I
A
E
S
I
n
t
e
r
n
at
io
n
al
Jou
r
n
al
of
A
r
t
if
ic
ia
l
I
n
t
e
ll
ig
e
n
c
e
(
I
J
-
AI
)
V
ol
.
15
, N
o.
1
,
F
e
br
ua
r
y
2026
, pp.
580
~
591
I
S
S
N
:
2252
-
8938
,
D
O
I
:
10.11591/
ij
a
i.
v
15
.i
1
.pp
580
-
591
580
Jou
r
n
al
h
om
e
page
:
ht
tp
:
//
ij
ai
.
ia
e
s
c
or
e
.c
om
D
yn
am
i
c
at
t
ac
k
p
at
t
e
r
n
-
aw
a
r
e
i
n
t
e
l
l
i
ge
n
t
c
yb
e
r
-
p
h
ysi
c
al
i
n
t
r
u
si
on
d
e
t
e
c
t
i
on
sys
t
e
m
f
or
i
n
t
e
r
n
e
t
of
t
h
i
n
gs
-
e
d
ge
n
e
t
w
o
r
k
s
V
is
h
al
a I
b
as
ap
u
r
a L
ak
s
h
m
in
ar
ayan
ap
p
a
1
, K
e
m
p
ah
an
u
m
ai
ah
M
.
R
avi
k
u
m
ar
2
1
D
e
pa
r
t
m
e
nt
of
E
l
e
c
t
r
oni
c
s
a
nd C
om
m
uni
c
a
t
i
on, S
J
C
I
ns
t
i
t
ut
e
of
T
e
c
hnol
ogy,
B
a
nga
l
or
e
, I
ndi
a
2
D
e
pa
r
t
m
e
nt
of
E
l
e
c
t
r
oni
c
s
a
nd C
om
m
uni
c
a
t
i
on, V
i
ve
ka
na
nd I
ns
t
i
t
ut
e
of
T
e
c
hnol
ogy, B
a
nga
l
or
e
, I
ndi
a
A
r
t
ic
le
I
n
f
o
A
B
S
T
R
A
C
T
A
r
ti
c
le
h
is
to
r
y
:
R
e
c
e
iv
e
d
J
ul
24
,
2025
R
e
vi
s
e
d
D
e
c
26
,
2025
A
c
c
e
pt
e
d
J
a
n
10
,
2026
The
proliferation
of
internet
of
things
(IoT)
technologies,
coupled
w
ith
the
convergence
of
edge
computi
ng
infrastru
ctures,
has
revoluti
onized
modern
cyber
-
physical
systems
(CPS).
However,
the
inherently
distributed
architectu
re
of
these
systems
increases
their
vulnerabi
lity
to
ad
vanced
network
-
level
cyber
threats,
posing
signifi
cant
challenges
to
data
in
tegrity
and
system
reliability.
Traditional
machine
learning
(ML)
and
deep
le
arning
(DL)
-
based
intrusion
detection
systems
(IDS)
often
fall
short
in
iden
ti
fying
evolvin
g
attack
vectors
due
to
th
eir
limit
ed
adaptabil
ity.
To
address
these
limitations,
this
paper
introduces
a
novel
dynamic
attack
pattern
-
aware
improvised
weighted
gradient
boosting
(DAPA
-
IWGB)
model
desig
ned
to
enhance
real
-
time
threat
detection
and
adaptive
response
within
IoT
-
edge
-
enabled
CPS
environm
ents.
The
DAPA
-
IWGB
framework
syn
ergizes
gradient
tree
boosting
with
an
enhanced
loss
function
handling
co
variate
shift,
while
incorporating
statistical
monitoring
mechanisms
for
d
ynamic
covariate
shift
recognition
and
continuous
learning.
Compreh
ensive
experiment
al
validati
on
using
two
prominen
t
benchmark
datasets
To
N
-
IoT
and
UNSW
-
NB15
demonstrates
the
proposed
model’s
robustnes
s
and
superior
performanc
e,
achieving
detection
accurac
ies
of
99.921
%
and
99.93%,
respectively.
Comparative
evaluations
highlight
substantial
improvements
in
detection
accura
cy,
adaptability,
and
reliability
over
existin
g
IDS
soluti
ons.
The
results
affirm
the
effe
ctiveness
of
the
DAPA
-
IWGB
model
in
fortifying
the
security
pos
ture
of
distributed
IoT
-
base
d
CPS
against
sophis
ticated an
d evolvi
ng cyber t
hreats.
K
e
y
w
o
r
d
s
:
A
da
pt
iv
e
t
hr
e
a
t
de
te
c
ti
on
C
ybe
r
-
phys
ic
a
l
s
ys
t
e
m
s
E
dge
c
om
put
in
g
I
nt
e
r
ne
t
of
t
hi
ngs
I
nt
r
us
io
n de
te
c
ti
on s
ys
te
m
This is an
open
acce
ss artic
le unde
r the
CC BY
-
SA
license.
C
or
r
e
s
pon
di
n
g A
u
th
or
:
V
is
ha
la
I
ba
s
a
pur
a
L
a
k
s
hm
in
a
r
a
ya
na
ppa
D
e
pa
r
tm
e
nt
of
E
le
c
tr
oni
c
s
a
nd C
om
m
uni
c
a
ti
on, S
J
C
I
ns
ti
tu
te
o
f
T
e
c
hnol
ogy
B
a
nga
lo
r
e
, I
ndi
a
E
m
a
il
:
vi
s
ha
la
il
_12@
r
e
di
f
f
m
a
il
.c
om
1.
I
N
T
R
O
D
U
C
T
I
O
N
T
he
in
te
r
n
e
t
of
th
in
g
s
(
I
o
T
)
ha
s
e
m
e
r
g
e
d
a
s
a
tr
a
ns
f
or
m
a
ti
v
e
pa
r
a
di
gm
,
e
n
a
bl
in
g
r
e
a
l
-
ti
m
e
d
a
ta
e
xc
h
a
ng
e
a
n
d
in
t
e
ll
ig
e
nt
d
e
c
i
s
io
n
-
m
a
ki
ng
th
r
ough
in
te
r
c
onn
e
c
t
e
d
phy
s
i
c
a
l
d
e
vi
c
e
s
.
A
ppl
ic
a
ti
on
s
s
pa
n
a
c
r
os
s
s
m
a
r
t
c
it
ie
s
,
in
du
s
tr
ia
l
a
ut
o
m
a
ti
o
n,
h
e
a
lt
h
c
a
r
e
,
a
nd
v
e
hi
c
ul
a
r
ne
twor
k
s
[
1]
,
[
2]
.
T
he
s
e
de
pl
oym
e
nt
s
a
r
e
ty
pi
c
a
ll
y
s
up
por
te
d
by
e
dg
e
a
nd
c
lo
ud
c
om
put
in
g
i
nf
r
a
s
tr
uc
tu
r
e
s
to
m
a
na
g
e
l
a
te
n
c
y,
c
om
put
a
ti
on,
a
nd
s
to
r
a
g
e
li
m
it
a
ti
on
s
[
3]
,
[
4]
.
H
o
w
e
v
e
r
,
hi
ghl
y
di
s
tr
ib
ut
e
d
a
nd
he
t
e
r
oge
n
e
ous
na
t
ur
e
of
I
o
T
e
c
o
s
y
s
te
m
s
i
nt
r
odu
c
e
s
ne
w
c
ybe
r
s
e
c
ur
it
y
r
i
s
ks
.
D
e
vi
c
e
s
w
it
h
li
m
it
e
d
c
om
put
i
ng
pow
e
r
a
n
d
m
in
im
a
l
bui
l
t
-
in
s
e
c
ur
it
y
a
r
e
s
us
c
e
pt
ib
le
to
c
ybe
r
t
hr
e
a
ts
s
u
c
h a
s
s
po
of
in
g,
di
s
tr
ib
ut
e
d
de
n
ia
l
of
s
e
r
vi
c
e
(
D
D
oS
)
, a
nd
da
t
a
t
a
m
pe
r
in
g
[
5]
–
[
7]
a
s
i
n F
ig
ur
e
1.
A
m
ong
th
e
s
e
,
ne
twor
k
-
le
ve
l
a
tt
a
c
k
s
li
ke
li
nk
f
lo
odi
ng,
ba
c
k
door
s
,
s
pa
m
m
in
g,
r
a
n
s
om
w
a
r
e
,
a
nd
D
D
O
S
a
r
e
pa
r
ti
c
ul
a
r
ly
da
nge
r
ous
,
be
c
a
us
e
th
e
y
im
pa
ir
e
dge
node
a
va
il
a
bi
li
ty
a
nd
in
te
r
f
e
r
e
w
it
h
r
e
a
l
-
ti
m
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
D
y
nam
ic
at
ta
c
k
pat
te
r
n
-
aw
a
r
e
i
nt
e
ll
ig
e
nt
c
y
b
e
r
-
phy
s
ic
al
…
(
V
is
hal
a I
bas
apur
a L
ak
s
hm
in
a
r
ay
anappa
)
581
da
ta
f
lo
w
s
[
8]
–
[
10]
.
W
he
n
it
c
om
e
s
to
ke
e
pi
ng
a
n
e
ye
out
f
o
r
unus
ua
l
a
c
ti
vi
ty
in
th
e
s
e
s
e
tt
in
gs
,
in
tr
us
io
n
de
te
c
ti
on
s
ys
te
m
s
(
I
D
S
)
a
r
e
e
s
s
e
nt
ia
l
[
5]
,
[
11]
.
T
he
dyna
m
ic
t
r
a
f
f
ic
a
nd
c
ha
ngi
ng
a
tt
a
c
k
pa
tt
e
r
ns
in
I
oT
-
e
dge
de
pl
oym
e
nt
s
a
r
e
uns
ui
ta
bl
e
f
or
tr
a
di
ti
ona
l
I
D
S
te
c
hni
que
s
,
w
hi
c
h
m
os
tl
y
r
e
ly
on
r
ul
e
-
ba
s
e
d
pr
oc
e
s
s
e
s
or
s
ta
ti
c
s
ig
na
tu
r
e
s
[
12]
,
[
13]
.
T
he
us
e
of
de
e
p
le
a
r
ni
ng
(
D
L
)
a
nd
m
a
c
hi
ne
le
a
r
ni
ng
(
M
L
)
to
e
nha
nc
e
I
D
S
c
a
pa
bi
li
ti
e
s
ha
s
be
e
n
th
e
s
ubj
e
c
t
of
r
e
c
e
nt
s
tu
dy.
M
L
m
e
th
od
s
a
na
ly
z
e
ne
twor
k
tr
a
f
f
ic
pa
tt
e
r
ns
,
w
hi
le
D
L
a
r
c
hi
te
c
tu
r
e
s
c
a
n
c
a
pt
ur
e
de
e
p
hi
e
r
a
r
c
hi
c
a
l
f
e
a
tu
r
e
s
f
r
om
r
a
w
da
ta
[
12]
,
[
14]
.
