
s
eh
o
ld
ap
p
lian
ce
s
,
m
e
d
ical
eq
u
ip
m
en
t,
in
d
u
s
tr
ial
s
y
s
tem
s
,
h
u
m
an
s
,
an
im
als,
b
u
ild
in
g
s
,
all
co
n
n
ec
te
d
an
d
co
m
m
u
n
icatin
g
a
n
d
s
h
ar
in
g
in
f
o
r
m
atio
n
b
ased
o
n
estab
lis
h
ed
p
r
o
to
co
ls
to
ac
h
iev
e
in
tellig
en
t
r
eo
r
g
an
iza
tio
n
,
p
lace
m
e
n
t,
tr
ac
k
i
n
g
,
s
e
cu
r
ity
,
an
d
c
o
n
tr
o
l,
as
well
as
p
er
s
o
n
al
o
n
lin
e
m
o
n
ito
r
in
g
,
p
r
o
ce
s
s
co
n
tr
o
l,
a
n
d
ad
m
i
n
is
tr
atio
n
[
2
1
]
.
T
h
e
n
u
m
b
er
o
f
I
o
T
d
e
v
ices
is
r
ap
id
ly
in
cr
ea
s
in
g
,
a
n
d
th
e
lack
o
f
s
ec
u
r
ity
in
t
h
ese
d
ev
ices
h
as
m
ad
e
th
e
m
tar
g
e
ts
f
o
r
cr
im
in
al
ac
tiv
ities
.
Fig
u
r
e
1
p
r
esen
ts
th
e
v
ar
iatio
n
s
o
f
cy
b
e
r
s
ec
u
r
ity
attac
k
s
th
at
o
cc
u
r
at
th
e
I
o
T
lay
er
s
s
u
ch
as
th
e
p
er
ce
p
tio
n
,
s
u
p
p
o
r
t,
n
etwo
r
k
,
an
d
ap
p
licatio
n
lay
er
s
[
2
2
]
.
C
N
N
is
a
h
i
g
h
l
y
p
o
w
e
r
f
u
l
c
l
a
s
s
o
f
d
e
e
p
l
e
a
r
n
i
n
g
t
h
a
t
i
s
w
id
e
l
y
a
p
p
l
i
e
d
i
n
v
a
r
i
o
u
s
t
a
s
k
s
,
i
n
c
l
u
d
i
n
g
o
b
j
e
c
t
d
e
t
e
ct
i
o
n
,
s
p
e
e
c
h
r
e
c
o
g
n
i
t
i
o
n
,
c
o
m
p
u
t
e
r
v
i
s
i
o
n
,
i
m
ag
e
c
l
a
s
s
i
f
i
ca
t
i
o
n
,
b
i
o
i
n
f
o
r
m
a
t
ic
s
,
a
n
d
t
i
m
e
s
e
r
i
e
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
7
2
2
-
3
2
2
1
C
o
m
p
u
t Sci
I
n
f
T
ec
h
n
o
l
,
Vo
l.
6
,
No
.
2
,
J
u
ly
2
0
2
5
:
20
2
-
2
1
3
204
p
r
e
d
i
c
t
i
o
n
t
as
k
s
.
C
N
Ns
a
r
e
f
e
ed
f
o
r
w
a
r
d
n
e
u
r
a
l
n
e
t
w
o
r
k
s
t
h
at
u
t
i
l
i
z
e
c
o
n
v
o
l
u
t
i
o
n
a
l
s
t
r
u
c
t
u
r
e
s
t
o
e
x
t
r
ac
t
f
e
at
u
r
es
f
r
o
m
d
a
t
a
.
U
n
l
i
k
e
t
r
a
d
i
t
i
o
n
al
m
e
t
h
o
d
s
,
C
NN
s
a
u
t
o
m
at
i
c
al
l
y
l
e
a
r
n
a
n
d
r
e
c
o
g
n
i
z
e
f
e
a
t
u
r
e
s
f
r
o
m
d
a
t
a
w
i
t
h
o
u
t
r
e
q
u
i
r
i
n
g
m
a
n
u
a
l
f
e
at
u
r
e
e
x
tr
a
c
t
i
o
n
b
y
h
u
m
a
n
s
.
T
h
e
d
esi
g
n
o
f
C
N
Ns
is
i
n
s
p
i
r
e
d
b
y
v
i
s
u
al
p
e
r
c
e
p
ti
o
n
.
T
h
e
m
a
i
n
c
o
m
p
o
n
e
n
ts
o
f
C
N
N
s
i
n
c
l
u
d
e
c
o
n
v
o
l
u
t
i
o
n
a
l
l
ay
e
r
s
,
p
o
o
l
i
n
g
la
y
e
r
s
,
a
n
d
f
u
l
ly
c
o
n
n
e
c
t
e
d
l
a
y
e
r
s
.
T
h
e
c
o
n
v
o
l
u
t
i
o
n
l
a
y
e
r
is
a
c
r
u
c
i
a
l
c
o
m
p
o
n
e
n
t
o
f
C
N
Ns
.
T
h
r
o
u
g
h
s
e
v
e
r
a
l
c
o
n
v
o
l
u
ti
o
n
a
l
l
a
y
e
r
s
,
t
h
e
c
o
n
v
o
l
u
t
i
o
n
o
p
e
r
a
t
i
o
n
e
x
t
r
a
c
ts
d
i
f
f
e
r
e
n
t
f
e
at
u
r
e
s
f
r
o
m
t
h
e
i
n
p
u
t
.
P
o
o
l
i
n
g
la
y
e
r
:
t
y
p
i
c
al
l
y
f
o
l
l
o
w
i
n
g
t
h
e
co
n
v
o
l
u
t
i
o
n
l
a
y
e
r
,
t
h
e
p
o
o
l
i
n
g
l
a
y
e
r
r
e
d
u
c
e
s
t
h
e
n
u
m
b
e
r
o
f
c
o
n
n
e
c
t
i
o
n
s
i
n
t
h
e
n
e
t
wo
r
k
b
y
p
e
r
f
o
r
m
i
n
g
d
i
m
e
n
s
i
o
n
a
l
i
t
y
r
e
d
u
c
t
i
o
n
o
n
t
h
e
i
n
p
u
t
d
a
t
a
.
I
ts
p
r
i
m
a
r
y
o
b
j
e
ct
iv
e
i
s
t
o
d
ec
r
e
a
s
e
t
h
e
c
o
m
p
u
ta
t
i
o
n
a
l
l
o
a
d
a
n
d
a
d
d
r
es
s
t
h
e
is
s
u
e
o
f
o
v
e
r
f
i
tt
i
n
g
.
T
h
e
p
o
o
l
i
n
g
o
p
e
r
a
t
i
o
n
p
r
o
d
u
c
e
s
a
n
o
u
t
p
u
t
f
e
a
t
u
r
e
m
a
p
t
h
a
t
i
s
m
o
r
e
r
e
s
i
l
i
e
n
t
t
o
d
i
s
t
o
r
t
i
o
n
s
a
n
d
e
r
r
o
r
s
i
n
t
h
e
n
e
u
r
o
n
s
.
F
u
l
l
y
c
o
n
n
e
c
t
e
d
:
t
h
e
f
u
l
l
y
c
o
n
n
e
c
t
e
d
l
a
y
e
r
i
s
t
y
p
i
c
a
l
l
y
l
o
c
a
t
e
d
a
t
t
h
e
e
n
d
o
f
t
h
e
C
N
N
a
r
c
h
it
e
c
t
u
r
e
.
In
t
h
i
s
l
a
y
e
r
,
e
a
c
h
n
e
u
r
o
n
is
c
o
n
n
e
c
t
e
d
t
o
a
l
l
n
e
u
r
o
n
s
i
n
t
h
e
p
r
e
v
i
o
u
s
l
a
y
e
r
,
f
o
l
l
o
wi
n
g
t
h
e
p
r
i
n
c
i
p
le
s
o
f
c
o
n
v
e
n
t
i
o
n
a
l
m
u
l
ti
-
l
a
y
e
r
p
e
r
ce
p
t
r
o
n
n
e
u
r
a
l
n
e
t
w
o
r
k
s
.
T
h
e
f
u
l
l
y
c
o
n
n
e
c
t
e
d
l
a
y
e
r
r
e
c
ei
v
e
s
i
n
p
u
t
f
r
o
m
t
h
e
la
s
t
p
o
o
l
i
n
g
o
r
c
o
n
v
o
l
u
t
i
o
n
l
a
y
e
r
,
w
h
i
c
h
is
a
v
e
c
t
o
r
c
r
e
a
t
e
d
b
y
f
l
at
t
e
n
i
n
g
t
h
e
f
e
a
t
u
r
e
m
a
p
s
.
T
h
e
f
u
l
l
y
c
o
n
n
e
c
t
e
d
l
a
y
e
r
s
e
r
v
e
s
as
t
h
e
cl
as
s
i
f
i
c
a
ti
o
n
c
o
m
p
o
n
e
n
t
i
n
C
N
Ns
,
e
n
a
b
li
n
g
t
h
e
n
e
t
w
o
r
k
t
o
m
a
k
e
p
r
e
d
i
c
t
i
o
n
s
[
2
3
]
.
Fig
u
r
e
1
.
Var
iatio
n
s
in
cy
b
e
r
s
ec
u
r
ity
attac
k
s
at
th
e
I
o
T
lay
er
[
2
2
]
2
.
2
.
M
et
ho
do
lo
g
y
I
n
th
is
r
esear
ch
,
th
e
m
eth
o
d
o
l
o
g
y
em
p
lo
y
e
d
is
d
iv
id
ed
i
n
to
5
s
tag
es,
wh
ich
ar
e
ex
p
lain
ed
as f
o
llo
ws:
‒
Ph
ase
1
:
d
ata
p
r
ep
r
o
ce
s
s
in
g
I
o
T
B
o
T
d
ataset
‒
Ph
ase
2
:
d
ata
p
r
ep
r
o
ce
s
s
in
g
U
NSW
-
N
B
1
5
d
ataset
‒
Ph
ase
3
:
b
u
ild
an
d
tr
ain
–
test
m
o
d
el
1
with
I
o
T
B
o
T
d
ataset
f
o
r
b
ase
m
o
d
el
a
n
d
UNSW
-
NB
1
5
f
o
r
test
in
g
m
o
d
el
1
with
o
u
t tr
an
s
f
er
lear
n
in
g
‒
Ph
ase
4
:
b
u
ild
an
d
tr
ain
–
test
m
o
d
el
2
wh
ic
h
m
o
d
el
1
u
p
d
at
ed
with
tr
an
s
f
er
lear
n
in
g
f
o
r
m
b
ase
m
o
d
el
‒
Ph
ase
5
:
ev
alu
atio
n
m
o
d
el
2
T
h
e
r
esear
ch
m
et
h
o
d
o
lo
g
y
d
ia
g
r
am
ca
n
b
e
s
ee
n
in
Fig
u
r
e
2
.
2
.
3
.
Da
t
a
prepa
ra
t
i
o
n,
t
re
a
t
m
ent
,
a
nd
prepro
ce
s
s
ing
I
n
th
is
r
esear
ch
,
d
ev
elo
p
m
en
t
o
f
th
e
m
o
d
el
f
o
cu
s
es
o
n
b
u
i
ld
in
g
a
m
o
d
e
l
ca
p
a
b
le
o
f
d
et
ec
tin
g
an
d
class
if
y
in
g
v
ar
io
u
s
ty
p
es
o
f
I
o
T
attac
k
s
.
I
n
th
is
r
eg
a
r
d
,
tw
o
d
atasets
ar
e
u
tili
ze
d
f
o
r
th
e
s
o
u
r
ce
an
d
tr
an
s
f
e
r
d
o
m
ain
s
co
n
tain
in
g
s
tr
ea
m
s
o
f
n
o
r
m
al
I
o
T
tr
af
f
ic
an
d
cy
b
e
r
-
attac
k
s
.
B
o
T
-
I
o
T
d
ataset
is
u
s
ed
f
o
r
th
e
s
o
u
r
ce
d
o
m
ain
b
ec
au
s
e
it
co
n
tain
s
a
s
u
b
s
tan
tial
am
o
u
n
t
o
f
I
o
T
n
et
wo
r
k
ac
tiv
ity
d
ata,
wh
er
ea
s
U
NSW
-
N
B
1
5
d
ataset
is
u
s
ed
f
o
r
t
h
e
tar
g
et
d
o
m
ai
n
d
u
e
to
its
r
a
r
ity
a
n
d
d
is
p
r
o
p
o
r
tio
n
ate
r
e
p
r
esen
tatio
n
o
f
I
o
T
n
etwo
r
k
tr
af
f
ic
co
m
p
r
is
ed
o
f
cy
b
e
r
-
attac
k
s
.