D
e
s
pi
te
th
e
ir
pr
om
is
e
,
th
e
s
e
s
ys
te
m
s
f
a
c
e
c
r
it
ic
a
l
li
m
it
a
ti
ons
in
de
te
c
ti
ng
uns
e
e
n
or
e
vol
v
in
g
a
tt
a
c
ks
due
to
c
ova
r
ia
te
s
hi
f
t
w
he
r
e
d
a
ta
di
s
tr
ib
ut
io
ns
c
ha
nge
ove
r
ti
m
e
[
15]
–
[
17]
.
A
ddi
ti
ona
ll
y,
m
os
t
e
xi
s
ti
ng
I
D
S
s
ol
ut
io
ns
s
tr
uggl
e
w
it
h
r
e
a
l
-
ti
m
e
a
da
pt
a
ti
on,
s
c
a
l
a
bi
li
ty
,
a
nd
da
ta
im
ba
la
nc
e
,
r
e
s
ul
ti
ng
in
in
c
r
e
a
s
e
d
f
a
l
s
e
pos
it
iv
e
s
a
nd
m
is
s
e
d
de
te
c
ti
ons
.
T
he
r
e
f
or
e
,
a
da
pt
iv
e
a
nd
li
ght
w
e
ig
ht
I
D
S
s
ol
ut
io
ns
ta
il
or
e
d
to
e
dge
-
I
oT
a
r
c
hi
te
c
tu
r
e
s
a
r
e
e
s
s
e
nt
ia
l
f
or
r
obus
t
a
nd s
c
a
la
bl
e
c
ybe
r
-
phys
ic
a
l
s
y
s
te
m
s
(
C
P
S
)
s
e
c
ur
it
y.
F
ig
ur
e
1. B
a
s
ic
a
r
c
hi
te
c
tu
r
e
of
di
f
f
e
r
e
nt
a
tt
a
c
ks
i
nduc
e
d i
n I
oT
-
e
dge
c
om
put
in
g ne
twor
ks
T
he
de
te
c
ti
on
of
in
tr
us
io
ns
in
I
oT
ne
twor
ks
ha
s
a
tt
r
a
c
te
d
s
ig
ni
f
ic
a
nt
r
e
s
e
a
r
c
h
in
te
r
e
s
t,
w
it
h
num
e
r
ous
s
tu
di
e
s
a
ppl
yi
ng
in
te
ll
ig
e
nt
le
a
r
ni
ng
te
c
hni
que
s
f
or
r
e
a
l
-
ti
m
e
s
e
c
ur
it
y
m
oni
to
r
in
g
[
18
]
.
S
a
iy
e
d
a
nd
A
l
-
A
nba
gi
[
18]
in
tr
oduc
e
d
a
D
D
oS
de
te
c
ti
on
f
r
a
m
e
w
or
k
us
in
g
ge
ne
ti
c
a
lg
or
it
hm
s
a
nd
s
ta
ti
s
ti
c
a
l
te
s
ti
ng,
de
m
ons
tr
a
ti
ng
hi
gh
pr
e
c
is
io
n
on
I
oT
tr
a
f
f
ic
da
ta
s
e
ts
.
C
ui
e
t
al
.
[
19]
pr
opos
e
d a
de
e
p
r
e
s
id
ua
l
ne
twor
k
w
it
h
a
tt
e
nt
io
n
m
e
c
ha
ni
s
m
s
(
D
R
N
-
A
M
)
f
or
im
pr
ove
d
de
te
c
ti
on
in
m
ul
ti
-
de
vi
c
e
s
e
tt
in
gs
.
J
a
ve
e
d
e
t
al
.
[
20]
a
ddr
e
s
s
e
d
in
tr
us
io
n
de
te
c
ti
on
in
s
m
a
r
t
a
gr
ic
ul
tu
r
e
us
in
g
e
dge
-
ba
s
e
d
le
a
r
ni
ng
(
I
D
S
-
S
A
E
L
)
f
r
a
m
e
w
or
ks
f
o
r
hos
ti
le
e
nvi
r
onm
e
nt
s
.
G
r
a
ph
le
a
r
ni
ng
f
r
a
m
e
w
or
ks
a
nd
f
e
de
r
a
te
d
m
ode
ls
a
r
e
ga
in
in
g
tr
a
c
ti
on
f
or
di
s
tr
ib
ut
e
d
a
nd
pr
iv
a
c
y
-
pr
e
s
e
r
vi
ng
de
te
c
ti
on.
Y
a
ng
e
t
al
.
[
21]
ut
il
iz
e
d
gr
a
ph
-
ba
s
e
d
a
nom
a
ly
d
e
te
c
ti
on
to
e
nha
n
c
e
li
nk
a
nom
a
ly
r
e
c
ogni
ti
on
(
G
A
D
-
E
L
A
R
)
.
B
ouz
in
is
e
t
al
.
[
22]
in
tr
od
uc
e
d
S
ta
tAvg,
a
f
e
de
r
a
te
d
le
a
r
ni
ng
m
e
th
od
to
a
ddr
e
s
s
d
a
ta
h
e
te
r
oge
ne
it
y
in
I
D
S
.
S
im
il
a
r
ly
,
F
a
r
e
s
e
t
al
.
[
23]
in
tr
oduc
e
d
S
T
-
L
S
T
M
-
D
T
L
by
in
te
gr
a
ti
ng
s
w
in
tr
a
ns
f
or
m
e
r
s
(
S
T
)
a
nd
lo
ng
s
hor
t
-
te
r
m
m
e
m
or
y
(
L
S
T
M
)
in
a
hybr
id
DL
-
ba
s
e
d
tr
a
ns
f
e
r
le
a
r
ni
ng
(
D
T
L
)
m
ode
l
f
or
s
c
a
la
bl
e
I
D
S
in
I
oT
.
T
r
a
ns
f
e
r
a
nd
m
ul
ti
-
vi
e
w
le
a
r
ni
ng
a
ppr
oa
c
he
s
h
a
ve
a
l
s
o
be
e
n
e
xpl
or
e
d.
L
i
e
t
al
.
[
24]
pe
r
f
or
m
e
d
c
om
pa
r
a
ti
ve
s
tu
di
e
s
on
s
in
gl
e
-
a
nd
m
ul
ti
-
vi
e
w
le
a
r
ni
ng
(
S
M
V
L
)
m
ode
ls
,
id
e
nt
if
y
in
g
be
ne
f
it
s
of
di
ve
r
s
if
ie
d
f
e
a
tu
r
e
r
e
pr
e
s
e
nt
a
ti
ons
us
in
g
a
ut
o
-
e
nc
ode
r
(
A
E
)
a
nd
c
onvolut
io
na
l
ne
ur
a
l
ne
twor
k
s
(
C
N
N
s
)
.
L
i
e
t
al
.
[
25]
a
ppl
ie
d
s
e
m
i
-
s
upe
r
vi
s
e
d
le
a
r
ni
ng
(
S
L
)
c
om
bi
ne
d
w
it
h
r
a
ndom
f
or
e
s
ts
(
R
F
)
f
or
in
tr
us
io
n
de
te
c
ti
on, e
f
f
e
c
ti
ve
ly
ha
ndl
in
g pa
r
ti
a
ll
y l
a
be
le
d da
ta
s
e
ts
.
T
r
a
n
s
f
or
m
e
r
-
b
a
s
e
d m
od
e
l
s
s
u
c
h a
s
a
tt
a
c
k
-
a
w
a
r
e
di
vi
de
-
a
nd
-
c
on
que
r
t
r
a
n
s
f
or
m
e
r
(
T
R
A
C
E
R
)
[
26]
a
nd
r
e
c
on
s
tr
uc
ti
on
m
e
m
or
y
ne
tw
or
k
(
R
e
M
e
N
e
t
)
[
27]
ha
v
e
s
how
n
pr
om
is
e
in
i
ndu
s
tr
ia
l
in
te
r
n
e
t
of
th
in
g
s
(
I
I
o
T
)
a
nd
tr
a
ns
por
ta
ti
on
C
P
S
r
e
s
p
e
c
ti
v
e
ly
.
T
he
s
e
m
ode
ls
e
x
c
e
l
in
c
a
pt
ur
i
ng
lo
ng
-
r
a
ng
e
d
e
pe
nde
nc
ie
s
,
but
of
te
n
r
e
qui
r
e
hi
gh
c
om
p
ut
a
ti
ona
l
r
e
s
our
c
e
s
.
B
ia
n
a
nd
L
iu
[
28]
pr
opo
s
e
d
a
r
e
pr
e
s
e
nt
a
ti
o
n
l
e
a
r
ni
n
g
m
od
e
l
G
a
u
s
s
ia
n
-
m
ix
tu
r
e
C
r
a
m
é
r
-
w
ol
d
a
ut
o
-
e
n
c
od
e
r
(
G
M
C
W
A
E
)
,
a
c
hi
e
vi
ng
im
pr
ove
d
a
c
c
ur
a
c
y
a
c
r
os
s
m
ul
ti
pl
e
da
t
a
s
e
ts
.
C
ha
n
dna
n
i
e
t
al
.
[
29]
i
nt
r
odu
c
e
d
f
e
de
r
a
t
e
d
m
ul
t
i
-
la
y
e
r
e
d
d
e
e
p
-
l
e
a
r
ni
n
g
(
F
e
d
-
M
L
D
L
)
a
nd
a
c
hi
e
ve
98%
a
c
c
ur
a
c
y
a
c
r
o
s
s
C
I
C
I
oT
, T
oN
-
I
o
T
,
a
nd
E
dge
-
I
I
o
T
s
e
t
da
t
a
s
e
t
s
. E
la
z
iz
e
t
al
.
[
30]
pr
o
pos
e
d a
f
e
de
r
a
te
d
in
tr
u
s
io
n
de
te
c
ti
on
f
r
a
m
e
w
or
k
u
s
in
g
ta
b
tr
a
ns
f
or
m
e
r
s
a
nd
m
e
t
a
he
ur
is
ti
c
tu
ni
ng,
d
e
m
on
s
tr
a
ti
n
g
e
f
f
e
c
ti
ve
n
e
s
s
a
c
r
os
s
be
nc
h
m
a
r
k
s
li
ke
N
-
B
a
I
o
T
a
nd
C
I
C
I
o
T
2023
.
W
hi
l
e
th
e
s
e
m
e
t
hods
s
how
im
pr
ove
d
pe
r
f
or
m
a
nc
e
i
n
s
p
e
c
if
i
c
c
ont
e
xt
s
,
m
a
ny
do
not
a
de
qua
t
e
ly
a
ddr
e
s
s
c
ov
a
r
ia
t
e
s
hi
f
t,
d
a
ta
i
m
ba
la
n
c
e
,
or
r
e
a
l
-
ti
m
e
a
da
pt
a
bi
li
t
y
c
r
it
i
c
a
l
f
or
dyna
m
i
c
e
dge
-
I
o
T
s
e
tt
in
g
s
.
A
r
e
c
ur
r
in
g
li
m
i
ta
ti
o
n
i
s
r
e
li
a
n
c
e
on
s
ta
ti
c
m
od
e
ls
tr
a
in
e
d
on
f
i
xe
d
ba
l
a
nc
e
d
da
ta
s
e
t
s
,
w
hi
c
h a
r
e
un
a
bl
e
t
o
a
da
pt
t
o n
ove
l
a
tt
a
c
k
pa
tt
e
r
ns
e
m
e
r
gi
ng i
n r
e
a
l
-
w
or
ld
i
m
b
a
la
nc
e
d t
r
a
f
f
i
c
[
31]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
15
, N
o.
1
,
F
e
br
ua
r
y
20
26
:
580
-
591
582
D
e
s
pi
te
e
xt
e
n
s
iv
e
r
e
s
e
a
r
c
h
in
M
L
/DL
-
ba
s
e
d
I
D
S
f
r
a
m
e
w
or
ks
,
e
xi
s
ti
ng
s
ys
te
m
s
of
te
n
s
uf
f
e
r
f
r
om
c
r
it
ic
a
l
s
hor
tc
om
in
gs
w
he
n
de
pl
oye
d
in
dyna
m
ic
I
oT
-
e
dge
e
nvi
r
onm
e
nt
s
[
31]
.
T
he
s
e
in
c
lu
de
:
i)
l
a
c
k
of
a
da
pt
a
bi
li
ty
:
m
os
t
m
ode
ls
a
r
e
tr
a
in
e
d
on
s
ta
ti
c
da
ta
s
e
ts
a
nd
c
a
nnot
r
e
s
pond
to
c
ha
nge
s
in
a
tt
a
c
k
be
ha
vi
or
s
(
i.
e
.,
c
ova
r
ia
te
s
hi
f
t)
;
ii
)
d
a
ta
im
ba
la
nc
e
:
a
tt
a
c
k
d
a
ta
is
of
te
n
un
de
r
r
e
pr
e
s
e
nt
e
d,
le
a
di
ng
to
hi
gh
f
a
ls
e
ne
ga
ti
ve
r
a
te
s
;
ii
i)
i
ns
uf
f
ic
ie
nt
r
e
a
l
-
ti
m
e
pe
r
f
o
r
m
a
nc
e
:
m
a
ny
DL
-
ba
s
e
d
m
ode
ls
a
r
e
r
e
s
our
c
e
-
he
a
vy,
li
m
it
in
g
de
pl
oym
e
nt
on e
dge
de
vi
c
e
s
;
a
nd
iv
)
l
im
it
e
d
ge
ne
r
a
li
z
a
ti
on:
s
ol
ut
io
ns
tu
ne
d t
o s
pe
c
if
ic
da
ta
s
e
ts
of
te
n pe
r
f
or
m
poor
ly
in
he
te
r
oge
ne
ous
e
nvi
r
onm
e
nt
s
.
T
o
a
ddr
e
s
s
th
e
s
e
ga
ps
,
th
is
pa
pe
r
in
tr
oduc
e
s
th
e
dyna
m
ic
a
tt
a
c
k
pa
tt
e
r
n
-
a
w
a
r
e
(
D
A
P
A
)
im
pr
ovi
s
e
d
w
e
ig
ht
e
d
gr
a
di
e
nt
boos
t
in
g
(
I
W
G
B
)
f
r
a
m
e
w
or
k.
I
t
le
ve
r
a
ge
s
r
e
a
l
-
ti
m
e
s
ta
ti
s
ti
c
a
l
m
oni
to
r
in
g
a
nd
a
da
pt
iv
e
l
e
a
r
ni
ng
to
de
te
c
t
c
ov
a
r
ia
te
s
hi
f
ts
,
e
n
s
ur
in
g
r
obus
t
a
nd
lo
w
-
la
te
nc
y
de
te
c
ti
on. T
hi
s
w
or
k a
im
s
t
o f
il
l
th
e
ne
e
d f
or
s
c
a
la
bl
e
, a
da
pt
iv
e
,
a
nd l
ig
ht
w
e
ig
ht
I
D
S
f
or
di
s
tr
ib
ut
e
d C
P
S
.
C
onve
nt
io
na
l
M
L
a
nd
D
L
-
ba
s
e
d
I
D
S
a
r
e
in
s
uf
f
ic
ie
nt
f
or
I
oT
-
e
dge
e
nvi
r
onm
e
nt
s
due
to
th
e
ir
in
a
bi
li
ty
to
a
da
pt
to
e
vol
vi
ng
ne
twor
k
be
ha
vi
or
s
,
ha
ndl
e
da
ta
im
ba
la
nc
e
,
a
nd
op
e
r
a
te
e
f
f
ic
ie
nt
ly
on
c
ons
tr
a
in
e
d
de
vi
c
e
s
.
T
he
r
e
is
a
ne
e
d
f
or
a
dyna
m
ic
,
li
ght
w
e
ig
ht
,
a
nd
r
obus
t
in
tr
us
io
n
de
te
c
ti
on
f
r
a
m
e
w
or
k
th
a
t
c
a
n
e
f
f
e
c
ti
ve
ly
de
te
c
t
a
nd
r
e
s
pond
to
ne
twor
k
-
le
ve
l
th
r
e
a
ts
in
r
e
a
l
ti
m
e
a
c
r
os
s
h
e
te
r
oge
ne
ous
C
P
S
pl
a
tf
or
m
s
.
T
hi
s
pa
p
e
r
pr
opos
e
s
th
e
dyna
m
ic
a
tt
a
c
k
pa
tt
e
r
n
-
a
w
a
r
e
im
pr
ovi
s
e
d
w
e
ig
ht
e
d
gr
a
di
e
nt
boos
ti
ng
(
D
A
P
A
-
I
W
G
B
)
m
ode
l,
a
nove
l
hybr
id
in
tr
us
io
n
de
te
c
ti
on
a
ppr
oa
c
h
ta
il
or
e
d
f
or
I
oT
-
e
dge
-
e
na
bl
e
d
C
P
S
.
T
he
m
ode
l
in
te
gr
a
te
s
e
xt
r
e
m
e
gr
a
di
e
nt
boo
s
ti
ng
(
X
G
B
oos
t)
w
it
h
a
n
im
pr
ovi
s
e
d
w
e
ig
ht
e
d
lo
s
s
f
unc
ti
on
th
a
t
dyna
m
ic
a
ll
y
a
dj
us
ts
to
c
ova
r
ia
te
s
hi
f
ts
in
n
e
twor
k
tr
a
f
f
ic
.
A
s
ta
ti
s
ti
c
a
l
m
oni
to
r
in
g
m
e
c
ha
ni
s
m
is
e
m
b
e
dde
d
to
de
te
c
t
r
e
a
l
-
ti
m
e
di
s
tr
ib
ut
io
na
l
c
ha
nge
s
,
a
ll
ow
in
g
th
e
m
ode
l
to
a
da
pt
c
ont
in
uous
ly
.
T
he
a
r
c
hi
te
c
tu
r
e
is
de
s
ig
ne
d
f
or
e
dge
d
e
pl
oym
e
nt
,
m
in
im
iz
in
g
c
om
put
a
ti
ona
l
lo
a
d
w
hi
le
m
a
in
ta
in
in
g
hi
gh
de
te
c
ti
on
a
c
c
ur
a
c
y.
T
he
m
ode
l
is
tr
a
in
e
d
a
nd
e
va
lu
a
t
e
d
on
two
be
nc
hm
a
r
k
da
t
a
s
e
t
s
U
N
S
W
-
N
B
15
[
32]
a
nd
T
oN
-
I
oT
[
33]
us
in
g
s
tr
a
ti
f
ie
d
s
a
m
pl
in
g
a
nd
c
la
s
s
-
ba
la
n
c
in
g
s
tr
a
te
gi
e
s
to
ove
r
c
o
m
e
da
ta
im
ba
la
nc
e
.
T
h
e
s
ys
t
e
m
de
m
ons
tr
a
te
s
e
nha
nc
e
d a
da
pt
a
bi
li
ty
, l
ow
e
r
f
a
ls
e
pos
it
iv
e
r
a
te
s
, a
nd f
a
s
te
r
de
t
e
c
ti
on t
im
e
s
c
om
pa
r
e
d t
o e
xi
s
ti
ng I
D
S
m
ode
ls
.
P
r
opos
e
s
a
n
a
d
a
pt
iv
e
I
D
S
th
a
t
de
te
c
t
s
nove
l
a
tt
a
c
ks
u
s
in
g
dy
na
m
ic
w
e
ig
ht
e
d
le
a
r
ni
ng.
E
f
f
e
c
ti
ve
ly
ha
ndl
e
s
c
ova
r
ia
te
s
hi
f
t
us
in
g
r
e
a
l
-
ti
m
e
s
ta
ti
s
ti
c
a
l
m
oni
to
r
in
g.
E
nha
nc
e
s
de
te
c
ti
on
a
c
c
ur
a
c
y
on
be
n
c
hm
a
r
k
da
ta
s
e
ts
w
it
h
m
in
im
a
l
f
a
ls
e
pos
it
iv
e
s
.
O
pt
im
iz
e
d
f
or
de
pl
oym
e
nt
on
e
d
ge
de
vi
c
e
s
w
it
h
lo
w
c
om
put
a
ti
ona
l
ov
e
r
he
a
d.
S
tr
e
ngt
he
ns
I
oT
-
e
dge
C
P
S
r
e
s
il
ie
nc
e
a
g
a
in
s
t
e
vol
vi
ng c
ybe
r
t
hr
e
a
ts
.
S
e
c
ti
on
1
pr
e
s
e
nt
s
th
e
ba
c
kgr
ound
a
nd
a
de
ta
il
e
d
li
te
r
a
tu
r
e
r
e
v
ie
w
of
r
e
c
e
nt
I
D
S
us
in
g
ML
a
nd
DL
in
I
oT
-
e
dge
e
nvi
r
onm
e
nt
s
.
T
he
n,
hi
ghl
ig
ht
s
th
e
r
e
s
e
a
r
c
h
g
a
p
a
nd
pr
ovi
de
s
th
e
m
ot
iv
a
ti
on
be
hi
nd
de
ve
lo
pi
ng
th
e
D
A
P
A
-
I
W
G
B
m
ode
l.