T
o
ev
alu
ate
attac
k
d
etec
tio
n
,
f
o
u
r
d
if
f
e
r
en
t d
atasets
will b
e
cr
ea
ted
as d
escr
ib
ed
.
B
o
t
-
I
o
T
d
ataset
as
s
o
u
r
ce
d
o
m
ain
d
ataset
h
as
a
v
o
l
u
m
e
o
f
7
3
m
illi
o
n
s
am
p
le
[
2
4
]
.
Fo
r
th
is
r
esear
ch
,
5
%
o
f
th
e
d
ataset
will
b
e
u
s
e
d
,
to
talin
g
a
p
p
r
o
x
im
ately
3
.
6
m
illi
o
n
s
am
p
les
[
2
5
]
.
T
h
e
5
%
d
ataset
in
clu
d
es
n
o
r
m
al
tr
af
f
ic
s
am
p
les
an
d
4
ca
teg
o
r
ies
o
f
attac
k
s
,
n
am
el
y
d
en
ial
o
f
s
er
v
ice
(
Do
S),
d
i
s
tr
ib
u
ted
d
en
ial
o
f
s
er
v
ice
(
DDo
S),
r
ec
o
n
n
aiss
an
ce
,
an
d
in
f
o
r
m
atio
n
th
ef
t w
h
ic
h
s
h
o
wn
in
T
a
b
le
1
.
UNS
W
-
NB
1
5
as
t
a
r
g
e
t
d
o
m
a
i
n
d
a
t
as
e
t
[
2
6
]
,
c
r
e
a
t
e
d
u
s
i
n
g
t
h
e
I
X
I
A
P
e
r
f
e
c
tSt
o
r
m
t
o
o
l
i
n
t
h
e
C
y
b
e
r
R
a
n
g
e
L
a
b
a
t
U
N
SW
C
a
n
b
e
r
r
a
t
o
g
e
n
e
r
a
t
e
a
c
o
m
b
in
a
t
i
o
n
o
f
m
o
d
e
r
n
n
o
r
m
a
l
a
c
t
i
v
i
t
i
es
a
n
d
s
y
n
t
h
e
t
ic
c
o
n
t
e
m
p
o
r
a
r
y
a
t
t
a
c
k
b
e
h
a
v
i
o
r
s
.
T
h
e
A
r
g
u
s
t
o
o
l
a
n
d
B
r
o
-
I
D
S
w
e
r
e
u
t
i
li
z
e
d
,
a
n
d
1
2
a
l
g
o
r
i
t
h
m
s
w
e
r
e
d
e
v
el
o
p
e
d
t
o
Evaluation Warning : The document was created with Spire.PDF for Python.
C
o
m
p
u
t Sci
I
n
f
T
ec
h
n
o
l
I
SS
N:
2722
-
3
2
2
1
A
tta
ck
d
etec
tio
n
in
i
n
tern
et
o
f
th
in
g
s
n
etw
o
r
ks w
ith
d
ee
p
lea
r
n
in
g
u
s
in
g
…
(
R
iki
A
b
d
illa
h
H
a
s
a
n
u
d
d
in
)
205
p
r
o
d
u
c
e
4
9
f
e
a
t
u
r
e
s
l
a
b
e
l
e
d
as
n
o
r
m
a
l
a
n
d
a
t
t
a
c
k
.
T
h
e
a
t
t
ac
k
c
a
t
e
g
o
r
i
e
s
w
e
r
e
c
l
ass
i
f
i
e
d
i
n
t
o
9
g
r
o
u
p
s
b
a
s
e
d
o
n
t
h
e
n
a
t
u
r
e
o
f
t
h
e
a
t
ta
c
k
s
[
2
7
]
.
T
h
e
t
o
t
a
l
n
u
m
b
e
r
o
f
r
e
c
o
r
d
s
is
2
,
5
4
0
,
0
4
7
w
h
i
c
h
s
h
o
w
n
i
n
T
a
b
l
e
2
.
Fig
u
r
e
2
.
R
esear
ch
m
eth
o
d
o
l
o
g
y
d
iag
r
am
T
ab
le
1
.
Dis
tr
ib
u
tio
n
o
f
B
o
t
-
Io
T
5
% d
ataset
N
a
me
S
u
b
c
a
t
e
g
o
r
y
D
e
scri
p
t
i
o
n
R
e
c
o
r
d
%
N
o
r
mal
N
o
r
mal
N
a
t
u
r
a
l
t
r
a
n
sa
c
t
i
o
n
4
7
7
0
.
0
0
D
D
o
s
TC
P
UDP
H
TTP
A
t
t
a
c
k
w
h
e
r
e
m
u
l
t
i
p
l
e
c
o
m
p
r
o
m
i
se
d
c
o
m
p
u
t
e
r
sy
st
e
ms
a
t
t
a
c
k
a
t
a
r
g
e
t
,
c
a
u
s
i
n
g
a
D
O
S
1
,
9
2
6
,
6
2
4
5
2
.
5
1
D
o
s
TC
P
UDP
H
TTP
A
ma
l
i
c
i
o
u
s
a
t
t
a
c
k
t
o
c
r
i
p
p
l
e
t
h
e
s
e
r
v
i
c
e
s
o
f
f
e
r
e
d
b
y
a
s
i
t
e
,
serv
e
r
o
r
n
e
t
w
o
r
k
o
v
e
r
l
o
a
d
i
n
g
t
h
e
t
a
r
g
e
t
o
f
i
t
s
a
ss
o
c
i
a
t
e
d
b
y
f
l
o
o
d
i
n
g
t
h
e
si
t
e
w
i
t
h
ma
n
y
r
e
q
u
e
st
s
1
,
6
5
0
,
2
6
0
4
4
.
9
7
R
e
c
o
n
n
a
i
ss
a
n
c
e
OS
F
i
n
g
e
r
p
r
i
n
t
i
n
g
S
e
r
v
i
c
e
S
c
a
n
n
i
n
g
A
l
l
t
h
e
d
i
f
f
e
r
e
n
t
st
r
i
k
e
s s
i
mu
l
a
t
i
n
g
a
t
t
a
c
k
s
g
a
t
h
e
r
i
n
g
i
n
f
o
r
ma
t
i
o
n
9
1
,
8
8
2
2
.
5
0
I
n
f
o
r
mat
i
o
n
t
h
e
f
t
K
e
y
l
o
g
g
i
n
g
D
a
t
a
e
x
f
i
l
t
r
a
t
i
o
n
S
t
e
a
l
i
n
g
o
f
p
e
r
so
n
a
l
u
ser
i
n
f
o
r
ma
t
i
o
n
79
0
.
00
T
ab
le
2
.
Dis
tr
ib
u
tio
n
o
f
UNSW
-
NB
1
5
d
ataset
N
a
me
D
e
scri
p
t
i
o
n
R
e
c
o
r
d
%
N
o
r
mal
N
a
t
u
r
a
l
t
r
a
n
sa
c
t
i
o
n
d
a
t
a
2
,
2
1
8
,
7
6
4
8
7
.
3
5
G
e
n
e
r
i
c
A
t
t
a
c
k
a
g
a
i
n
s
t
b
l
o
c
k
c
i
p
h
e
r
s wi
t
h
a
g
i
v
e
n
b
l
o
c
k
a
n
d
k
e
y
s
i
z
e
(
n
o
t
c
o
n
si
d
e
r
i
n
g
i
t
s s
t
r
u
c
t
u
r
e
)
2
1
5
,
4
8
1
8
.
48
Ex
p
l
o
i
t
s
A
t
t
a
c
k
t
h
a
t
e
x
p
l
o
i
t
s v
u
l
n
e
r
a
b
i
l
i
t
i
e
s,
t
a
k
i
n
g
a
d
v
a
n
t
a
g
e
o
f
se
c
u
r
i
t
y
p
r
o
b
l
e
ms
(
o
f
a
n
o
p
e
r
a
t
i
n
g
sy
st
e
m
o
r
a
p
i
e
c
e
o
f
s
o
f
t
w
a
r
e
)
k
n
o
w
n
b
y
t
h
e
a
t
t
a
c
k
e
r
s.
4
4
,
5
2
5
1
.
75
F
u
z
z
e
r
s
A
t
t
a
c
k
t
h
a
t
s
u
sp
e
n
d
s
a
p
r
o
g
r
a
m
o
r
n
e
t
w
o
r
k
,
f
e
e
d
i
n
g
i
t
w
i
t
h
r
a
n
d
o
m
l
y
g
e
n
e
r
a
t
e
d
d
a
t
a
.
2
4
,
2
4
6
0
.
95
D
o
s
A
mal
i
c
i
o
u
s
a
t
t
a
c
k
t
h
a
t
ma
k
e
s
a
ser
v
e
r
o
r
n
e
t
w
o
r
k
r
e
so
u
r
c
e
u
n
a
v
a
i
l
a
b
l
e
,
o
v
e
r
l
o
a
d
i
n
g
t
h
e
t
a
r
g
e
t
o
f
t
h
e
a
sso
c
i
a
t
e
d
i
n
f
r
a
st
r
u
c
t
u
r
e
w
i
t
h
a
f
l
o
o
d
o
f
I
n
t
e
r
n
e
t
t
r
a
f
f
i
c
.
1
6
,
3
5
3
0
.
64
R
e
c
o
n
n
a
i
ss
a
n
c
e
C
o
m
p
r
i
s
e
s
d
i
f
f
e
r
e
n
t
a
t
t
a
c
k
s
t
h
a
t
g
a
t
h
e
r
i
n
f
o
r
ma
t
i
o
n
.
1
3
,
9
8
7
0
.
55
A
n
a
l
y
s
i
s
D
i
f
f
e
r
e
n
t
a
t
t
a
c
k
s
o
n
p
e
n
e
t
r
a
t
i
o
n
s (H
T
M
L
f
i
l
e
s,
s
p
a
m
,
a
n
d
p
o
r
t
s
c
a
n
)
2
,
6
7
7
0
.
11
B
a
c
k
d
o
o
r
A
n
a
t
t
a
c
k
t
h
a
t
b
y
p
a
sses
a
s
y
s
t
e
m
sec
u
r
i
t
y
me
c
h
a
n
i
sm
t
o
a
c
c
e
ss a
c
o
m
p
u
t
e
r
o
r
i
t
s
d
a
t
a
.
2
,
3
2
9
0
.
09
S
h
e
l
l
c
o
d
e
A
t
t
a
c
k
t
h
a
t
e
x
p
l
o
i
t
s s
o
f
t
w
a
r
e
v
u
l
n
e
r
a
b
i
l
i
t
i
e
s
u
si
n
g
sm
a
l
l
p
i
e
c
e
s
o
f
c
o
d
e
a
s
p
a
y
l
o
a
d
s.
1
,
5
1
1
0
.
06
W
o
r
ms
A
t
t
a
c
k
w
h
e
r
e
t
h
e
a
t
t
a
c
k
e
r
r
e
p
l
i
c
a
t
e
s
i
t
sel
f
t
o
sp
r
e
a
d
t
o
o
t
h
e
r
c
o
m
p
u
t
e
r
s.
1
7
4
0
.
01
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
7
2
2
-
3
2
2
1
C
o
m
p
u
t Sci
I
n
f
T
ec
h
n
o
l
,
Vo
l.
6
,
No
.
2
,
J
u
ly
2
0
2
5
:
20
2
-
2
1
3
206
As
in
d
icate
d
b
y
th
e
d
is
tr
ib
u
ti
o
n
o
f
UNSW
-
NB
1
5
,
wh
ich
e
x
h
ib
its
a
r
atio
o
f
8
7
%
n
o
r
m
al
tr
af
f
ic
an
d
1
3
%
attac
k
tr
af
f
ic,
th
e
s
y
s
tem
f
u
n
ctio
n
s
in
a
b
alan
ce
d
m
an
n
er
b
etwe
en
th
ese
two
tr
a
f
f
ic
c
ateg
o
r
ies.
T
h
is
was
s
u
b
s
eq
u
en
tly
b
alan
ce
d
t
o
ap
p
r
o
x
im
ate
a
r
ea
lis
tic
s
ce
n
ar
io
,
r
esu
ltin
g
in
5
0
%
o
f
n
o
r
m
al
tr
af
f
ic
an
d
5
0
%
o
f
attac
k
s
.