F
ur
th
e
r
,
de
f
in
e
s
th
e
pr
obl
e
m
s
ta
te
m
e
nt
a
nd
di
s
c
u
s
s
e
s
th
e
li
m
it
a
ti
ons
of
e
xi
s
ti
ng
s
ys
te
m
s
, a
nd r
e
s
e
a
r
c
h m
e
th
odol
ogy. S
e
c
ti
on 2, de
s
c
r
ib
e
s
t
he
pr
opos
e
d D
A
P
A
-
I
W
G
B
m
e
th
odol
ogy, inc
lu
di
n
g
m
ode
l
c
om
pone
nt
s
,
a
r
c
hi
te
c
tu
r
e
,
a
nd
r
e
a
l
-
ti
m
e
a
da
pt
a
ti
on
s
tr
a
te
gy.
S
e
c
ti
on
3
pr
ovi
de
s
e
xpe
r
im
e
nt
a
l
s
e
tu
p
de
ta
il
s
, da
ta
s
e
ts
us
e
d,
e
va
lu
a
ti
on me
tr
ic
s
, a
nd a
c
om
pr
e
he
n
s
iv
e
pe
r
f
or
m
a
nc
e
a
na
ly
s
is
. S
e
c
ti
on 4 c
onc
lu
de
s
t
he
pa
pe
r
w
it
h a
s
um
m
a
r
y of
ke
y f
in
di
ngs
a
nd outl
in
e
s
pos
s
ib
le
f
ut
ur
e
r
e
s
e
a
r
c
h di
r
e
c
ti
ons
.
2.
M
E
T
H
O
D
T
he
D
A
P
A
-
I
W
G
B
m
ode
l,
s
pe
c
if
ic
a
ll
y
d
e
s
ig
ne
d
f
or
in
tr
us
io
n
d
e
te
c
ti
on
in
I
oT
-
e
dg
e
-
e
na
bl
e
d
C
P
S
,
is
pr
e
s
e
nt
e
d
in
th
i
s
pa
r
t
a
lo
ng
w
it
h
a
th
or
ough
m
e
th
odol
ogy
f
or
it
s
de
s
ig
n
a
nd
im
pl
e
m
e
nt
a
ti
on.
T
h
e
pr
opos
e
d
f
r
a
m
e
w
or
k
is
s
tr
uc
tu
r
e
d
to
id
e
nt
if
y
a
nd
r
e
s
pond
to
e
vol
vi
ng
a
n
d
di
ve
r
s
e
c
ybe
r
th
r
e
a
ts
a
c
r
os
s
di
s
tr
ib
ut
e
d
I
oT
e
nvi
r
onm
e
nt
s
. I
ni
ti
a
ll
y, w
e
de
s
c
r
ib
e
t
he
a
r
c
hi
te
c
tu
r
a
l
de
s
ig
n of
t
he
I
oT
-
e
dge
th
r
e
a
t
de
te
c
ti
on s
y
s
te
m
, f
ol
lo
w
e
d
by
a
f
or
m
a
l
pr
e
s
e
nt
a
ti
on
of
th
e
unde
r
ly
in
g
m
a
th
e
m
a
ti
c
a
l
m
ode
l
s
uppor
ti
ng
th
e
D
A
P
A
-
I
W
G
B
hybr
id
c
la
s
s
if
ic
a
ti
on a
ppr
oa
c
h.
F
ig
ur
e
2
il
lu
s
tr
a
te
s
a
r
e
pr
e
s
e
nt
a
ti
ve
I
oT
-
e
dge
a
r
c
hi
te
c
tu
r
e
c
a
pt
u
r
in
g
ty
pi
c
a
l
ne
twor
k
in
te
r
a
c
ti
ons
a
nd
pot
e
nt
ia
l
s
e
c
ur
it
y vulne
r
a
bi
li
ti
e
s
w
it
hi
n a
he
te
r
oge
ne
ous
C
P
S
. I
n t
hi
s
pa
r
a
di
gm
, va
r
io
us
I
oT
de
vi
c
e
s
i
nc
lu
di
ng
s
m
a
r
t
s
e
ns
or
s
,
w
e
a
r
a
bl
e
s
,
in
du
s
tr
ia
l
m
oni
to
r
s
,
a
nd
hom
e
a
ut
om
a
ti
on
s
ys
te
m
s
c
om
m
uni
c
a
t
e
da
ta
to
lo
c
a
li
z
e
d
e
dge
c
om
put
in
g
node
s
,
w
hi
c
h
pe
r
f
or
m
lo
w
-
la
te
nc
y
pr
e
pr
oc
e
s
s
in
g
ta
s
ks
.
D
e
s
pi
te
th
e
ope
r
a
ti
ona
l
be
ne
f
it
s
of
e
dge
c
om
put
in
g,
th
e
s
ta
te
m
e
nt
pa
th
w
a
y
a
m
ong
I
oT
de
vi
c
e
s
a
nd
e
dge
nod
e
s
r
e
m
a
in
s
s
us
c
e
pt
ib
le
to
a
w
id
e
s
pe
c
tr
um
of
s
e
c
ur
it
y t
hr
e
a
ts
. A
s
hi
ghl
ig
ht
e
d i
n r
e
d i
n t
he
a
r
c
hi
te
c
tu
r
e
, c
e
r
ta
in
node
s
m
a
y be
c
om
pr
om
is
e
d a
nd
e
xpl
oi
te
d
to
in
it
ia
te
c
ybe
r
-
a
tt
a
c
ks
,
s
uc
h
a
s
:
s
poof
in
g
a
nd
i
m
pe
r
s
ona
ti
on
w
it
hi
n
dom
a
in
na
m
e
r
e
s
ol
ut
io
n
pr
ot
oc
ol
s
,
de
ni
a
l
of
s
e
r
vi
c
e
(
D
o
S
)
/DD
oS
a
tt
a
c
ks
le
a
di
ng
to
s
e
r
vi
c
e
di
s
r
upt
io
n,
a
nd
li
nk
-
le
ve
l
o
r
r
out
in
g
a
tt
a
c
ks
,
c
or
r
upt
in
g
th
e
da
ta
f
lo
w
a
nd
de
vi
c
e
s
ync
hr
oni
z
a
ti
o
n.
T
he
s
e
vul
ne
r
a
bi
li
ti
e
s
c
a
n
de
gr
a
de
s
ys
te
m
r
e
li
a
bi
li
ty
,
in
tr
oduc
e
de
la
ys
,
a
nd
c
om
pr
om
is
e
s
e
ns
it
iv
e
in
f
or
m
a
ti
on.
T
o
c
ount
e
r
a
c
t
th
e
s
e
e
vol
vi
ng
th
r
e
a
t
ve
c
to
r
s
,
w
e
pr
opos
e
th
e
D
A
P
A
-
I
W
G
B
in
tr
us
io
n
de
te
c
ti
on
m
ode
l
e
m
be
dde
d
w
it
hi
n
th
e
e
dge
la
ye
r
.
B
y
le
ve
r
a
gi
ng
li
ght
w
e
ig
ht
e
ns
e
m
bl
e
le
a
r
ni
ng
w
it
h
dyna
m
ic
r
e
-
w
e
ig
ht
in
g
m
e
c
ha
ni
s
m
s
,
th
e
m
ode
l
p
e
r
f
or
m
s
c
ont
in
uous
th
r
e
a
t
m
oni
to
r
in
g,
a
da
pt
s
to
c
onc
e
pt
dr
if
t
in
ne
twor
k
be
ha
vi
or
,
a
nd
m
a
in
ta
in
s
de
te
c
ti
on
pe
r
f
or
m
a
nc
e
ove
r
ti
m
e
.
T
w
o
w
e
ll
-
known
r
e
a
l
-
w
or
ld
in
tr
us
io
n
de
te
c
ti
on
d
a
ta
s
e
t
s
a
r
e
us
e
d
to
va
li
da
t
e
th
e
m
ode
l:
T
he
U
N
S
W
-
N
B
15
da
ta
s
e
t,
w
hi
c
h
s
pa
ns
ni
ne
th
r
e
a
t
c
la
s
s
e
s
a
nd
in
c
lu
de
s
a
c
om
bi
na
ti
on
of
a
r
ti
f
ic
ia
l
a
nd
r
e
a
l
ne
twor
k
tr
a
f
f
ic
th
a
t
r
e
pr
e
s
e
nt
s
bot
h
c
ont
e
m
por
a
r
y
a
nd
c
onve
nt
io
na
l
a
tt
a
c
k
c
a
te
gor
ie
s
.
T
e
l
e
m
e
tr
y,
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
D
y
nam
ic
at
ta
c
k
pat
te
r
n
-
aw
a
r
e
i
nt
e
ll
ig
e
nt
c
y
b
e
r
-
phy
s
ic
al
…
(
V
is
hal
a I
bas
apur
a L
ak
s
hm
in
a
r
ay
anappa
)
583
ope
r
a
ti
ng
s
ys
te
m
lo
gs
,
a
nd
ne
twor
k
-
ba
s
e
d
in
tr
us
io
n
e
ve
nt
s
ga
th
e
r
e
d
f
r
om
s
m
a
r
t
de
vi
c
e
s
a
nd
e
dge
s
ys
te
m
s
m
a
ke
up
th
e
T
oN
-
I
oT
da
ta
s
e
t,
w
hi
c
h
w
a
s
c
r
e
a
t
e
d
f
or
he
te
r
oge
ne
ous
I
o
T
e
nvi
r
onm
e
nt
s
.
T
og
e
th
e
r
,
th
e
s
e
da
ta
s
e
ts
s
uppor
t
c
om
pr
e
he
ns
iv
e
e
v
a
lu
a
ti
on
of
th
e
D
A
P
A
-
I
W
G
B
m
ode
l
a
c
r
os
s
va
r
ie
d
a
tt
a
c
k
pa
tt
e
r
ns
,
da
t
a
m
oda
li
ti
e
s
,
a
nd
r
e
a
l
-
w
or
ld
C
P
S
de
pl
oym
e
nt
s
c
e
na
r
io
s
.
B
y
d
e
pl
oyi
ng
th
e
pr
opos
e
d
de
te
c
ti
on
m
e
c
ha
ni
s
m
di
r
e
c
tl
y
a
t
th
e
e
dge
,
th
e
s
y
s
te
m
be
ne
f
it
s
f
r
om
r
e
duc
e
d
de
te
c
t
io
n
la
te
nc
y,
lo
w
e
r
e
d
ne
twor
k
ov
e
r
he
a
d,
a
nd
in
c
r
e
a
s
e
d
r
e
s
pon
s
iv
e
ne
s
s
to
lo
c
a
li
z
e
d
th
r
e
a
ts
.
T
hi
s
a
r
c
hi
te
c
t
ur
a
l
de
s
ig
n
pr
ovi
de
s
a
r
e
s
il
ie
nt
a
nd
s
c
a
la
bl
e
f
ounda
ti
on f
or
s
e
c
ur
in
g I
oT
-
e
dge
-
e
na
bl
e
d C
P
S
a
ga
in
s
t
dyna
m
ic
a
nd mul
ti
f
a
c
e
te
d c
ybe
r
t
hr
e
a
ts
.
F
ig
ur
e
2. N
e
twor
k c
om
m
uni
c
a
ti
on a
tt
a
c
k s
c
e
na
r
io
i
n I
oT
-
e
dge
c
om
put
in
g ne
twor
ks
T
hi
s
s
e
c
ti
on
in
tr
oduc
e
s
th
e
ba
s
ic
w
or
ki
ng
of
gr
a
di
e
nt
boo
s
ti
ng
(
X
G
B
oos
t
)
m
ode
l
.
F
ur
th
e
r
,
pr
e
s
e
nt
a
n
I
W
G
B
m
ode
l
f
or
D
A
P
A
to
opt
im
iz
e
th
e
pr
e
di
c
ti
on
e
r
r
or
.
T
he
s
e
c
ti
on
in
tr
oduc
e
s
a
ne
w
w
e
ig
ht
e
d
s
um
pr
e
di
c
ti
on
e
r
r
or
m
in
im
iz
a
ti
on
m
ode
l
c
om
bi
ne
d
w
it
h
lo
ga
r
it
h
m
-
ba
s
e
d
lo
s
t
f
unc
ti
on
to
de
s
ig
n
a
nove
l
a
tt
a
c
k
pr
e
di
c
ti
ve
c
la
s
s
if
ie
r
;
th
e
D
A
P
A
-
I
W
G
B
e
ns
ur
e
s
to
r
e
duc
e
th
e
m
is
c
la
s
s
if
ic
a
ti
on
in
th
e
C
P
S
I
oT
-
e
dge
c
om
put
in
g ne
twor
ks
. L
e
t
th
e
i
nput
da
ta
s
e
t
(
i.
e
., U
N
S
W
a
nd
T
o
N
-
I
oT
)
be
de
not
e
d i
n (
1)
.
=
{
(
1
,
1
)
,
(
2
,
2
)
,
…
…
(
,
)
}
(
1)
H
e
r
e
,
e
a
c
h
r
e
pr
e
s
e
nt
s
a
f
e
a
tu
r
e
ve
c
to
r
a
t
ti
m
e
s
te
p
,
a
nd
∈
{
0
,
1
}
de
not
e
s
th
e
tr
ue
la
be
l,
w
he
r
e
=
0
c
or
r
e
s
ponds
to
no
a
tt
a
c
k
a
nd
=
1
in
di
c
a
te
s
th
e
pr
e
s
e
n
c
e
of
a
n
a
tt
a
c
k.
T
he
obj
e
c
ti
ve
f
unc
ti
on
of
th
e
D
A
P
A
-
I
W
G
B
m
ode
l
is
f
o
r
m
ul
a
te
d
a
s
(
2)
.
I
n
(
2)
,
is
th
e
num
be
r
of
ti
m
e
s
te
ps
,
ℓ
(
⋅
)
is
th
e
lo
s
s
f
unc
ti
o
n
be
twe
e
n
tr
ue
la
be
l
a
nd
pr
e
di
c
te
d
la
be
l
,
a
nd
r
e
pr
e
s
e
nt
s
th
e
ℎ
ba
s
e
le
a
r
ne
r
(
de
c
is
io
n
tr
e
e
)
.
T
he
r
e
gul
a
r
iz
a
ti
on
te
r
m
(
)
,
w
hi
c
h
c
ont
r
ol
s
m
ode
l
c
om
pl
e
xi
ty
,
is
de
f
in
e
d
a
s
(
3)
.
W
he
r
e
a
nd
a
r
e
r
e
gul
a
r
iz
a
ti
on
hype
r
pa
r
a
m
e
te
r
s
,
a
nd
de
f
in
e
s
th
e
ve
c
to
r
of
le
a
f
w
e
ig
ht
s
.
E
a
c
h
ba
s
e
le
a
r
ne
r
is
r
e
pr
e
s
e
nt
e
d
a
s
a
w
e
ig
ht
e
d
s
um
of
l
e
a
f
pr
e
di
c
ti
ons
, a
s
i
n
(
4)
.
(
)
=
∑
ℓ
(
,
̂
)
=
1
+
∑
(
)
=
1
(
2)
(
)
=
+
1
2
|
|
|
|
2
(
3)
(
)
=
∑
(
)
ℎ
(
)
=
1
(
4)
H
e
r
e
,
ℎ
(
)
is
th
e
de
c
i
s
io
n
r
ul
e
f
or
th
e
ℎ
node
a
t
ti
m
e
,
a
nd
(
)
is
th
e
c
or
r
e
s
ponding
w
e
ig
ht
.
T
o
a
da
pt
to
va
r
yi
ng
pr
e
di
c
ti
on
e
r
r
or
s
,
dyna
m
ic
w
e
ig
ht
s
a
r
e
c
o
m
put
e
d
a
s
in
(
5)
.
̅
(
)
r
e
pr
e
s
e
nt
s
th
e
a
ve
r
a
ge
e
r
r
or
f
or
m
ode
l
iii
ove
r
a
r
e
c
e
nt
ti
m
e
w
in
dow
,
a
nd
(
)
is
th
e
in
s
t
a
nt
a
ne
ous
e
r
r
or
.
T
o
m
it
ig
a
te
vol
a
ti
li
ty
in
w
e
ig
ht
upda
te
s
a
nd
e
ns
ur
e
te
m
por
a
l
s
ta
bi
li
ty
,
a
n
a
ve
r
a
ge
d
upd
a
te
m
e
c
ha
ni
s
m
is
in
tr
oduc
e
d
a
s
in
(
6)
.
W
he
r
e
de
not
e
s
th
e
num
be
r
of
pr
io
r
s
te
ps
c
ons
id
e
r
e
d
f
or
s
m
oot
hi
ng
a
nd
(
−
)
de
f
in
e
s
pr
e
c
e
di
ng
ti
m
e
s
te
ps
w
e
ig
ht
of
. T
he
m
ode
l'
s
pr
e
di
c
ti
on e
r
r
or
a
t
ti
m
e
is
de
f
in
e
d
in
(
7)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
15
, N
o.
1
,
F
e
br
ua
r
y
20
26
:
580
-
591
584
(
)
=
̅
(
)
(
)
(
5)
(
)
=
1
∙
∑
(
−
)
=
1
(
6)
,
=
∑
(
)
∙
(
)
−
(
)
=
1
(
7)
(
)
de
f
in
in
g
w
e
ig
ht
e
d
va
lu
e
a
s
s
ig
n
e
d
a
t
ti
m
e
in
s
ta
nc
e
,
(
)
de
f
in
e
s
out
c
om
e
of
ba
s
e
li
ne
c
la
s
s
if
ie
r
f
or
,
a
nd
(
)
de
f
in
e
s
th
e
r
e
a
l
va
lu
e
a
t
in
s
t
a
nc
e
.
T
he
r
e
f
or
e
,
th
e
e
r
r
or
is
dyna
m
ic
a
ll
y
r
e
c
a
lc
ul
a
te
d
f
or
th
e
ne
xt
ti
m
e
s
te
p
a
s
in
(
8)
.
T
he
f
in
a
l
pr
e
di
c
ti
on
of
th
e
e
ns
e
m
bl
e
is
a
w
e
ig
ht
e
d
s
um
of
tr
e
e
pr
e
di
c
ti
ons
a
s
s
how
n
in
(
9)
.
T
he
lo
s
s
f
unc
ti
on
is
in
it
ia
ll
y
f
or
m
ul
a
te
d
a
s
a
bi
na
r
y
c
r
os
s
-
e
nt
r
opy
w
it
h
c
l
a
s
s
im
ba
la
nc
e
c
on
s
id
e
r
a
ti
on
a
s
s
how
n
in
(
10)
.
H
ow
e
v
e
r
,
th
e
lo
s
s
f
unc
ti
on
w
it
h
c
la
s
s
im
ba
la
nc
e
pa
r
a
m
e
te
r
ig
nor
in
g bi
a
s
e
d w
e
ig
ht
in
g i
s
s
ue
s
c
ons
id
e
r
in
g i
m
ba
la
nc
e
d da
ta
.
+
1
,
=
∑
+
1
(
)
∙
+
1
(
+
1
)
−
(
)
=
1
(
8)
̂
=
∑
(
)
(
)
=
1
(
9)
=
−
∑
(
l
o
g
(
=
1
̂
)
+
(
1
−
)
l
o
g
(
1
−
̂
)
)
(
10)
T
hus
,
to
f
ur
th
e
r
a
ddr
e
s
s
d
a
ta
im
ba
la
nc
e
,
th
e
c
la
s
s
-
s
pe
c
if
ic
w
e
ig
ht
s
0
a
nd
1
a
r
e
in
c
or
por
a
te
d
a
s
in
(
11)
.
T
he
0
w
e
ig
ht
is
ke
pt
lo
w
e
r
th
a
n
th
e
1
to
r
e
duc
e
th
e
f
a
ls
e
pos
it
iv
e
s
.
T
he
r
e
f
or
e
,
th
e
w
e
ig
ht
s
a
r
e
de
te
r
m
in
e
d
ba
s
e
d
on
in
ve
r
s
e
c
la
s
s
f
r
e
que
nc
y
a
s
in
(
12)
.
W
he
r
e
de
f
in
e
s
th
e
r
a
ti
on
b
e
twe
e
n
nor
m
a
l
a
nd
m
a
li
c
io
us
pa
c
ke
t.
T
he
upda
te
d m
ode
l
w
e
ig
ht
e
xpr
e
s
s
io
n i
nt
e
gr
a
ti
ng c
la
s
s
i
m
ba
la
nc
e
be
c
om
e
s
(
13)
.
=
−
∑
(
1
l
o
g
(
=
1
̂
)
+
0
(
1
−
)
l
o
g
(
1
−
̂
)
)
(
11)
=
1
,
∈
{
0
,
1
}
(
12)
(
)
=
∑
(
̅
(
)
∙
)
=
1
(
)
(
13)
F
ur
th
e
r
,
in
di
vi
dua
l
tr
e
e
w
e
ig
ht
s
a
r
e
pe
n
a
li
z
e
d
if
m
is
c
la
s
s
if
yi
ng
m
in
or
it
y
c
la
s
s
in
s
ta
nc
e
s
a
s
in
(
14)
.
W
he
r
e
de
f
in
in
g
th
e
pe
na
li
z
in
g
te
r
m
us
e
d
f
or
r
e
duc
in
g
th
e
w
e
i
ght
of
f
a
ls
e
pos
it
iv
e
c
la
s
s
if
ie
r
f
or
m
m
in
or
it
y
c
la
s
s
e
s
.
T
he
r
e
f
or
e
,
a
ll
tr
e
e
w
e
ig
ht
s
a
r
e
nor
m
a
li
z
e
d
a
f
te
r
e
a
c
h
pe
r
io
d
a
s
in
(
15)
.