T
o
ascer
tain
th
e
ef
f
ic
ac
y
o
f
th
e
tr
an
s
f
e
r
l
ea
r
n
in
g
-
b
a
s
ed
m
o
d
el
in
d
etec
tin
g
b
o
th
k
n
o
wn
a
n
d
u
n
k
n
o
w
n
attac
k
s
,
th
r
ee
d
is
tin
ct
d
atasets
wer
e
g
en
er
ated
in
th
e
tar
g
et
d
o
m
ain
an
d
ar
e
p
r
esen
ted
in
T
a
b
le
3
.
a)
UNSW
-
N
B
1
5
_
b
ase4
:
d
ataset
co
n
tain
in
g
n
o
r
m
al
tr
af
f
ic
an
d
th
r
ee
ty
p
e
o
f
k
n
o
wn
attac
k
s
(
g
en
er
ic,
Do
S,
an
d
r
ec
o
n
n
aiss
an
ce
)
u
s
ed
f
o
r
t
r
ain
in
g
an
d
d
iv
i
d
ed
in
to
two
s
u
b
d
ataset:
‒
UNSW
-
N
B
1
5
_
b
ase4
_
tr
ain
: 7
5
% o
f
UNSW
-
NB
1
5
_
b
ase4
to
tr
ain
m
o
d
el
1
.
‒
UNSW
-
N
B
1
5
_
b
ase4
_
test
:
2
5
%
o
f
UNSW
-
NB
1
5
_
b
ase4
to
ev
alu
ate
th
e
ef
f
ec
tiv
e
n
ess
in
d
etec
tio
n
o
f
k
n
o
wn
attac
k
s
.
b)
UNSW
-
N
B
1
5
_
f
ir
s
t
-
test
:
d
ataset
to
ev
alu
ate
ef
f
ec
tiv
en
ess
in
d
etec
tio
n
o
f
u
n
k
n
o
w
n
at
tack
s
(
ex
p
lo
its
,
f
u
zz
er
s
,
an
alis
is
,
b
ac
k
d
o
o
r
,
s
h
ellco
d
e,
an
d
wo
r
m
s
)
.
c)
UNSW
-
N
B
1
5
_
f
u
ll
-
test
:
d
atas
et
to
ev
alu
ate
ef
f
ec
tiv
en
ess
in
d
etec
tio
n
o
f
attac
k
ty
p
e
k
n
o
wn
attac
k
s
(
g
en
er
ic
,
Do
S,
an
d
r
ec
o
n
n
aiss
an
ce
)
an
d
u
n
k
n
o
wn
attac
k
s
(
ex
p
lo
its
,
f
u
zz
er
s
,
an
alis
is
,
b
ac
k
d
o
o
r
,
s
h
ellco
d
e
,
an
d
wo
r
m
s
)
.
I
t
is
im
p
o
r
tan
t
to
n
o
te
th
at
p
h
a
s
es
1
an
d
2
o
f
th
e
f
r
am
ewo
r
k
s
i
n
v
o
lv
e
d
ata
p
r
o
ce
s
s
in
g
.
I
n
th
e
tr
an
s
f
er
lear
n
in
g
-
b
ased
m
o
d
el
u
n
d
er
d
esig
n
,
th
e
o
u
tp
u
t
f
r
o
m
th
e
co
n
v
o
l
u
tio
n
al
b
ase
is
u
tili
ze
d
as
in
p
u
t
f
o
r
th
e
class
if
ier
.
Su
b
s
eq
u
en
tly
,
th
e
d
atasets
f
o
r
th
e
s
o
u
r
ce
an
d
tar
g
et
d
o
m
ain
s
m
u
s
t
b
e
tr
ain
e
d
with
th
e
s
am
e
in
p
u
t
f
o
r
m
ats
an
d
f
ea
tu
r
es.
Fo
r
t
h
is
s
tu
d
y
,
n
ew
v
er
s
io
n
s
wer
e
g
e
n
er
ated
f
o
r
b
o
th
d
atasets
with
th
eir
1
5
c
o
m
m
o
n
f
ea
tu
r
es,
wh
ich
ar
e
s
h
o
wn
in
T
ab
le
4
.
I
n
th
is
p
h
ase,
th
e
c
o
lu
m
n
s
wi
th
s
tr
in
g
f
o
r
m
ats
ar
e
co
n
v
er
te
d
to
n
u
m
er
ic
f
o
r
m
ats
u
s
in
g
t
h
e
o
n
e
h
o
t
en
co
d
in
g
(
OHE
)
m
eth
o
d
.
T
h
e
co
lu
m
n
s
f
o
r
m
atted
in
h
ex
a
d
ec
im
al
(
HE
X)
ar
e
co
n
v
er
ted
to
d
ec
im
al
(
DE
C
)
f
o
r
m
at.
A
lo
g
ar
ith
m
ic
p
r
o
ce
d
u
r
e
is
ap
p
lied
to
th
e
co
lu
m
n
s
(
ie.
,
d
u
r
,
s
b
y
tes,
d
b
y
tes,
an
d
s
p
k
ts
)
wh
ich
h
av
e
co
n
ce
n
tr
ated
v
al
u
es
at
0
.
St
an
d
ar
d
n
o
r
m
aliza
tio
n
o
f
th
e
d
ataset
is
p
er
f
o
r
m
ed
to
p
r
ev
en
t
o
v
er
f
itti
n
g
an
d
p
o
ten
tial
b
ias
in
th
e
r
esu
lts
.
Fin
ally
,
th
e
d
ata
is
tr
an
s
f
o
r
m
ed
in
to
an
im
ag
e
f
o
r
m
at
in
to
a
1
D
f
o
r
m
at
(
i.e
.
,
a
v
ec
to
r
o
f
len
g
th
2
4
y
ield
in
g
d
im
e
n
s
io
n
s
o
f
[
1
]
,
[
2
4
]
).
T
ab
le
3
.
Dis
tr
ib
u
tio
n
o
f
UNSW
-
NB
1
5
s
u
b
d
ataset
U
N
S
W
-
NB
1
5
_
b
a
se
4
_
t
r
a
i
n
U
N
S
W
-
N
B
1
5
_
b
a
se
4
_
t
e
st
U
N
S
W
-
N
B
1
5
_
f
i
r
st
-
t
e
s
t
U
N
S
W
-
N
B
1
5
_
f
u
l
l
-
t
e
st
N
a
me
R
e
c
o
r
d
%
R
e
c
o
r
d
%
R
e
c
o
r
d
%
R
e
c
o
r
d
%
N
o
r
mal
1
8
3
,
9
6
9
5
1
.
6
7
6
1
,
4
6
5
5
0
.
0
1
7
5
,
4
6
2
5
0
.
0
0
3
2
1
,
2
8
3
5
0
.
0
0
G
e
n
e
r
i
c
1
6
0
,
1
9
8
4
4
.
9
9
5
3
,
9
3
8
4
3
.
8
8
2
1
5
,
4
8
1
3
3
.
5
3
Ex
p
l
o
i
t
s
4
4
,
5
2
5
2
9
.
5
0
4
4
,
5
2
5
6
.
9
3
F
u
z
z
e
r
s
2
4
,
2
4
6
1
6
.
0
7
2
4
,
2
4
6
3
.
7
7
D
o
s
8
,
9
7
1
2
.
5
2
4
,
0
7
3
3
.
3
1
1
6
,
3
5
3
2
.
5
4
R
e
c
o
n
n
a
i
ss
a
n
c
e
2
,
9
0
9
0
.
8
2
3
,
4
3
4
2
.
7
9
1
3
,
9
8
7
2
.
1
8
A
n
a
l
y
s
i
s
2
,
6
7
7
1
.
7
7
2
,
6
7
7
0
.
4
2
B
a
c
k
d
o
o
r
2
,
3
2
9
1
.
5
4
2
,
3
2
9
0
.
3
6
S
h
e
l
l
c
o
d
e
1
,
5
1
1
1
.
0
0
1
,
5
1
1
0
.
2
4
W
o
r
ms
1
7
4
0
.
1
2
1
7
4
0
.
0
3
T
ab
le
4
.
1
5
co
m
m
o
n
f
ea
tu
r
es i
n
b
o
th
d
ataset
No
B
o
t
-
I
o
T
U
N
S
W
N
B
-
15
Ty
p
e
D
e
scri
p
t
i
o
n
1
P
r
o
t
o
P
r
o
t
o
N
o
mi
n
a
l
Te
x
t
u
a
l
r
e
p
r
e
se
n
t
a
t
i
o
n
o
f
t
r
a
n
sac
t
i
o
n
p
r
o
t
o
c
o
l
s
p
r
e
s
e
n
t
s
i
n
n
e
t
w
o
r
k
f
l
o
w
2
S
a
d
d
r
S
r
c
i
p
N
o
mi
n
a
l
S
o
u
r
c
e
I
P
a
d
d
r
e
ss
.
3
S
p
o
r
t
S
p
o
r
t
I
n
t
e
g
e
r
S
o
u
r
c
e
p
o
r
t
n
u
mb
e
r
.
4
D
a
d
d
r
D
st
i
p
N
o
mi
n
a
l
D
e
st
i
n
a
t
i
o
n
I
P
a
d
d
r
e
ss.
5
D
p
o
r
t
D
sp
o
r
t
I
n
t
e
g
e
r
D
e
st
i
n
a
t
i
o
n
P
o
r
t
n
u
mb
e
r
.
6
S
p
k
t
s
S
p
k
t
s
F
l
o
a
t
S
o
u
r
c
e
-
to
-
d
e
st
i
n
a
t
i
o
n
p
a
c
k
e
t
c
o
u
n
t
7
D
p
k
t
s
D
p
k
t
s
F
l
o
a
t
D
e
st
i
n
a
t
i
o
n
-
to
-
so
u
r
c
e
p
a
c
k
e
t
c
o
u
n
t
.
8
S
b
y
t
e
s
S
b
y
t
e
s
F
l
o
a
t
S
o
u
r
c
e
-
to
-
d
e
st
i
n
a
t
i
o
n
b
y
t
e
c
o
u
n
t
9
D
b
y
t
e
s
D
b
y
t
e
s
F
l
o
a
t
D
e
st
i
n
a
t
i
o
n
-
to
-
so
u
r
c
e
b
y
t
e
c
o
u
n
t
.
10
S
t
a
t
e
S
t
a
t
e
N
o
mi
n
a
l
Tr
a
n
s
a
c
t
i
o
n
st
a
t
e
.
11
S
t
i
me
S
t
i
me
Ti
mes
t
a
m
p
R
e
c
o
r
d
st
a
r
t
t
i
m
e
.
12
Lt
i
m
e
Lt
i
m
e
Ti
mes
t
a
m
p
R
e
c
o
r
d
e
n
d
t
i
me
.
13
D
u
r
D
u
r
F
l
o
a
t
R
e
c
o
r
d
t
o
t
a
l
d
u
r
a
t
i
o
n
.
14
A
t
t
a
c
k
La
b
e
l
B
i
n
a
r
y
C
l
a
s
s l
a
b
e
l
:
0
f
o
r
n
o
r
ma
l
t
r
a
f
f
i
c
,
1
f
o
r
a
t
t
a
c
k
.
15
C
a
t
e
g
o
r
y
a
t
t
a
c
k
_
c
a
t
N
o
mi
n
a
l
C
y
b
e
r
a
t
t
a
c
k
f
a
mi
l
y
.
2
.
4
.
T
ra
ns
f
er
lea
rning
a
nd
m
o
del desig
n
I
n
p
h
ase
3
,
m
o
d
el
1
is
co
n
s
t
r
u
cted
as
a
b
ase
m
o
d
el
u
s
in
g
th
e
B
o
T
-
I
o
T
d
ataset,
wh
ich
h
as
b
ee
n
r
an
d
o
m
l
y
p
ar
titi
o
n
e
d
in
to
7
5
%
tr
ain
in
g
d
ata
an
d
2
5
%
test
i
n
g
d
ata.
I
n
p
h
ase
4
,
m
o
d
el
1
e
x
p
er
ien
ce
d
an
u
p
d
ate
Evaluation Warning : The document was created with Spire.PDF for Python.
C
o
m
p
u
t Sci
I
n
f
T
ec
h
n
o
l
I
SS
N:
2722
-
3
2
2
1
A
tta
ck
d
etec
tio
n
in
i
n
tern
et
o
f
th
in
g
s
n
etw
o
r
ks w
ith
d
ee
p
lea
r
n
in
g
u
s
in
g
…
(
R
iki
A
b
d
illa
h
H
a
s
a
n
u
d
d
in
)
207
th
at
in
teg
r
ated
th
e
k
n
o
wled
g
e
ac
q
u
ir
ed
in
th
e
s
o
u
r
ce
d
o
m
ain
in
to
th
e
tar
g
et
d
o
m
ain
,
r
esu
lti
n
g
in
th
e
f
o
r
m
atio
n
o
f
m
o
d
el
2
.