U
nde
r
pe
r
f
or
m
in
g
tr
e
e
s
w
it
h
w
e
ig
ht
s
be
lo
w
a
pr
e
de
f
in
e
d
th
r
e
s
hol
d
a
r
e
pr
une
d
a
s
in
(
16)
.
W
he
r
e
is
us
e
d
a
s
a
th
r
e
s
hol
d
pa
r
a
m
e
te
r
f
o
r
opt
im
iz
in
g t
he
t
r
e
e
s
iz
e
t
o a
tt
a
in
de
s
ir
e
d a
tt
a
c
k de
te
c
ti
on a
c
c
ur
a
c
y w
it
hout
ove
r
f
it
ti
ng pr
obl
e
m
s
.
=
, i
f
(
)
≠
a
nd
=
1
(
14)
=
∑
(
15)
ℋ
=
ℋ
{
|
<
}
(
16)
T
hi
s
s
e
c
ti
on
in
tr
oduc
e
s
D
A
P
A
c
om
bi
ne
d
w
it
h
I
W
G
B
f
or
c
ova
r
ia
te
s
hi
f
t
ha
ndl
in
g
a
nd
m
ode
l
a
da
pt
a
ti
on.
T
h
e
m
ode
l
a
ddr
e
s
s
e
s
c
onc
e
pt
dr
if
t
a
nd
e
vol
vi
ng
a
tt
a
c
k
pa
tt
e
r
ns
us
in
g
a
n a
da
pt
iv
e
e
n
s
e
m
bl
e
-
ba
s
e
d
tr
a
in
in
g s
tr
a
te
gy, s
um
m
a
r
iz
e
d i
n A
lg
or
it
hm
1.
A
lg
or
it
hm
1. D
A
P
A
-
I
W
G
B
e
ns
e
m
bl
e
a
da
pt
a
ti
on mode
l
f
or
i
nt
r
us
io
n de
te
c
ti
on i
n I
oT
-
e
dge
ne
twor
ks
I
nput
D
a
ta
s
e
t
, D
A
P
A
-
I
W
G
B
e
ns
e
m
bl
e
m
ode
l
in
de
x
, numbe
r
of
m
ode
ls
, a
tt
a
c
k c
la
s
s
, t
r
e
e
pr
uni
ng t
hr
e
s
hol
d
, dyna
m
ic
pe
na
lt
y f
a
c
to
r
, upda
te
i
nt
e
r
va
l
, a
nd N
um
be
r
of
f
ol
ds
O
ut
put
U
pda
te
d e
ns
e
m
bl
e
in
D
A
P
A
-
I
W
G
B
a
nd node
w
e
ig
ht
s
1.
I
ni
ti
a
li
z
e
:
S
e
t
num
be
r
of
t
r
e
e
s
=
1
, i
ni
ti
a
li
z
e
a
ll
w
e
ig
ht
s
=
1
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
D
y
nam
ic
at
ta
c
k
pat
te
r
n
-
aw
a
r
e
i
nt
e
ll
ig
e
nt
c
y
b
e
r
-
phy
s
ic
al
…
(
V
is
hal
a I
bas
apur
a L
ak
s
hm
in
a
r
ay
anappa
)
585
P
a
r
ti
ti
on da
ta
s
e
t
in
to
f
ol
ds
2.
F
or
e
a
c
h f
ol
d
=
1
to
do
U
s
e
f
ol
d
a
s
va
li
da
ti
on
s
e
t
a
nd r
e
m
a
in
in
g f
ol
ds
a
s
t
r
a
in
in
g
da
ta
T
r
a
in
D
A
P
A
-
I
W
G
B
m
ode
l
on t
r
a
in
in
g da
ta
I
ni
ti
a
li
z
e
e
ns
e
m
bl
e
=
{
}
3.
F
or
e
a
c
h t
im
e
s
te
p
=
1
to
do
F
or
e
a
c
h m
ode
l
=
1
to
do
F
or
e
a
c
h t
r
e
e
=
1
to
do
If
(
)
≠
a
nd
(
)
≠
0
th
e
n
U
pda
te
t
r
e
e
w
e
ig
ht
:
=
⋅
E
nd I
f
E
nd F
or
C
om
put
e
e
ns
e
m
bl
e
pr
e
di
c
ti
on u
s
in
g:
(
)
=
∑
(
)
ℎ
(
)
=
1
If
(
)
=
0
th
e
n
N
or
m
a
li
z
e
w
e
ig
ht
s
:
=
∑
P
r
une
l
ow
-
w
e
ig
ht
t
r
e
e
s
:
ℋ
=
ℋ
{
|
<
}
I
f
pr
e
di
c
te
d
̂
≠
th
e
n
I
nc
r
e
a
s
e
numbe
r
of
t
r
e
e
s
:
=
+
1
G
e
ne
r
a
te
ne
w
t
r
e
e
vi
a
f
e
a
tu
r
e
-
s
pl
it
:
ℎ
(
)
A
dd ne
w
t
r
e
e
:
ℋ
=
∪
ℎ
(
)
, s
e
t
=
1
E
nd I
f
E
nd I
f
F
or
e
a
c
h t
r
e
e
=
1
to
do
U
pda
te
t
r
e
e
w
it
h ne
w
da
ta
:
=
(
,
,
)
E
nd F
or
E
nd F
or
E
nd F
or
4.
R
e
tu
r
n:
F
in
a
l
e
ns
e
m
bl
e
a
nd w
e
ig
ht
s
A
lg
or
it
hm
1
d
e
s
c
r
ib
e
s
th
e
tr
a
i
ni
n
g
a
nd
a
da
pt
a
ti
on
pr
o
c
e
s
s
of
t
he
D
A
P
A
-
I
W
G
B
m
od
e
l
f
or
i
nt
r
u
s
io
n
de
t
e
c
ti
on
i
n
I
oT
-
e
dg
e
e
n
a
bl
e
d
C
P
S
.
T
h
i
s
m
o
de
l
d
yn
a
m
i
c
a
ll
y
h
a
nd
le
s
c
o
va
r
ia
te
s
hi
f
t
,
c
o
nc
e
pt
dr
i
f
t,
a
n
d
c
l
a
s
s
im
b
a
l
a
n
c
e
w
hi
le
im
pr
ovi
ng
c
l
a
s
s
if
i
c
a
ti
o
n
a
c
c
ur
a
c
y
a
c
r
o
s
s
e
vo
lv
in
g
a
tt
a
c
k
p
a
tt
e
r
ns
.
T
h
e
m
a
in
c
o
m
po
ne
nt
s
a
nd
w
or
kf
lo
w
a
r
e
e
xp
la
in
e
d
a
s
:
i)
I
ni
ti
a
l
iz
a
ti
on
a
n
d
K
-
f
ol
d
s
e
t
up
:
th
e
d
a
t
a
s
e
t
is
pa
r
ti
ti
on
e
d
in
t
o
non
-
o
ve
r
la
ppi
ng
f
ol
d
s
to
e
n
a
b
le
c
r
o
s
s
-
va
l
id
a
ti
on.
O
n
e
-
f
ol
d
i
s
u
s
e
d
f
or
v
a
l
id
a
ti
on
w
hi
l
e
th
e
r
e
m
a
i
ni
n
g
−
1
f
ol
d
s
a
r
e
u
s
e
d
f
or
tr
a
in
in
g
.
T
hi
s
a
pp
r
o
a
c
h
i
m
pr
o
ve
s
th
e
ge
ne
r
a
l
iz
a
b
il
i
ty
of
t
he
m
o
de
l
by
r
e
d
uc
in
g
ov
e
r
f
i
tt
in
g a
n
d e
n
s
ur
e
s
r
obu
s
tn
e
s
s
a
c
r
o
s
s
va
r
i
e
d
d
a
ta
s
e
g
m
e
nt
s
.
E
a
c
h
d
e
c
i
s
i
on
tr
e
e
in
th
e
e
n
s
e
m
bl
e
i
s
i
ni
t
i
a
li
z
e
d
w
it
h
e
qu
a
l
w
e
i
ght
=
1
,
a
nd
th
e
num
b
e
r
of
tr
e
e
s
m
m
m
b
e
gi
n
s
f
r
om
on
e
.
ii)
E
ns
e
m
bl
e
tr
a
in
in
g
o
ve
r
ti
m
e
s
t
e
p
s
:
f
or
e
ve
r
y
t
im
e
s
t
e
p
t
tt
,
th
e
a
l
gor
it
hm
it
e
r
a
te
s
ov
e
r
e
a
c
h
b
a
s
e
le
a
r
ne
r
ii
i
in
th
e
e
n
s
e
m
b
le
of
s
i
z
e
.
T
he
m
o
d
e
l
u
s
e
s
th
e
D
A
P
A
-
I
W
G
B
f
r
a
m
e
w
or
k
t
o
bui
ld
d
e
c
is
io
n
t
r
e
e
s
w
hi
c
h
a
r
e
c
a
p
a
bl
e
of
c
a
p
tu
r
i
ng
di
s
t
in
c
t
a
t
ta
c
k
p
a
tt
e
r
ns
.
iii)
E
r
r
or
-
ba
s
e
d
w
e
ig
ht
a
dj
u
s
tm
e
nt
:
i
f
a
de
c
is
io
n
tr
e
e
H
jH_j
H
j
in
c
or
r
e
c
tl
y
c
la
s
s
if
ie
s
th
e
in
put
s
a
m
pl
e
(
(
)
≠
)
a
nd
th
e
it
e
r
a
ti
on
in
de
x
is
not
a
li
gne
d
w
it
h
th
e
tr
e
e
upda
te
in
te
r
va
l
,
th
e
w
e
ig
ht
is
pe
na
li
z
e
d
u
s
in
g
a
d
e
c
a
y
f
a
c
to
r
.
T
hi
s
m
e
c
h
a
ni
s
m
e
n
s
ur
e
s
th
a
t
le
s
s
a
c
c
ur
a
te
tr
e
e
s
gr
a
dua
ll
y
lo
s
e
in
f
lu
e
nc
e
ove
r
t
im
e
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
15
, N
o.
1
,
F
e
br
ua
r
y
20
26
:
580
-
591
586
iv
)
P
r
e
d
ic
ti
o
n
a
n
d
e
n
s
e
m
bl
e
a
g
gr
e
ga
ti
o
n:
th
e
ov
e
r
a
ll
pr
e
di
c
ti
o
n
a
t
ti
m
e
s
t
e
p
i
s
obt
a
i
ne
d
u
s
in
g
a
w
e
i
ght
e
d
s
u
m
of
ou
tp
u
t
s
f
r
o
m
a
ll
d
e
c
is
io
n
tr
e
e
s
b
a
s
e
d
o
n
(
4)
.