I
n
m
o
d
el
2
,
t
h
e
c
o
n
v
o
lu
ti
o
n
al
b
ase
o
f
th
e
o
r
ig
i
n
al
b
ase
m
o
d
el
is
f
ix
ed
,
an
d
th
e
class
if
ier
is
f
ed
with
its
o
u
tp
u
ts
.
M
o
d
el
2
is
t
r
ain
ed
u
s
in
g
th
e
UNSW
-
NB
1
5
_
B
ase4
_
tr
ain
d
ataset,
w
h
ich
r
ep
r
es
en
ts
7
5
%
o
f
th
e
UNSW
-
N
B
1
5
_
B
ase4
d
ataset.
I
n
th
e
f
in
al
p
h
ase
o
f
th
e
ev
alu
atio
n
,
m
o
d
el
2
'
s
ca
p
ab
ilit
ies
to
d
etec
t
b
o
th
k
n
o
wn
an
d
u
n
k
n
o
wn
att
ac
k
s
will
b
e
ass
e
s
s
ed
by
u
s
in
g
th
e
UNSW
-
NB
1
5
_
f
ir
s
t_
test
an
d
UNSW
-
N
B
1
5
_
f
u
ll_
test
d
atasets
.
T
h
is
ap
p
r
o
ac
h
e
n
co
m
p
as
s
es
b
o
th
k
n
o
wn
a
n
d
u
n
k
n
o
w
n
attac
k
s
ce
n
ar
io
s
,
th
er
eb
y
e
n
s
u
r
in
g
a
co
m
p
r
eh
en
s
iv
e
ass
es
s
m
en
t o
f
Mo
d
el
2
'
s
ca
p
ab
ilit
ies.
T
h
is
s
tu
d
y
d
esig
n
s
a
n
d
d
ev
elo
p
s
a
1
D
-
C
NN
f
o
r
attac
k
d
etec
tio
n
in
I
o
T
n
etwo
r
k
s
.
Mo
d
el
1
s
er
v
es
as
th
e
b
ase
m
o
d
el,
an
d
m
o
d
el
2
a
d
d
s
k
n
o
wled
g
e
f
r
o
m
it.
T
h
e
p
r
o
p
o
s
ed
m
o
d
els
a
r
e
o
u
tlin
ed
in
Fig
u
r
e
3
.
T
h
e
f
ir
s
t
m
o
d
el
(
h
er
ea
f
ter
r
ef
er
r
e
d
to
as
"
m
o
d
el
1
")
co
n
s
is
ts
o
f
an
in
p
u
t
lay
er
,
two
b
lo
ck
s
o
f
co
n
v
o
lu
tio
n
lay
er
s
,
a
f
latten
lay
er
,
a
f
u
lly
co
n
n
ec
t
ed
lay
er
,
a
n
d
an
o
u
tp
u
t
lay
e
r
.
E
ac
h
b
lo
c
k
c
o
m
p
r
is
es
a
co
n
v
o
lu
tio
n
al
lay
er
,
a
n
o
r
m
aliza
tio
n
lay
er
,
a
p
o
o
li
n
g
lay
er
,
an
d
a
d
r
o
p
o
u
t
lay
e
r
.
T
h
e
co
n
v
o
l
u
tio
n
la
y
er
is
a
cr
i
tical
co
m
p
o
n
en
t
o
f
d
ee
p
lear
n
in
g
alg
o
r
ith
m
s
,
r
esp
o
n
s
ib
le
f
o
r
ex
tr
ac
ti
n
g
f
ea
tu
r
es
f
r
o
m
r
aw
d
ata.
I
t
ac
co
m
p
lis
h
es
th
is
b
y
lear
n
in
g
d
ata
attr
ib
u
tes
f
r
o
m
s
m
all
s
u
b
-
s
am
p
les
o
f
th
e
in
p
u
t
d
at
a,
en
s
u
r
in
g
th
e
p
r
eser
v
atio
n
o
f
th
e
u
n
d
er
l
y
in
g
ass
o
ciatio
n
s
b
etwe
en
d
ata
p
o
i
n
ts
.
T
h
e
o
b
jectiv
e
o
f
lay
er
n
o
r
m
aliza
tio
n
is
to
n
o
r
m
alize
all
in
p
u
ts
to
a
s
p
ec
if
ic
n
eu
r
al
n
etwo
r
k
lay
er
.
T
h
e
l
ay
er
n
o
r
m
aliza
tio
n
lay
er
is
r
esp
o
n
s
ib
le
f
o
r
s
tan
d
ar
d
izin
g
th
e
o
u
tp
u
t
o
f
th
e
co
n
v
o
l
u
tio
n
lay
er
f
o
r
th
e
m
ax
p
o
o
lin
g
lay
e
r
.
T
h
e
m
a
x
p
o
o
li
n
g
lay
er
is
r
esp
o
n
s
ib
le
f
o
r
d
eter
m
in
in
g
th
e
to
ta
l
n
u
m
b
er
o
f
f
ea
tu
r
es
p
r
esen
t
in
ea
ch
p
atc
h
.
I
t
id
en
tifie
s
th
e
f
ea
tu
r
es
th
at
d
em
o
n
s
tr
ate
th
e
g
r
ea
test
d
o
m
in
a
n
c
e
with
in
s
p
ec
if
ic
lo
ca
lized
r
e
g
io
n
s
.
T
h
is
p
r
o
ce
s
s
en
a
b
les
th
e
n
etwo
r
k
to
p
r
io
r
itize
th
e
m
o
s
t
s
ig
n
if
ican
t
elem
en
ts
wh
ile
ef
f
ec
tiv
ely
r
ed
u
cin
g
r
e
d
u
n
d
a
n
cy
.
A
s
p
atial
d
r
o
p
o
u
t
lay
er
is
em
p
lo
y
ed
to
elim
in
at
e
th
e
en
tire
f
ea
tu
r
e
m
ap
r
at
h
er
t
h
an
in
d
iv
id
u
al
u
n
its
.
T
h
e
f
latten
lay
er
is
f
u
lly
c
o
n
n
ec
ted
to
a
d
en
s
e
lay
e
r
.
T
h
e
f
in
al
lay
er
o
f
th
e
m
o
d
el
is
d
esig
n
ated
as th
e
o
u
t
p
u
t la
y
er
f
o
r
b
in
ar
y
class
if
icati
o
n
.
Mo
d
el
2
r
ep
r
esen
ts
an
u
p
d
ate
d
v
er
s
io
n
o
f
m
o
d
el
1
.
I
t
co
m
p
r
is
es
an
in
p
u
t
lay
er
,
a
f
r
o
ze
n
l
ay
er
with
tr
an
s
f
er
lear
n
in
g
f
o
r
th
e
b
ase
m
o
d
el,
a
f
latten
lay
er
,
th
r
ee
b
lo
ck
s
o
f
f
u
lly
c
o
n
n
ec
ted
la
y
er
s
with
a
d
r
o
p
o
u
t
lay
er
,
a
n
d
a
n
o
u
tp
u
t
lay
er
.
T
h
e
in
p
u
t
lay
er
co
n
tai
n
s
an
eq
u
al
n
u
m
b
er
o
f
f
ea
tu
r
es.
Du
r
in
g
th
e
tr
ai
n
in
g
o
f
th
e
b
in
ar
y
class
if
icatio
n
m
o
d
el,
th
e
co
n
v
o
lu
tio
n
lay
er
s
,
n
o
r
m
ali
za
tio
n
lay
er
s
,
p
o
o
lin
g
lay
er
s
,
d
r
o
p
o
u
t
lay
er
s
,
an
d
f
latten
lay
er
s
wer
e
f
r
o
ze
n
.
Du
r
in
g
th
e
tr
ain
in
g
p
h
ase,
lear
n
in
g
was
p
er
m
itted
ex
clu
s
iv
ely
in
th
e
d
en
s
e
an
d
o
u
tp
u
t la
y
e
r
s
.
As d
em
o
n
s
tr
ate
d
in
Fig
u
r
e
3
,
a
g
en
er
al
p
er
s
p
e
ctiv
e
f
r
o
m
b
o
th
m
o
d
els is
p
r
e
s
en
ted
.
Fig
u
r
e
3
.
Gen
e
r
al
v
iew
o
f
m
o
d
el
1
an
d
m
o
d
el
2
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
7
2
2
-
3
2
2
1
C
o
m
p
u
t Sci
I
n
f
T
ec
h
n
o
l
,
Vo
l.
6
,
No
.
2
,
J
u
ly
2
0
2
5
:
20
2
-
2
1
3
208
I
n
m
o
d
el
1
,
th
e
in
p
u
t f
o
r
1
D
-
C
NN
is
two
-
d
im
en
s
io
n
al.
I
n
itially
,
an
in
p
u
t v
ec
to
r
[
1
]
,
[
2
4
]
is
g
en
er
ated
to
ac
co
m
m
o
d
ate
th
e
1
5
c
o
m
m
o
n
f
ea
t
u
r
es
in
p
h
ases
1
an
d
2
.
Su
b
s
eq
u
en
t
to
t
h
e
in
co
r
p
o
r
atio
n
o
f
th
e
i
n
p
u
t
lay
er
,
th
e
m
o
d
el
was
au
g
m
en
ted
with
f
o
u
r
co
n
v
o
l
u
tio
n
lay
er
b
lo
ck
s
.
T
h
e
co
n
v
o
l
u
tio
n
o
p
er
atio
n
in
th
e
f
i
r
s
t
lay
er
is
co
n
f
ig
u
r
ed
with
a
6
4
-
f
ilt
er
co
n
f
ig
u
r
atio
n
,
a
2
4
-
f
i
lter
k
er
n
el
s
ize,
a
r
ec
tifie
d
l
in
ea
r
u
n
it
(
R
eL
U)
ac
tiv
atio
n
f
u
n
ctio
n
,
an
d
th
e
s
am
e
p
ad
d
in
g
p
ar
am
eter
.
T
h
e
lay
er
n
o
r
m
aliza
tio
n
ad
j
u
s
ts
t
h
e
p
r
ec
ed
in
g
lay
er
ac
tiv
atio
n
s
ep
ar
ately
f
o
r
ea
ch
s
am
p
le
in
a
g
iv
en
b
atch
.
T
h
e
m
ax
im
u
m
p
o
o
lin
g
l
a
y
er
p
r
o
v
i
d
es
a
s
o
lu
tio
n
th
at
en
s
u
r
es
th
e
p
r
eser
v
atio
n
o
f
th
e
m
o
s
t
s
alien
t
f
ea
tu
r
es,
th
e
r
eb
y
en
h
a
n
cin
g
th
e
ef
f
icien
cy
o
f
th
e
tr
ain
in
g
p
r
o
ce
s
s
an
d
im
p
r
o
v
i
n
g
t
h
e
p
e
r
f
o
r
m
an
ce
o
f
th
e
m
o
d
el.
A
s
p
atial
d
r
o
p
o
u
t
lay
er
is
em
p
lo
y
ed
to
r
eg
u
lar
ize
th
e
t
r
ain
in
g
d
ata
m
o
d
el
an
d
m
itig
ate
o
v
e
r
f
itti
n
g
,
with
a
d
r
o
p
v
alu
e
o
f
0
.
0
5
.
I
t
is
n
o
tewo
r
th
y
t
h
at
ea
ch
o
f
th
e
f
o
u
r
co
n
v
o
l
u
tio
n
lay
e
r
s
em
p
lo
y
s
i
d
en
tical
p
ar
am
ete
r
s
.
T
h
e
class
if
icatio
n
co
m
p
o
n
en
t
c
o
m
p
r
is
es
f
u
lly
co
n
n
ec
ted
,
f
latten
ed
,
an
d
d
e
n
s
e
lay
er
s
.
T
h
e
f
latten
lay
er
is
ap
p
li
ed
to
th
e
m
o
d
el,
th
er
e
b
y
tr
an
s
f
o
r
m
i
n
g
th
e
ten
s
o
r
in
to
a
s
h
ap
e
th
at
is
eq
u
iv
alen
t to
th
e
ten
s
o
r
elem
en
ts
.
T
h
e
f
latten
lay
er
is
co
n
n
ec
ted
to
a
f
u
lly
co
n
n
ec
ted
d
en
s
e
lay
er
,
an
d
th
e
d
e
n
s
e
lay
er
is
co
n
n
e
cted
to
th
e
o
u
tp
u
t
lay
er
.
T
h
e
d
en
s
e
lay
er
co
n
tain
s
5
1
2
n
e
u
r
o
n
s
,
an
d
a
s
in
g
le
n
eu
r
o
n
is
r
esp
o
n
s
ib
le
f
o
r
im
p
lem
en
tin
g
th
e
s
ig
m
o
id
ac
tiv
at
io
n
f
u
n
ctio
n
in
t
h
e
o
u
tp
u
t
lay
er
.
T
h
e
m
o
d
el
h
as
b
ee
n
tr
ain
e
d
o
v
er
th
e
c
o
u
r
s
e
o
f
1
5
ep
o
ch
s
,
with
a
b
atc
h
s
ize
o
f
2
5
6
an
d
an
Ad
am
o
p
tim
i
ze
r
with
a
lea
r
n
in
g
r
ate
o
f
2
×
1
0
-
5
T
h
is
tr
ain
in
g
p
r
o
ce
s
s
h
as
b
ee
n
im
p
lem
e
n
ted
to
m
in
im
ize
th
e
e
r
r
o
r
f
u
n
ctio
n
an
d
th
e
b
i
n
ar
y
cr
o
s
s
-
en
tr
o
p
y
lo
s
s
f
u
n
ctio
n
.
I
n
m
o
d
el
2
,
th
e
co
n
v
o
l
u
tio
n
lay
er
f
r
o
m
th
e
m
o
d
el
1
is
to
b
e
f
r
o
ze
n
in
o
r
d
er
to
p
r
e
v
e
n
t
weig
h
t
ch
an
g
es
d
u
r
in
g
tr
ain
i
n
g
.
T
h
is
lay
er
is
to
b
e
f
o
llo
wed
b
y
th
r
e
e
b
lo
ck
s
o
f
f
u
lly
co
n
n
ec
ted
la
y
er
s
.
T
h
e
f
ir
s
t
lay
e
r
co
n
s
is
ts
o
f
5
1
2
n
eu
r
o
n
s
with
a
d
r
o
p
v
alu
e
o
f
0
.
4
,
th
e
s
ec
o
n
d
lay
er
co
n
s
is
ts
o
f
2
5
6
n
eu
r
o
n
s
with
a
d
r
o
p
v
alu
e
o
f
0
.
3
,
a
n
d
th
e
last
lay
er
co
n
s
is
ts
o
f
1
2
8
n
eu
r
o
n
s
with
a
d
r
o
p
v
alu
e
o
f
0
.
2
.
T
h
e
o
u
t
p
u
t
lay
er
is
co
m
p
r
is
ed
o
f
a
s
in
g
le
n
eu
r
o
n
th
at
ex
h
ib
its
s
ig
m
o
id
ac
tiv
atio
n
.
T
h
e
m
o
d
el
h
as
b
ee
n
tr
ain
ed
o
v
er
5
0
e
p
o
ch
s
,
with
a
b
atch
s
ize
o
f
4
0
9
6
,
an
Ad
am
o
p
tim
ize
r
with
a
lear
n
in
g
r
ate
o
f
6
×1
0
-
5
,
an
d
a
b
in
ar
y
cr
o
s
s
-
en
tr
o
p
y
lo
s
s
f
u
n
ctio
n
.
T
h
e
tr
ain
in
g
p
a
r
am
eter
s
f
o
r
b
o
th
m
o
d
els ar
e
s
u
m
m
ar
ized
in
T
ab
le
5
.
2
.
5
.
E
v
a
lua
t
i
o
n m
et
rics
I
n
p
h
ase
5
,
th
e
two
m
o
d
els
ar
e
v
alid
ated
u
s
in
g
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all,
F1
-
s
co
r
e,
a
n
d
ar
ea
u
n
d
er
th
e
r
ec
eiv
er
o
p
er
atin
g
ch
a
r
ac
ter
is
tic
cu
r
v
e
(
AUC
-
R
OC
)
s
co
r
e.
Acc
u
r
ac
y
is
d
ef
i
n
ed
as
th
e
p
r
o
p
o
r
tio
n
o
f
s
am
p
les
th
at
ar
e
co
r
r
ec
tly
id
en
tifie
d
,
ex
p
r
ess
ed
as
a
p
er
ce
n
tag
e
o
f
th
e
to
tal
n
u
m
b
er
o
f
s
am
p
les.
Pre
cisi
o
n
i
s
q
u
an
tifie
d
b
y
th
e
p
r
o
p
o
r
tio
n
o
f
in
s
tan
ce
s
th
at
ar
e
ac
cu
r
ately
class
if
ied
to
th
e
to
tal
tr
u
e
p
o
s
itiv
e
(
T
P)
an
d
f
alse
p
o
s
itiv
e
(
FP
)
ca
s
es.
T
h
e
ca
lcu
latio
n
o
f
r
ec
all
is
d
eter
m
in
ed
b
y
t
h
e
d
iv
is
io
n
o
f
th
e
t
o
tal
n
u
m
b
e
r
o
f
T
P
m
ea
s
u
r
em
en
ts
b
y
th
e
to
tal
T
P
an
d
f
alse
n
eg
ativ
e
(
FN)
m
ea
s
u
r
em
en
ts
.
T
h
e
F1
-
s
co
r
e
is
d
eter
m
in
ed
b
y
th
e
ca
lcu
latio
n
o
f
th
e
weig
h
te
d
a
v
er
ag
e
o
f
p
r
ec
is
io
n
an
d
r
ec
al
l.
Fu
r
th
er
m
o
r
e,
th
e
C
NN
m
o
d
el'
s
v
alid
atio
n
is
s
u
p
p
o
r
ted
b
y
th
e
f
alse
p
o
s
itiv
e
r
ate
(
FP
R
)
,
d
ef
in
ed
as
th
e
n
u
m
b
er
o
f
n
o
r
m
al
s
am
p
les
th
at
ar
e
class
if
ied
as
p
o
s
itiv
e,
an
d
th
e
f
alse
n
eg
ativ
e
r
ate
(
FNR
)
,
d
en
o
tin
g
th
e
n
u
m
b
er
o
f
ab
n
o
r
m
al
s
am
p
les
th
at
ar
e
id
en
tifie
d
as
n
eg
ativ
e.
T
h
e
Ma
tth
ews
co
r
r
e
latio
n
co
ef
f
icien
t
(
MCC
)
is
a
m
etr
ic
th
at
co
n
s
id
er
s
tr
u
e
a
n
d
f
alse
p
o
s
itiv
es
an
d
n
eg
ativ
es,
th
er
e
b
y
p
r
o
v
i
d
in
g
a
b
alan
ce
d
m
ea
s
u
r
e
o
f
class
if
icatio
n
p
er
f
o
r
m
a
n
ce
.
T
h
e
r
a
n
g
e
o
f
p
o
s
s
ib
le
v
alu
es
is
f
r
o
m
-
1
to
1
.
I
n
th
is
co
n
te
x
t,
1
in
d
icate
s
p
er
f
ec
t
p
r
ed
ictio
n
,
0
in
d
icate
s
n
o
b
etter
th
an
r
an
d
o
m
ch
a
n
ce
,
a
n
d
-
1
in
d
icate
s
to
tal
d
is
ag
r
ee
m
en
t
b
etwe
en
p
r
ed
ictio
n
an
d
tr
u
e
p
r
ed
ictio
n
[
2
8
]
.
=
(
+
)
(
+
+
+
)
(
1
)
=
(
+
)
(
2
)
=
(
+
)
(
3
)
1
=
2
×
(
∗
)
(
+
)
(
4
)
=
×
−
×
√
(
+
)
×
(
+
)
×
(
+
)
×
(
+
)
(
5
)
2
.
6
.
Sy
s
t
e
m
s
et
up
T
h
e
tr
an
s
f
er
lear
n
in
g
ex
p
e
r
im
en
tal
en
v
ir
o
n
m
en
t
was
co
n
s
tr
u
cted
o
n
an
Asu
s
X4
4
1
UB
lap
to
p
,
wh
ich
is
s
u
p
p
o
r
ted
b
y
th
e
f
o
llo
wi
n
g
h
ar
d
war
e
s
p
ec
if
icatio
n
s
:
an
I
n
tel®
C
o
r
e™
i3
-
6
0
0
6
U
C
PU
@
2
.
0
GH
z
p
r
o
ce
s
s
o
r
an
d
8
GB
o
f
R
AM
DDR4
.
T
h
is
lap
to
p
m
o
d
el
is
s
u
p
p
o
r
ted
b
y
an
NVI
DI
A®
GeFo
r
ce
®
MX
1
1
0
2
GB
GD
DR
3
g
r
ap
h
ics
ca
r
d
with
a
GPU
d
r
iv
er
,
c
u
d
a
1
2
.
2
.
2
,
C
u
DNN
8
.
9
.
7
.
2
9
,
an
d
T
en
s
o
r
R
T
1
0
.
2
.
0
.
T
h
e
Evaluation Warning : The document was created with Spire.PDF for Python.
C
o
m
p
u
t Sci
I
n
f
T
ec
h
n
o
l
I
SS
N:
2722
-
3
2
2
1
A
tta
ck
d
etec
tio
n
in
i
n
tern
et
o
f
th
in
g
s
n
etw
o
r
ks w
ith
d
ee
p
lea
r
n
in
g
u
s
in
g
…
(
R
iki
A
b
d
illa
h
H
a
s
a
n
u
d
d
in
)
209
o
p
er
atin
g
s
y
s
tem
u
tili
ze
d
in
th
is
s
tu
d
y
is
Ub
u
n
tu
2
2
.
0
4
OS,
wh
ich
is
eq
u
i
p
p
ed
with
Py
th
o
n
3
.
1
1
.
0
,
T
en
s
o
r
f
lo
w
2
.
1
5
.
1
,
Scik
it
-
L
ea
r
n
,
an
d
Ker
as p
ac
k
a
g
es.
T
ab
le
5
.
T
h
e
p
ar
am
ete
r
o
f
m
o
d
el
1
an
d
m
o
d
el
2
La
y
e
r
N
a
me
M
o
d
e
l
1
M
o
d
e
l
2
N
u
mb
e
r
D
e
scri
p
t
i
o
n
N
u
mb
e
r
D
e
scri
p
t
i
o
n
I
n
p
u
t
I
n
p
u
t
l
a
y
e
r
1
2
4
f
i
t
u
r
1
2
4
f
i
t
u
r
H
i
d
d
e
n
l
a
y
e
r
C
o
n
v
1
D
2
F
i
l
t
e
r
=
6
4
,
P
a
d
d
i
n
g
=
sam
e
,
K
e
r
n
e
l
_
s
i
z
e
=
2
4
,
A
c
t
i
v
a
t
i
o
n
=
R
e
L
u
N
o
r
mal
i
z
a
t
i
o
n
2
M
a
x
P
o
o
l
1
D
2
P
o
o
l
_
s
i
z
e
=
2
,
P
a
d
d
i
n
g
=
s
a
me
S
p
a
t
i
a
l
d
r
o
p
o
u
t
2
R
a
t
e
=
0
.
0
5
C
l
a
s
si
f
i
c
a
t
i
o
n
D
e
n
se
1
N
e
u
r
o
n
=
5
1
2
,
A
c
t
i
v
a
t
i
o
n
=
R
e
l
u
1
N
e
u
r
o
n
=
5
1
2
,
A
c
t
i
v
a
t
i
o
n
=
R
e
l
u
,
D
r
o
p
o
u
t
=0
.
4
1
N
e
u
r
o
n
=
2
5
6
,
A
c
t
i
v
a
t
i
o
n
=
R
e
l
u
,
D
r
o
p
o
u
t
=0
.
4
1
N
e
u
r
o
n
=
1
2
8
,
A
c
t
i
v
a
t
i
o
n
=
R
e
l
u
,
D
r
o
p
o
u
t
=0
.