T
h
e
m
od
e
l
dyn
a
m
i
c
a
ll
y
a
d
a
pt
s
it
s
pr
e
di
c
ti
on
s
t
r
a
t
e
g
y
to
c
ur
r
e
n
t
i
np
ut
c
o
ndi
ti
o
n
s
.
v)
P
e
r
io
di
c
tr
e
e
w
e
ig
ht
nor
m
a
li
z
a
ti
on
a
nd
pr
uni
ng:
a
t
s
pe
c
if
ie
d
in
te
r
va
ls
(
e
ve
r
y
it
e
r
a
ti
ons
)
,
th
e
f
ol
lo
w
in
g
a
c
ti
ons
a
r
e
tr
ig
ge
r
e
d:
w
e
ig
ht
nor
m
a
li
z
a
ti
on:
tr
e
e
w
e
ig
ht
s
a
r
e
s
c
a
le
d
r
e
la
ti
ve
to
th
e
to
ta
l
s
um
to
m
a
in
ta
i
n
a
c
ons
i
s
te
nt
in
f
lu
e
nc
e
r
a
ng
e
.
T
r
e
e
pr
uni
ng:
tr
e
e
s
w
hos
e
w
e
ig
ht
s
f
a
ll
be
lo
w
a
pr
e
de
f
in
e
d
th
r
e
s
hol
d
a
r
e
c
ons
id
e
r
e
d unde
r
pe
r
f
or
m
in
g a
nd a
r
e
r
e
m
ove
d f
r
om
t
he
e
ns
e
m
bl
e
.
vi
)
E
ns
e
m
bl
e
e
xpa
n
s
io
n:
if
th
e
e
ns
e
m
bl
e
m
is
c
l
a
s
s
if
ie
s
th
e
c
ur
r
e
n
t
in
put
(
̂
≠
)
,
a
ne
w
de
c
i
s
io
n
tr
e
e
is
a
dde
d
by
pe
r
f
or
m
in
g
a
f
r
e
s
h
f
e
a
tu
r
e
-
ba
s
e
d
s
pl
it
.
T
hi
s
e
ns
ur
e
s
th
a
t
th
e
m
ode
l
le
a
r
ns
f
r
om
r
e
c
e
nt
e
r
r
o
r
s
a
nd a
da
pt
s
t
o e
vol
vi
ng a
tt
a
c
k b
e
ha
vi
or
.
vi
i)
C
ova
r
ia
te
s
hi
f
t
a
da
pt
a
ti
on
:
t
o
a
ddr
e
s
s
th
e
pr
obl
e
m
o
f
c
ova
r
ia
t
e
s
hi
f
t,
e
a
c
h
tr
e
e
is
upda
te
d
a
t
e
ve
r
y
ti
m
e
s
te
p
w
it
h
th
e
la
te
s
t
in
s
ta
nc
e
(
,
)
.
T
hi
s
a
ll
ow
s
th
e
de
c
is
io
n
tr
e
e
s
to
e
vol
ve
c
ont
in
uous
ly
b
y
le
a
r
ni
ng
f
r
om
ne
w
a
nd
pot
e
nt
ia
ll
y
s
hi
f
te
d
da
ta
di
s
tr
ib
ut
io
ns
,
e
ns
ur
in
g
th
a
t
th
e
m
ode
l
s
ta
ys
r
e
le
va
nt
in
dyna
m
ic
C
P
S
e
nvi
r
onm
e
nt
s
.
vi
ii
)
F
in
a
l
out
put
:
th
e
a
lg
or
it
hm
ou
tp
ut
s
th
e
upda
te
d
e
ns
e
m
bl
e
of
de
c
is
io
n
tr
e
e
s
a
nd
th
e
ir
r
e
s
pe
c
ti
ve
w
e
ig
ht
s
,
w
hi
c
h
a
r
e
opt
im
iz
e
d
th
r
ough
dyna
m
ic
a
dj
us
tm
e
nt
,
pr
uni
ng,
a
nd
c
r
os
s
-
va
li
da
ti
on.
T
hi
s
opt
im
iz
e
d e
ns
e
m
bl
e
i
s
t
he
n u
s
e
d t
o pr
e
di
c
t
a
tt
a
c
k
s
on f
ut
ur
e
uns
e
e
n i
nput
s
.
K
e
y
a
dva
nt
a
g
e
s
of
th
e
a
lg
or
it
hm
a
r
e
:
i)
dyna
m
ic
w
e
ig
ht
in
g:
poo
r
ly
pe
r
f
or
m
in
g
tr
e
e
s
lo
s
e
in
f
lu
e
nc
e
ove
r
ti
m
e
,
a
nd
tr
e
e
s
th
a
t
c
a
pt
ur
e
ne
w
a
tt
a
c
k
pa
tt
e
r
ns
a
r
e
a
dde
d
a
da
pt
iv
e
ly
;
ii
)
c
ova
r
ia
te
s
hi
f
t
ha
ndl
in
g:
e
a
c
h
tr
e
e
is
upda
te
d
w
it
h
ne
w
s
a
m
pl
e
s
,
m
a
in
ta
in
in
g
pr
e
di
c
ti
on
r
e
le
va
nc
e
ove
r
ti
m
e
;
ii
i)
i
m
ba
la
nc
e
-
a
w
a
r
e
le
a
r
ni
ng:
c
la
s
s
-
s
p
e
c
if
ic
w
e
ig
ht
in
g
(
in
c
or
por
a
te
d
in
lo
s
s
f
unc
ti
on
a
nd
tr
e
e
w
e
ig
ht
in
g)
e
ns
ur
e
s
th
a
t
m
in
or
it
y
c
la
s
s
a
tt
a
c
ks
a
r
e
not
ove
r
lo
oke
d
;
iv
)
K
-
f
ol
d
opt
im
iz
a
ti
on:
pr
e
ve
nt
s
ov
e
r
f
it
ti
ng
by
va
li
da
ti
ng
p
e
r
f
or
m
a
nc
e
a
c
r
os
s
m
ul
ti
pl
e
da
ta
pa
r
ti
ti
ons
,
e
nha
nc
in
g
ge
ne
r
a
li
z
a
ti
on
;
a
nd
v)
m
ode
l
s
c
a
la
bi
l
it
y:
e
ns
e
m
bl
e
s
iz
e
i
s
a
da
pt
iv
e
ly
c
ont
r
ol
le
d
vi
a
pr
uni
ng
a
nd
a
ddi
ti
on
of
tr
e
e
s
ba
s
e
d
on
p
e
r
f
or
m
a
nc
e
.
T
hi
s
a
ppr
oa
c
h
e
n
s
ur
e
s
c
ont
in
uou
s
le
a
r
ni
ng
f
r
om
r
e
a
l
-
ti
m
e
da
ta
by
dyna
m
ic
a
ll
y
a
dj
us
ti
ng
de
c
is
io
n
tr
e
e
s
,
upda
ti
ng
le
a
f
w
e
ig
ht
s
ba
s
e
d
on
m
is
c
la
s
s
if
ic
a
ti
on
of
r
a
r
e
a
tt
a
c
k
ty
pe
s
,
a
nd
pr
uni
ng
le
s
s
in
f
or
m
a
ti
ve
m
ode
ls
.
T
he
in
c
lu
s
io
n
of
c
la
s
s
-
ba
s
e
d
w
e
ig
ht
in
g,
c
ova
r
ia
te
s
hi
f
t
ha
ndl
in
g,
a
nd
pe
r
io
di
c
nor
m
a
li
z
a
ti
on
e
ns
ur
e
s
r
e
s
il
ie
nc
e
a
ga
in
s
t
c
la
s
s
im
ba
la
nc
e
a
nd
te
m
por
a
l
da
ta
dr
if
t
in
I
oT
-
e
dge
-
ba
s
e
d C
P
S
e
nvi
r
onm
e
nt
s
.
3.
E
X
P
E
R
I
M
E
N
T
A
L
S
E
T
U
P
A
N
D
R
E
S
U
L
T
S
T
he
pr
opos
e
d
D
A
P
A
-
I
W
G
B
f
r
a
m
e
w
or
k
w
a
s
de
ve
lo
pe
d
a
nd
de
pl
oye
d
w
it
hi
n
a
s
im
ul
a
te
d
I
oT
-
e
dge
c
om
put
in
g
e
nvi
r
onm
e
nt
to
e
va
lu
a
te
it
s
e
f
f
e
c
ti
ve
ne
s
s
in
r
e
a
l
-
ti
m
e
a
tt
a
c
k
de
te
c
ti
on.
P
yt
hon
3
w
a
s
us
e
d
f
or
th
e
im
pl
e
m
e
nt
a
ti
on,
a
nd
th
e
de
te
c
ti
on
c
om
pone
nt
w
a
s
in
te
g
r
a
te
d
a
t
th
e
e
dge
la
ye
r
to
e
na
bl
e
lo
c
a
l
de
c
is
io
n
-
m
a
ki
ng
a
nd
lo
w
-
la
te
nc
y
th
r
e
a
t
pr
e
di
c
ti
on.
A
c
om
put
e
r
s
ys
te
m
w
it
h
a
n
I
nt
e
l
C
or
e
i7
p
r
oc
e
s
s
or
a
nd
16
G
B
of
R
A
M
w
a
s
u
s
e
d
f
or
th
e
e
xp
e
r
im
e
nt
s
,
gua
r
a
nt
e
e
in
g
e
nough
pr
oc
e
s
s
in
g
pow
e
r
f
or
bot
h
tr
a
in
in
g
a
nd
in
f
e
r
e
nc
e
w
it
hout
s
a
c
r
if
ic
in
g pe
r
f
or
m
a
nc
e
.
T
w
o
popula
r
c
ybe
r
s
e
c
ur
it
y
be
nc
hm
a
r
k
d
a
ta
s
e
t
s
,
U
N
S
W
-
N
B
15
a
nd
T
oN
-
I
oT
,
w
hi
c
h
pr
ovi
de
th
or
ough
c
ove
r
a
ge
of
c
ont
e
m
por
a
r
y
c
ybe
r
-
a
tt
a
c
k
tr
e
nds
in
I
oT
c
ont
e
xt
s
,
w
e
r
e
us
e
d
in
th
e
e
va
lu
a
ti
on.
U
N
S
W
-
N
B
15
[
32]
:
t
hi
s
da
ta
s
e
t,
w
hi
c
h
w
a
s
c
r
e
a
te
d
by
th
e
A
us
tr
a
li
a
n
c
e
nt
r
e
f
or
c
ybe
r
s
e
c
ur
it
y
(
A
C
C
S
)
,
in
c
lu
de
s
bot
h
b
e
ni
gn
a
nd
m
a
li
c
io
us
ne
twor
k
tr
a
f
f
ic
f
r
om
ni
ne
di
f
f
e
r
e
nt
a
tt
a
c
k
ty
pe
s
,
s
u
c
h
a
s
ba
c
kdoor
s
,
w
or
m
s
,
D
oS
,
a
na
ly
s
is
,
e
xpl
oi
ts
,
a
nd
s
pa
m
m
in
g.