2
O
u
t
p
u
t
O
u
t
p
u
t
l
a
y
e
r
1
N
e
u
r
o
n
=
1
,
A
c
t
i
v
a
t
i
o
n
=
S
i
g
m
o
i
d
1
N
e
u
r
o
n
=
1
,
A
c
t
i
v
a
t
i
o
n
=
S
i
g
m
o
i
d
H
y
p
e
r
p
a
r
a
me
t
e
r
s
Ea
r
l
y
s
t
o
p
p
i
n
g
(
mo
n
i
t
o
r
=
v
a
l
_
l
o
ss,
p
a
t
i
e
n
c
e
=
5
)
O
p
t
i
mi
z
e
r
s
=
A
d
a
m
Lo
ss
f
u
n
c
t
i
o
n
=
B
i
n
a
r
y
C
r
o
sse
n
t
r
o
p
h
y
Le
a
r
n
i
n
g
r
a
t
e
=
2
e
-
5
B
a
t
c
h
si
z
e
=
2
5
6
E
p
o
c
h
=
1
5
Ea
r
l
y
s
t
o
p
p
i
n
g
(
mo
n
i
t
o
r
=
v
a
l
_
l
o
ss,
p
a
t
i
e
n
c
e
=
5
)
,
o
p
t
i
m
i
z
e
r
s
=
A
d
a
m
Lo
ss
f
u
n
c
t
i
o
n
=
B
i
n
a
r
y
C
r
o
sse
n
t
r
o
p
h
y
Le
a
r
n
i
n
g
r
a
te
=
6
e
-
5
B
a
t
c
h
si
z
e
=
4
0
9
6
E
p
o
c
h
=
5
0
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
3
.
1
.
E
v
a
lua
t
i
o
n m
o
del 1
a
s
ba
s
e
m
o
del
Mo
d
el
1
as
b
ase
m
o
d
el
is
tr
ai
n
ed
u
s
in
g
th
e
5
%
B
o
T
-
I
o
T
h
as
an
o
p
tim
al
attac
k
d
etec
tio
n
r
ate
with
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all,
an
d
F1
-
s
co
r
e
ac
h
iev
e
1
0
0
%
an
d
MCC
o
f
0
.
9
5
0
2
m
ea
n
in
g
th
a
t
alm
o
s
t
all
tr
af
f
ic
ca
n
b
e
d
etec
ted
ac
cu
r
ately
.
H
o
wev
er
,
s
o
m
e
n
o
r
m
al
tr
af
f
ic
is
id
en
tifie
d
b
y
th
e
m
o
d
el
as
attac
k
tr
af
f
ic,
a
n
d
s
o
m
e
attac
k
tr
af
f
ic,
p
ar
ticu
la
r
ly
r
ec
o
n
n
aiss
an
ce
ty
p
e
tr
af
f
ic,
is
d
etec
ted
as
n
o
r
m
al
tr
af
f
ic.
T
h
e
co
m
p
lete
r
esu
lts
o
f
tr
af
f
ic
d
etec
tio
n
f
r
o
m
th
is
m
o
d
el
ar
e
s
h
o
wn
in
T
a
b
les 6
an
d
7.
T
ab
le
6
.
Per
f
o
r
m
an
ce
o
f
b
ase
m
o
d
el
N
a
me
D
e
t
e
c
t
e
d
N
o
t
d
e
t
e
c
t
e
d
%
D
e
t
e
c
t
e
d
N
o
r
mal
1
0
4
2
9
8
.
1
1
D
D
o
s
3
8
5
,
2
8
2
0
1
0
0
.
0
0
D
o
s
3
3
0
,
0
9
7
0
1
0
0
.
0
0
R
e
c
o
n
n
a
i
ss
a
n
c
e
1
6
,
4
3
8
9
9
9
.
9
5
I
n
f
o
r
mat
i
o
n
t
h
e
f
t
16
0
1
0
0
.
0
0
T
ab
le
7
.
Me
tr
ics o
f
b
ase
m
o
d
e
l
M
e
t
r
i
c
s
V
a
l
u
e
A
c
c
u
r
a
c
y
1
0
0
.
0
0
%
P
r
e
c
i
s
i
o
n
1
0
0
.
0
0
%
R
e
c
a
l
l
1
0
0
.
0
0
%
F1
-
S
c
o
r
e
1
0
0
.
0
0
%
M
C
C
0
.
9
5
0
2
3
.
2
.
E
v
a
lua
t
i
o
n m
o
del 1
wit
ho
ut
t
ra
ns
f
er
lea
rning
I
n
th
is
s
tag
e,
m
o
d
el
1
is
tr
ai
n
ed
u
s
in
g
th
e
UNSW
-
N
B
1
5
-
B
ase4
_
tr
ain
d
ataset
an
d
tes
ted
u
s
in
g
th
e
UNSW
-
N
B
1
5
-
f
ir
s
t_
test
d
atas
et.
T
h
e
d
ataset
co
n
tain
s
ty
p
es
o
f
attac
k
s
th
at
h
a
v
e
n
o
t
b
ee
n
tr
ain
ed
(
u
n
k
n
o
wn
attac
k
s
)
,
allo
win
g
f
o
r
th
e
ass
ess
m
en
t o
f
m
o
d
el
1
'
s
ab
ilit
y
to
d
etec
t n
ew
ty
p
es o
f
attac
k
s
.
Mo
d
el
1
is
ca
p
ab
le
o
f
d
etec
tin
g
n
o
r
m
al
tr
a
f
f
ic
well,
b
u
t
f
o
r
u
n
k
n
o
wn
attac
k
,
esp
ec
ially
f
u
zz
er
s
an
d
an
aly
s
is
,
it
h
as
a
lo
w
d
etec
tio
n
r
ate
o
f
4
8
.
1
7
%
a
n
d
4
1
.
6
4
%
r
e
s
p
ec
tiv
ely
.
Fu
r
th
er
m
o
r
e
,
m
o
d
el
1
will
b
e
test
ed
with
UNSW
-
NB
1
5
_
f
u
ll_
test
to
d
eter
m
in
e
th
e
ex
ten
t
to
wh
ich
d
etec
tin
g
k
n
o
wn
attac
k
s
an
d
u
n
k
n
o
wn
attac
k
s
.
T
h
e
ca
p
ab
il
ity
o
f
m
o
d
el
1
f
o
r
attac
k
d
etec
tio
n
is
q
u
ite
g
o
o
d
.
Ho
wev
er
,
th
e
r
e
ar
e
s
till
ty
p
e
s
o
f
attac
k
s
th
at
r
em
ain
i
n
ad
e
q
u
ately
d
etec
tab
le,
s
u
ch
as
b
ac
k
d
o
o
r
an
d
s
h
el
lco
d
e,
it
h
as
a
lo
w
d
etec
tio
n
r
ate
o
f
8
2
.
0
7
%
an
d
8
4
.
6
5
%.
T
h
e
co
m
p
lete
r
esu
lts
o
f
tr
af
f
ic
d
etec
tio
n
f
r
o
m
m
o
d
el
1
ar
e
s
h
o
wn
in
T
ab
le
8
.
3
.
3
.
E
v
a
lua
t
i
o
n m
o
del 2
(
mo
del 1
wit
h t
ra
ns
f
er
lea
rning
)
At
th
is
s
tag
e,
m
o
d
el
2
is
co
n
s
t
r
u
cted
wh
er
e
m
o
d
el
1
is
en
h
an
ce
d
with
tr
an
s
f
er
lear
n
in
g
,
al
lo
win
g
f
o
r
u
n
d
er
s
tan
d
i
n
g
th
at
th
e
u
s
e
o
f
tr
an
s
f
er
lear
n
in
g
ca
n
ass
is
t
m
o
d
el
2
in
d
etec
tin
g
b
o
th
k
n
o
wn
an
d
u
n
k
n
o
wn
attac
k
s
m
o
r
e
ef
f
ec
tiv
ely
co
m
p
ar
ed
to
m
o
d
el
1
.
I
n
f
ir
s
t te
s
t u
s
in
g
UNSW
-
N
B
1
5
_
f
ir
s
t_
test
d
ataset,
m
o
d
el
1
h
as
ab
ilit
y
to
d
etec
t
u
n
k
n
o
wn
attac
k
s
an
d
h
as
im
p
r
o
v
ed
b
ec
au
s
e
o
f
tr
an
s
f
er
lea
r
n
in
g
c
o
m
p
a
r
ed
to
with
o
u
t
it.
On
ly
th
e
f
u
zz
er
s
an
d
wo
r
m
s
h
av
e
a
d
etec
tio
n
r
ate
o
f
9
6
%,
wh
ile
o
th
er
s
h
av
e
r
ates
ab
o
v
e
9
7
%.
T
h
e
s
ec
o
n
d
test
wil
l
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
7
2
2
-
3
2
2
1
C
o
m
p
u
t Sci
I
n
f
T
ec
h
n
o
l
,
Vo
l.
6
,
No
.
2
,
J
u
ly
2
0
2
5
:
20
2
-
2
1
3
210
u
s
e
th
e
UNSW
-
N
B
1
5
_
f
u
ll_
te
s
t
d
ataset,
wh
ich
in
clu
d
es
b
o
th
k
n
o
wn
a
n
d
u
n
k
n
o
w
n
attac
k
s
.
Fo
r
u
n
k
n
o
wn
attac
k
ty
p
es,
a
n
aly
s
is
h
as
th
e
h
ig
h
est d
etec
tio
n
r
ate
at
9
9
.
8
4
%,
wh
ile
b
ac
k
d
o
o
r
s
h
av
e
th
e
lo
west d
etec
tio
n
r
ate
at
8
9
.
3
6
%.
Fo
r
k
n
o
wn
attac
k
d
et
ec
tio
n
s
,
r
ec
o
n
n
aiss
an
ce
h
as th
e
h
ig
h
est d
etec
tio
n
r
ate
at
9
9
.
9
4
%,
an
d
e
x
p
l
o
its
h
av
e
th
e
lo
west
d
etec
tio
n
r
ate
at
9
4
.
5
3
%.
T
h
e
co
m
p
lete
r
es
u
lts
o
f
th
e
d
etec
tio
n
r
ate
f
r
o
m
m
o
d
el
2
ar
e
s
h
o
wn
in
T
ab
le
9
.
T
ab
le
8
.
Dete
ctio
n
r
ate
in
m
o
d
el
1
with
o
u
t
tr
an
s
f
e
r
lea
r
n
in
g
N
a
me
U
N
S
W
-
N
B
1
5
-
f
i
r
s
t
_
t
e
s
t
U
N
S
W
-
N
B
1
5
_
f
u
l
l
_
t
e
st
D
e
t
e
c
t
e
d
N
o
t
d
e
t
e
c
t
e
d
%
D
e
t
e
c
t
e
d
D
e
t
e
c
t
e
d
N
o
t
d
e
t
e
c
t
e
d
%
D
e
t
e
c
t
e
d
N
o
r
mal
7
4
,
5
1
8
7
9
4
9
8
.
9
5
3
1
6
,
0
8
6
4
,
5
0
7
9
8
.
5
9
G
e
n
e
r
i
c
2
1
3
,
5
4
5
1
7
3
9
9
.
9
2
Ex
p
l
o
i
t
s
1
8
,
5
5
6
9
,
7
6
1
6
5
.
5
3
2
5
,
7
4
1
2
,
5
7
6
9
0
.
9
0
F
u
z
z
e
r
s
1
0
,
3
6
5
1
1
,
1
5
2
4
8
.
1
7
2
0
,
6
4
3
8
7
4
9
5
.
9
4
R
e
c
o
n
n
a
i
ss
a
n
c
e
1
1
,
8
3
5
19
9
9
.
8
4
D
o
S
3
,
6
5
0
2
1
3
9
4
.
4
9
S
h
e
l
l
c
o
d
e
1
,
2
9
7
2
1
4
8
5
.
8
4
1
,
2
7
9
2
3
2
8
4
.
6
5
A
n
a
l
y
s
i
s
2
5
9
3
6
3
4
1
.
6
4
6
0
7
15
9
7
.
5
9
B
a
c
k
d
o
o
r
2
4
4
1
1
3
6
8
.
3
5
2
9
3
64
8
2
.
0
7
W
o
r
ms
67
67
5
0
.
0
0
1
6
9
5
9
7
.
1
3
T
ab
le
9
.
Dete
ctio
n
r
ate
in
m
o
d
el
2
(
m
o
d
el
1
with
tr
a
n
s
f
er
lea
r
n
in
g
)
N
a
me
U
N
S
W
-
N
B
1
5
-
f
i
r
s
t
_
t
e
s
t
U
N
S
W
-
N
B
1
5
_
f
u
l
l
_
t
e
st
D
e
t
e
c
t
e
d
N
o
t
d
e
t
e
c
t
e
d
%
D
e
t
e
c
t
e
d
D
e
t
e
c
t
e
d
N
o
t
d
e
t
e
c
t
e
d
%
D
e
t
e
c
t
e
d
N
o
r
mal
7
2
,
9
0
3
2
,
4
0
9
9
6
.
8
0
3
1
6
,
0
3
9
4
,
5
5
4
9
8
.
5
8
G
e
n
e
r
i
c
2
1
3
,
6
9
2
26
9
9
.
9
9
Ex
p
l
o
i
t
s
2
7
,
5
8
8
7
2
9
9
7
.
4
3
2
6
,
7
6
8
1
,
5
4
9
9
4
.
5
3
F
u
z
z
e
r
s
2
0
,
7
2
5
7
9
2
9
6
.
3
2
2
1
,
0
1
7
5
0
0
9
7
.
6
8
R
e
c
o
n
n
a
i
ss
a
n
c
e
1
1
,
8
4
7
7
9
9
.
9
4
D
o
S
3
,
7
7
9
84
9
7
.
8
3
S
h
e
l
l
c
o
d
e
1
4
7
1
40
9
7
.
3
5
1
,
3
5
9
1
5
2
8
9
.
9
4
A
n
a
l
y
s
i
s
6
2
1
1
9
9
.
8
4
6
2
1
1
9
9
.
8
4
B
a
c
k
d
o
o
r
3
5
0
7
9
8
.
0
4
3
1
9
38
8
9
.
3
6
W
o
r
ms
1
6
8
6
9
6
.
5
5
1
6
9
5
9
7
.
1
3
3
.
4
.
Va
lid
a
t
i
o
n o
f
t
he
m
o
del
First,
th
e
d
etec
tio
n
r
ate
f
r
o
m
b
o
th
test
s
co
m
p
ar
ed
an
d
in
d
ic
ated
a
s
ig
n
if
ican
t
im
p
r
o
v
em
e
n
t.
Ov
er
all,
in
f
ir
s
t
test
,
UNSW
-
NB
1
5
_
f
ir
s
t_
test
d
ataset
u
s
ed
,
ea
ch
t
y
p
e
o
f
u
n
k
n
o
wn
attac
k
s
h
o
wed
im
p
r
o
v
em
e
n
t,
p
ar
ticu
lar
ly
f
o
r
a
n
aly
s
is
,
wh
ic
h
in
cr
ea
s
ed
b
y
5
8
.
2
0
%
f
r
o
m
4
1
.
6
4
%
to
9
9
.
8
4
%,
an
d
f
u
zz
e
r
s
,
wh
ich
in
c
r
ea
s
ed
b
y
4
8
.
1
5
%
f
r
o
m
4
8
.
1
7
%
to
9
6
.
3
2
%,
w
h
ile
th
e
d
etec
tio
n
o
f
n
o
r
m
al
tr
af
f
ic
h
as
d
ec
r
ea
s
ed
f
r
o
m
9
8
.
9
5
t
o
9
6
.
8
0
%.
T
h
e
d
etec
tio
n
o
f
n
o
r
m
al
tr
af
f
ic
is
s
lig
h
tly
l
o
wer
b
e
ca
u
s
e
th
e
tr
ai
n
in
g
o
f
m
o
d
el
1
was
co
n
d
u
cte
d
with
th
e
B
o
T
-
I
o
T
d
ataset,
wh
ich
h
as
a
v
er
y
s
m
a
ll
p
er
ce
n
ta
g
e
o
f
n
o
r
m
al
tr
af
f
ic
co
m
p
ar
e
d
to
attac
k
tr
af
f
ic.
I
n
s
ec
o
n
d
test
u
s
ed
UNSW
-
NB
1
5
_
f
u
ll_
test
to
ass
ess
its
ca
p
ab
ilit
y
in
d
etec
tin
g
b
o
th
k
n
o
wn
a
n
d
u
n
k
n
o
wn
attac
k
s
.
Ov
er
all,
ea
ch
ty
p
e
o
f
u
n
k
n
o
wn
attac
k
h
as
also
s
h
o
w
n
an
in
cr
ea
s
e,
with
u
n
k
n
o
w
n
a
ttack
ty
p
es
s
u
ch
as
b
ac
k
d
o
o
r
s
an
d
ex
p
lo
its
ex
h
ib
i
tin
g
a
s
ig
n
if
ican
t
im
p
r
o
v
em
en
t
o
f
7
.
7
2
%
an
d
5
.
2
9
%
wh
ile
f
o
r
k
n
o
wn
attac
k
s
,
Do
S
in
cr
ea
s
ed
b
y
3
.
3
4
%
f
r
o
m
9
4
.
4
9
%
to
9
7
.
8
3
%,
r
esp
ec
tiv
ely
.
Ho
wev
er
,
th
e
d
etec
tio
n
r
a
te
o
f
n
o
r
m
al
tr
af
f
ic
s
till
s
l
ig
h
tly
d
ec
r
ea
s
ed
b
y
-
0
.
0
1
%.
T
h
e
co
m
p
lete
r
esu
lts
o
f
th
e
co
m
p
ar
is
o
n
an
d
im
p
r
o
v
em
e
n
t
d
etec
tio
n
r
ates
f
r
o
m
b
o
th
m
o
d
els
ar
e
s
h
o
wn
in
T
ab
le
1
0
.
T
h
e
f
o
llo
win
g
co
n
clu
s
io
n
s
ca
n
b
e
d
r
a
wn
f
r
o
m
th
is
d
ata.
I
t
h
as
b
ee
n
d
em
o
n
s
tr
ated
t
h
at
Mo
d
el
2
,
wh
en
e
m
p
lo
y
i
n
g
th
e
tr
an
s
f
er
lear
n
in
g
m
eth
o
d
,
ex
h
ib
its
an
en
h
an
ce
d
d
etec
tio
n
r
a
te
f
o
r
b
o
th
k
n
o
wn
an
d
u
n
k
n
o
wn
attac
k
s
.
T
h
e
k
n
o
wn
attac
k
d
etec
tio
n
r
ate
r
an
g
es
f
r
o
m
9
7
.
8
3
%
to
9
9
.
9
9
%,
with
th
e
u
n
k
n
o
wn
attac
k
d
etec
tio
n
r
ate
r
an
g
in
g
f
r
o
m
8
9
.
3
6
%
to
9
9
.
8
4
%.
A
co
m
p
ar
is
o
n
o
f
m
o
d
els
1
an
d
2
r
e
v
ea
ls
th
at
th
e
latter
,
u
tili
zin
g
tr
an
s
f
er
lear
n
i
n
g
,
en
h
an
ce
s
th
e
d
etec
tio
n
r
ate
f
o
r
u
n
k
n
o
wn
attac
k
s
b
y
u
p
to
5
8
.
2
0
%,
an
d
f
o
r
k
n
o
w
n
attac
k
s
b
y
u
p
to
7
.
2
9
%.
T
h
er
e
f
o
r
e,
m
o
d
el
2
with
tr
a
n
s
f
er
lea
r
n
in
g
d
em
o
n
s
tr
ates
a
s
ig
n
if
ican
t
en
h
a
n
ce
m
en
t
in
th
e
d
etec
tio
n
o
f
b
o
th
k
n
o
wn
an
d
u
n
k
n
o
w
n
attac
k
s
,
d
esp
ite
th
e
r
ed
u
ce
d
r
ep
r
esen
tatio
n
in
th
e
d
ataset,
wh
en
co
m
p
ar
ed
with
th
e
m
o
d
el.
C
o
m
p
ar
is
o
n
was
m
ad
e
b
etwe
en
th
e
ev
alu
atio
n
m
etr
ics
f
r
o
m
b
o
th
test
s
.
Mo
d
el
1
was
in
itiall
y
ev
alu
ated
f
o
r
its
ca
p
ac
ity
to
d
etec
t
u
n
k
n
o
wn
s
u
s
in
g
th
e
UN
SW
-
N
B
1
5
_
f
ir
s
t_
test
d
ataset.
T
h
e
m
o
d
el'
s
o
v
er
all
ac
cu
r
ac
y
was
f
o
u
n
d
to
b
e
8
2
.
4
2
%,
with
a
p
r
ec
is
io
n
o
f
5
8
.
6
9
%,
a
r
ec
all
o
f
9
7
.
4
9
%,
an
F1
-
s
co
r
e
o
f
6
3
.
6
4
%,
an
d
an
MCC
o
f
0
.
6
5
7
3
.
Mo
d
el
1
was
also
ev
alu
ated
f
o
r
th
e
d
etec
tio
n
o
f
b
o
th
u
n
k
n
o
w
n
an
d
k
n
o
wn
attac
k
s
u
s
in
g
th
e
UNSW
-
N
B
1
5
-
f
u
ll
_
test
d
ataset.
I
n
th
is
p
ar
ticu
lar
in
s
tan
ce
,
th
e
ac
cu
r
ac
y
r
at
e
was
r
ec
o
r
d
ed
at
Evaluation Warning : The document was created with Spire.PDF for Python.
C
o
m
p
u
t Sci
I
n
f
T
ec
h
n
o
l
I
SS
N:
2722
-
3
2
2
1
A
tta
ck
d
etec
tio
n
in
i
n
tern
et
o
f
th
in
g
s
n
etw
o
r
ks w
ith
d
ee
p
lea
r
n
in
g
u
s
in
g
…
(
R
iki
A
b
d
illa
h
H
a
s
a
n
u
d
d
in
)
211
9
8
.
5
5
%,
th
e
p
r
ec
is
io
n
r
ate
at
9
8
.
5
0
%,
t
h
e
r
ec
all
r
ate
at
9
8
.
3
4
%,
th
e
F1
-
s
co
r
e
at
8
7
.
2
9
%,
an
d
th
e
MCC
at
0
.
9
7
0
9
.
A
co
m
p
a
r
ativ
e
an
aly
s
is
o
f
th
e
o
v
er
all
m
etr
ics
f
o
r
b
o
th
s
o
lu
tio
n
s
r
ev
ea
ls
th
at
m
o
d
el
2
ex
h
ib
its
s
u
p
er
io
r
p
e
r
f
o
r
m
an
ce
in
all
m
etr
ics
wh
en
co
m
p
ar
e
d
to
m
o
d
el
1
.
T
h
e
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
an
d
F1
-
s
co
r
e
m
etr
ics
h
av
e
b
ee
n
e
n
h
an
ce
d
b
y
1
4
.
4
9
%,
3
8
.
3
1
%,
a
n
d
2
2
.
9
5
%,
r
esp
ec
tiv
ely
,
in
th
e
d
et
ec
tio
n
o
f
u
n
k
n
o
wn
attac
k
s
.
Fu
r
th
er
m
o
r
e,
th
ese
m
etr
ics
h
av
e
b
ee
n
im
p
r
o
v
ed
b
y
0
.
3
1
%,
0
.
6
7
%,
an
d
0
.
3
4
%,
r
esp
ec
tiv
ely
,
wh
en
co
n
s
id
er
in
g
b
o
th
k
n
o
wn
an
d
u
n
k
n
o
wn
attac
k
s
.
E
v
alu
atio
n
m
etr
ics
f
r
o
m
b
o
th
test
s
ar
e
co
m
p
ar
ed
.
Mo
d
el
1
was
ev
alu
ated
f
ir
s
t
o
n
th
e
d
etec
tio
n
o
f
u
n
k
n
o
wn
with
th
e
UNSW
-
NB
1
5
_
f
ir
s
t_
test
d
ataset.
I
t
ac
h
ie
v
es
an
o
v
er
all
ac
c
u
r
ac
y
o
f
8
2
.
4
2
%,
a
p
r
ec
is
io
n
o
f
5
8
.
6
9
%,
a
r
ec
all
o
f
9
7
.
4
9
%,
an
F1
-
s
co
r
e
o
f
6
3
.
6
4
%,
an
d
an
MCC
o
f
0
.
6
5
7
3
.
Mo
d
el
1
was
also
ev
alu
ated
f
o
r
th
e
d
etec
tio
n
o
f
b
o
th
u
n
k
n
o
w
n
an
d
k
n
o
wn
attac
k
s
u
s
in
g
th
e
UNSW
-
N
B
1
5
-
f
u
ll_
test
d
ata
s
et.
I
n
th
is
ca
s
e,
th
e
ac
cu
r
ac
y
is
9
8
.
5
5
%,
p
r
ec
is
io
n
is
9
8
.
5
0
%,
r
ec
all
is
9
8
.
3
4
%,
F1
-
s
co
r
e
i
s
8
7
.
2
9
%
,
an
d
MCC
is
0
.
9
7
0
9
.
I
f
th
e
o
v
er
all
m
etr
ics
f
o
r
b
o
t
h
s
o
lu
tio
n
s
ar
e
co
m
p
ar
e
d
,
we
ca
n
co
n
clu
d
e
th
at
m
o
d
el
2
o
u
tp
er
f
o
r
m
s
m
o
d
el
1
f
o
r
all
m
etr
ics.
Acc
u
r
ac
y
,
p
r
ec
is
io
n
a
n
d
F1
-
s
co
r
e
ar
e
im
p
r
o
v
ed
b
y
1
4
.
4
9
%,
3
8
.
3
1
%
,
an
d
2
2
.
9
5
%
in
th
e
d
etec
tio
n
o
f
u
n
k
n
o
wn
,
a
n
d
0
.
3
1
%,
0
.
6
7
%,
an
d
0
.
3
4
%
r
esp
ec
tiv
ely
wh
e
n
co
n
s
id
er
in
g
b
o
th
k
n
o
wn
an
d
u
n
k
n
o
wn
attac
k
s
.
T
h
e
p
r
im
ar
y
f
ac
to
r
co
n
t
r
ib
u
ti
n
g
to
th
is
en
h
an
ce
m
en
t
is
th
e
f
aster
co
n
v
er
g
en
ce
o
f
th
e
o
p
ti
m
izatio
n
alg
o
r
ith
m
to
o
p
tim
al
weig
h
ts
in
m
o
d
el
2
wh
en
th
e
weig
h
ts
ar
e
in
it
ial
ized
in
a
s
im
ilar
d
o
m
ain
.
T
h
is
co
n
tr
asts
with
m
o
d
el
1
,
wh
ich
c
o
m
m
en
ce
s
f
r
o
m
th
e
b
eg
in
n
in
g
an
d
u
tili
ze
s
r
an
d
o
m
weig
h
ts
.
An
o
th
er
m
et
r
ic
th
at
h
as
s
h
o
wn
n
o
tab
le
im
p
r
o
v
em
e
n
t
is
MCC
.
Mo
d
el
2
d
em
o
n
s
tr
ates
h
ig
h
er
MCC
v
alu
es
d
u
e
t
o
two
p
r
im
ar
y
f
ac
to
r
s
:
th
e
o
cc
u
r
r
e
n
ce
o
f
u
n
k
n
o
wn
attac
k
s
an
d
th
e
im
b
ala
n
ce
d
n
at
u
r
e
o
f
th
e
d
ataset.
I
t
ca
n
b
e
co
n
clu
d
ed
f
r
o
m
th
ese
f
ac
to
r
s
th
at
m
o
d
el
2
m
ak
es
s
u
p
er
io
r
p
r
ed
ictio
n
s
in
ter
m
s
o
f
attac
k
d
etec
tio
n
.
T
h
is
is
a
cr
itical
co
n
s
id
er
atio
n
i
n
ad
ap
tin
g
th
e
m
o
d
el
to
d
if
f
er
en
t I
o
T
d
ev
ice
tr
a
f
f
ic
p
atter
n
s
.
As d
em
o
n
s
tr
ated
in
T
ab
le
1
1
,
th
is
s
tu
d
y
p
r
o
v
id
es a
co
m
p
r
eh
e
n
s
iv
e
ev
alu
atio
n
o
f
t
h
e
p
er
f
o
r
m
an
ce
m
etr
ics f
r
o
m
m
o
d
els 1
an
d
2
.
T
ab
le
1
0
.
C
o
m
p
ar
is
o
n
a
n
d
im
p
r
o
v
em
e
n
t d
etec
tio
n
r
ate
in
m
o
d
el
1
a
n
d
m
o
d
el
2
N
a
me
U
N
S
W
-
N
B
1
5
-
f
i
r
s
t
_
t
e
s
t
U
N
S
W
-
N
B
1
5
-
f
u
l
l
_
t
e
st
M
o
d
e
l
1
(
%)
M
o
d
e
l
2
(
%)
I
mp
r
o
v
e
(
%)
M
o
d
e
l
1
(
%)
M
o
d
e
l
2
(
%)
I
mp
r
o
v
e
(
%)
N
o
r
mal
9
8
.
9
5
9
6
.
8
0
-
2
.
1
5
9
8
.
5
9
9
8
.
5
8
-
0
.
0
1
G
e
n
e
r
i
c
9
9
.
9
2
9
9
.
9
9
0
.
0
7
Ex
p
l
o
i
t
s
6
5
.
5
3
9
7
.
4
3
3
1
.
9
0
9
0
.
9
0
9
4
.
5
3
3
.
6
3
F
u
z
z
e
r
s
4
8
.
1
7
9
6
.
3
2
4
8
.
1
5
9
5
.
9
4
9
7
.
6
8
1
.
7
4
R
e
c
o
n
n
a
i
ss
a
n
c
e
9
9
.
8
4
9
9
.
9
4
0
.
1
D
o
S
9
4
.
4
9
9
7
.
8
3
3
.
3
4
S
h
e
l
l
c
o
d
e
8
5
.
8
4
9
7
.
3
5
1
1
.
5
1
8
4
.
6
5
8
9
.
9
4
5
.
2
9
A
n
a
l
y
s
i
s
4
1
.
6
4
9
9
.
8
4
5
8
.
2
0
9
7
.
5
9
9
9
.
8
4
2
.
2
5
B
a
c
k
d
o
o
r
6
8
.
3
5
9
8
.
0
4
2
9
.
6
9
8
2
.
0
7
8
9
.
3
6
7
.
2
9
W
o
r
ms
5
0
.
0
0
9
6
.
5
5
4
6
.
5
5
9
7
.
1
3
9
7
.
1
3
0
T
ab
le
1
1
.
C
o
m
p
ar
is
o
n
o
f
ev
al
u
atio
n
m
etr
ics f
r
o
m
m
o
d
el
1
a
n
d
m
o
d
el
2
M
e
t
r
i
c
U
N
S
W
-
N
B
1
5
-
f
i
r
s
t
_
t
e
s
t
U
N
S
W
-
N
B
1
5
-
f
u
l
l
_
t
e
st
M
o
d
e
l
1
M
o
d
e
l
2
I
mp
r
o
v
e
m
e
n
t
M
o
d
e
l
1
M
o
d
e
l
2
(
%)
I
mp
r
o
v
e
m
e
n
t
A
c
c
u
r
a
c
y
8
2
.
4
2
%
9
6
.
8
8
%
1
4
.
4
9
%
9
8
.
5
5
%
9
8
.
8
6
%
0
.
3
1
%
P
r
e
c
i
s
i
o
n
5
8
.
6
9
%
9
7
.
0
0
%
3
8
.
3
1
%
9
8
.
5
0
%
9
9
.
1
7
%
0
.
6
7
%
R
e
c
a
l
l
9
7
.
4
9
%
9
5
.
4
8
%
-
2
.
0
1
%
9
8
.
3
4
%
9
8
.
4
0
%
0
.
0
6
%
F1
-
S
c
o
r
e
7
3
.
2
7
%
9
6
.
2
2
%
2
2
.
9
5
%
9
8
.
4
4
%
9
8
.
7
2
%
0
.
3
4
%
M
C
C
0
.
6
5
7
3
0
.
9
3
5
8
0
.
2
7
8
5
0
.
9
7
0
9
0
.
9
7
7
0
0
.
0
0
6
1
4.
CO
NCLU
SI
O
N
I
n
t
h
i
s
r
es
e
a
r
c
h
,
w
e
i
n
v
es
t
i
g
ate
t
h
e
d
e
t
e
ct
i
o
n
c
a
p
a
b
il
i
ti
e
s
o
f
t
r
a
n
s
f
e
r
l
ea
r
n
i
n
g
-
b
a
s
e
d
m
e
t
h
o
d
s
f
o
r
b
o
t
h
u
n
k
n
o
w
n
a
n
d
k
n
o
w
n
a
t
t
a
c
k
s
in
I
o
T
n
e
t
w
o
r
k
s
w
i
t
h
s
c
a
r
ce
a
n
d
u
n
b
a
l
a
n
c
e
d
d
a
t
a
s
e
ts
.
T
o
t
h
i
s
e
n
d
,
w
e
d
e
v
e
l
o
p
a
n
e
f
f
i
c
i
e
n
t
a
tt
a
c
k
d
e
t
e
ct
i
o
n
f
r
a
m
e
w
o
r
k
t
h
a
t
c
o
m
b
i
n
e
s
k
n
o
w
l
e
d
g
e
t
r
a
n
s
f
e
r
a
n
d
m
o
d
e
l
r
e
f
i
n
e
m
e
n
t
,
a
c
h
i
e
v
i
n
g
h
i
g
h
d
e
t
e
c
t
i
o
n
a
c
c
u
r
ac
y
f
o
r
b
o
t
h
k
n
o
w
n
a
n
d
u
n
k
n
o
w
n
a
t
t
a
c
k
s
.
T
h
e
B
o
T
-
I
o
T
d
a
t
as
e
t
i
s
u
ti
l
iz
e
d
to
l
e
a
r
n
k
n
o
w
l
e
d
g
e
(
s
o
u
r
c
e
d
o
m
a
i
n
)
a
n
d
is
a
p
p
l
i
ed
t
o
t
h
e
U
NSW
-
NB
1
5
d
a
t
as
e
t
(
t
a
r
g
e
t
d
o
m
a
i
n
)
.
T
o
e
v
al
u
a
t
e
th
e
p
e
r
f
o
r
m
a
n
c
e
a
n
d
f
e
a
s
i
b
il
i
t
y
o
f
t
h
e
p
r
o
p
o
s
e
d
t
r
a
n
s
f
e
r
l
e
a
r
n
i
n
g
-
b
a
s
e
d
m
o
d
e
l
,
w
e
g
e
n
e
r
a
t
e
U
NSW
-
NB
1
5
_
f
i
r
s
t
-
t
es
t
d
a
t
as
e
t
t
o
e
v
a
l
u
a
t
e
e
f
f
e
ct
i
v
e
n
es
s
i
n
d
e
te
c
t
i
o
n
o
f
u
n
k
n
o
w
n
a
t
t
a
c
k
s
a
n
d
U
N
SW
-
NB
1
5
_
f
u
l
l
-
t
es
t
d
a
t
a
s
et
t
o
e
v
al
u
a
t
e
e
f
f
e
c
t
i
v
e
n
es
s
i
n
d
e
te
c
t
i
o
n
o
f
k
n
o
w
n
a
n
d
u
n
k
n
o
w
n
a
t
t
a
c
k
s
.
W
e
f
i
n
d
t
h
a
t
t
r
a
n
s
f
e
r
l
ea
r
n
in
g
a
n
d
f
i
n
e
-
tu
n
i
n
g
i
m
p
r
o
v
e
t
h
e
m
o
d
e
l
f
o
r
t
h
e
d
e
t
e
c
t
i
o
n
o
f
i
n
d
e
t
e
c
t
i
o
n
o
f
k
n
o
w
n
a
n
d
u
n
k
n
o
w
n
a
t
t
a
c
k
s
.
T
h
e
e
x
p
e
r
i
m
e
n
t
a
l
r
e
s
u
lts
s
h
o
w
t
h
a
t
t
h
e
t
r
a
n
s
f
e
r
l
e
a
r
n
i
n
g
-
b
a
s
e
d
m
o
d
e
l
a
c
h
i
e
v
e
s
h
i
g
h
a
c
c
u
r
a
c
y
,
p
r
e
c
i
s
i
o
n
,
F
1
-
s
c
o
r
e
a
n
d
M
C
C
.
F
o
r
f
u
t
u
r
e
r
e
s
e
a
r
c
h
,
g
r
e
a
t
e
r
e
f
f
o
r
ts
wi
l
l
b
e
m
a
d
e
t
o
e
x
p
l
o
r
e
a
n
d
i
m
p
l
em
e
n
t
a
d
v
a
n
c
e
d
a
r
c
h
i
t
e
ct
u
r
e
s
o
f
t
r
a
n
s
f
e
r
l
e
a
r
n
i
n
g
-
b
a
s
e
d
m
o
d
e
l
s
i
n
a
d
d
r
e
s
s
i
n
g
a
tt
a
c
k
d
e
t
e
c
t
i
o
n
is
s
u
es
b
y
u
t
i
l
i
z
in
g
s
o
u
r
c
e
a
n
d
t
a
r
g
e
t
d
o
m
a
i
n
s
t
h
a
t
h
a
v
e
d
i
f
f
e
r
e
n
t
c
l
a
s
s
l
a
b
el
s
,
i
n
o
r
d
e
r
t
o
f
u
r
t
h
e
r
e
n
h
a
n
c
e
t
h
e
p
e
r
f
o
r
m
a
n
c
e
a
n
d
f
e
a
s
i
b
il
i
t
y
o
f
I
o
T
a
tt
a
c
k
d
e
t
e
ct
i
o
n
s
y
s
t
e
m
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.