I
t
in
c
lu
de
s
49
ne
twor
k
f
e
a
tu
r
e
s
th
a
t
c
a
pt
ur
e
be
h
a
vi
or
a
l
c
ha
r
a
c
te
r
is
ti
c
s
c
r
it
ic
a
l
f
or
a
nom
a
ly
de
te
c
ti
on.
T
hi
s
da
ta
s
e
t
pr
ovi
de
s
a
r
obus
t
te
s
tb
e
d
f
or
e
va
lu
a
ti
ng
I
D
S
m
ode
ls
on
c
onve
nt
io
na
l
a
nd
c
om
pl
e
x
ne
twor
k
th
r
e
a
ts
.
T
oN
-
I
oT
[
33]
:
b
ui
ld
in
g
upon
th
e
U
N
S
W
-
N
B
15
f
r
a
m
e
w
or
k,
th
e
T
oN
-
I
oT
da
ta
s
e
t
in
c
or
por
a
te
s
da
ta
f
r
om
r
e
a
l
-
w
or
ld
I
oT
in
f
r
a
s
tr
uc
tu
r
e
s
,
e
nc
om
pa
s
s
in
g
s
ys
t
e
m
lo
gs
,
te
le
m
e
tr
y,
a
nd
ne
twor
k
a
c
ti
vi
ty
a
c
r
os
s
m
ul
ti
pl
e
la
ye
r
s
.
A
tt
a
c
k
ty
pe
s
in
c
lu
de
r
a
ns
om
w
a
r
e
,
ba
c
kdoor
s
,
li
nk
-
la
ye
r
a
tt
a
c
ks
,
m
a
lwa
r
e
in
f
e
c
ti
on
s
,
a
nd
D
oS
a
tt
a
c
ks
.
T
hi
s
da
ta
s
e
t
e
f
f
e
c
ti
ve
ly
s
im
ul
a
te
s
dyna
m
ic
I
o
T
e
nvi
r
onm
e
nt
s
a
nd
of
f
e
r
s
m
ul
ti
-
s
our
c
e
,
he
te
r
oge
ne
ou
s
da
t
a
s
tr
e
a
m
s
,
e
n
a
bl
in
g
r
obus
t
tr
a
in
in
g
a
nd
te
s
ti
ng
of
e
dge
-
ba
s
e
d
in
tr
us
io
n
de
te
c
ti
on
m
ode
ls
.
T
he
s
e
d
a
ta
s
e
ts
w
e
r
e
s
e
l
e
c
te
d
f
or
th
e
ir
hi
gh
r
e
le
v
a
nc
e
to
c
ont
e
m
por
a
r
y
I
oT
-
e
dge
C
P
S
de
pl
oym
e
nt
s
a
nd t
he
ir
a
bi
li
ty
t
o r
e
f
le
c
t
r
e
a
l
-
w
or
ld
t
hr
e
a
t
s
c
e
na
r
io
s
i
n s
m
a
r
t
e
nvi
r
onm
e
nt
s
.
S
ta
nda
r
d
c
la
s
s
if
ic
a
ti
on
m
e
a
s
ur
e
s
w
e
r
e
us
e
d
to
e
va
lu
a
te
th
e
D
A
P
A
-
I
W
G
B
m
ode
l'
s
de
te
c
ti
on
pe
r
f
or
m
a
nc
e
.
T
he
s
e
m
e
tr
ic
s
,
w
hi
c
h
m
e
a
s
ur
e
how
w
e
ll
th
e
m
ode
l
di
s
ti
ngui
s
he
s
be
tw
e
e
n
a
tt
a
c
k
a
nd
nor
m
a
l
tr
a
f
f
ic
, a
r
e
a
s
i
n (
17)
t
o (
20)
.
=
+
+
+
+
(
17)
=
+
(
18)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
D
y
nam
ic
at
ta
c
k
pat
te
r
n
-
aw
a
r
e
i
nt
e
ll
ig
e
nt
c
y
b
e
r
-
phy
s
ic
al
…
(
V
is
hal
a I
bas
apur
a L
ak
s
hm
in
a
r
ay
anappa
)
587
=
+
(
19)
1
−
=
2
×
×
+
(
20)
W
he
r
e
T
P
,
T
N
,
F
P
,
a
nd
F
N
in
di
c
a
te
tr
ue
po
s
it
iv
e
,
tr
ue
n
e
ga
ti
ve
,
f
a
ls
e
po
s
it
iv
e
s
,
a
nd
f
a
ls
e
ne
g
a
ti
ve
,
r
e
s
pe
c
ti
ve
ly
.
T
og
e
th
e
r
,
th
e
s
e
m
e
tr
ic
s
gi
v
e
s
a
hol
is
ti
c
vi
e
w
of
de
te
c
ti
on
a
c
c
ur
a
c
y,
s
e
ns
it
iv
it
y,
a
nd
r
obus
tn
e
s
s
to
f
a
ls
e
a
la
r
m
s
.
T
he
pe
r
f
or
m
a
nc
e
of
th
e
D
A
P
A
-
I
W
G
B
m
ode
l
w
a
s
e
va
lu
a
te
d
a
c
r
os
s
bot
h
da
ta
s
e
ts
us
in
g
th
e
a
f
or
e
m
e
nt
io
ne
d
m
e
tr
ic
s
.
E
xpe
r
im
e
nt
a
l
out
c
om
e
s
on
U
N
S
W
-
N
B
15
da
ta
s
e
t
a
r
e
pr
e
s
e
nt
e
d
in
F
ig
ur
e
3,
w
h
e
r
e
th
e
m
ode
l
a
c
hi
e
v
e
d
a
n
a
c
c
ur
a
c
y
of
99.93%
,
pr
e
c
is
io
n
of
9
9.965%
,
r
e
c
a
ll
of
99.967%
,
a
nd
F
1
-
s
c
or
e
of
99.961%
.
T
he
s
e
r
e
s
ul
ts
a
f
f
ir
m
th
e
m
ode
l'
s
hi
gh
pr
e
c
is
io
n
in
id
e
nt
if
yi
ng
m
a
li
c
io
us
tr
a
f
f
ic
a
nd
it
s
lo
w
r
a
te
of
f
a
ls
e
pos
it
iv
e
s
.
T
he
in
te
gr
a
ti
on
of
dyna
m
ic
s
ta
ti
s
ti
c
a
l
m
oni
to
r
in
g
a
nd
a
da
pt
iv
e
r
e
tr
a
in
in
g,
a
s
de
ta
il
e
d
in
(
12)
,
f
ur
th
e
r
s
tr
e
ngt
he
ns
de
te
c
ti
on s
ta
bi
li
ty
ove
r
t
im
e
by
r
e
c
a
li
br
a
ti
ng
t
r
e
e
w
e
ig
ht
s
i
n r
e
s
pons
e
t
o e
vol
vi
ng pa
tt
e
r
ns
.
S
im
il
a
r
ly
,
r
e
s
ul
ts
on
th
e
T
oN
-
I
oT
da
ta
s
e
t
a
r
e
s
how
n
in
F
ig
ur
e
4,
w
it
h
th
e
f
ol
lo
w
in
g
m
e
tr
ic
s
obs
e
r
ve
d
a
n
a
c
c
ur
a
c
y:
99.921%
,
pr
e
c
is
io
n
of
99.92%
,
r
e
c
a
ll
of
99.85%
,
a
nd
F
1
-
s
c
or
e
of
99.85%
.
T
he
hi
gh
de
te
c
ti
on
s
c
or
e
s
a
c
r
os
s
bot
h
da
ta
s
e
ts
va
li
da
te
th
e
ge
ne
r
a
li
z
a
bi
li
ty
a
nd
a
da
pt
a
bi
li
ty
of
D
A
P
A
-
I
W
G
B
unde
r
di
ve
r
s
e
a
tt
a
c
k
c
ondi
ti
ons
.
T
h
e
m
ode
l'
s
r
e
s
pons
iv
e
ne
s
s
to
t
e
m
por
a
l
s
hi
f
ts
in
da
ta
di
s
tr
ib
ut
io
n
is
f
a
c
il
it
a
te
d
by
a
r
e
tr
a
in
in
g
m
e
c
ha
ni
s
m
,
w
hi
c
h
upd
a
te
s
e
ns
e
m
bl
e
tr
e
e
s
tr
uc
tu
r
e
a
nd
w
e
ig
ht
di
s
tr
ib
ut
io
n
ba
s
e
d
on
r
e
a
l
-
ti
m
e
f
e
e
dba
c
k
a
nd
c
la
s
s
if
ic
a
ti
on
e
r
r
or
s
.
T
hi
s
a
da
pt
iv
e
le
a
r
ni
ng
m
e
c
ha
ni
s
m
e
ns
ur
e
s
lo
ng
-
te
r
m
r
e
li
a
bi
li
ty
in
f
a
s
t
-
c
ha
ngi
ng
I
oT
e
nvi
r
onm
e
nt
s
.
O
ve
r
a
ll
,
th
e
r
e
s
ul
t
s
de
m
ons
tr
a
te
th
a
t
D
A
P
A
-
I
W
G
B
e
f
f
e
c
ti
ve
ly
di
s
ti
ngui
s
he
s
be
twe
e
n
nor
m
a
l
a
nd
m
a
li
c
io
us
be
ha
vi
or
,
m
a
in
ta
in
s
r
e
s
il
i
e
nc
e
a
ga
in
s
t
c
on
c
e
pt
dr
if
t,
a
nd
m
in
im
iz
e
s
c
la
s
s
if
ic
a
ti
on e
r
r
or
s
i
n e
dge
-
ba
s
e
d C
P
S
i
nf
r
a
s
tr
uc
tu
r
e
s
.
F
ig
ur
e
3. D
A
P
A
-
I
W
G
B
p
e
r
f
or
m
a
nc
e
on UN
S
W
-
N
B
15 da
ta
s
e
t
F
ig
ur
e
4. D
A
P
A
-
I
W
G
B
p
e
r
f
or
m
a
nc
e
on T
oN
-
I
oT
da
ta
s
e
t
99.93
99.965
99.967
99.961
99.91
99.92
99.93
99.94
99.95
99.96
99.97
A
C
C
U
R
A
C
Y
P
R
E
C
I
S
I
O
N
R
E
C
A
L
L
F
-
S
C
O
R
E
P
e
r
c
e
n
t
age
(
%
)
P
e
r
f
or
m
an
c
e
M
e
t
r
i
c
s
D
A
P
A
-
I
W
G
B
P
E
R
F
O
R
M
A
N
C
E
-
U
N
S
W
-
N
B
1
5
99.921
99.92
99.85
99.85
99.8
99.82
99.84
99.86
99.88
99.9
99.92
99.94
A
C
C
U
R
A
C
Y
P
R
E
C
I
S
I
O
N
R
E
C
A
L
L
F
-
S
C
O
R
E
P
e
r
c
e
n
t
age
(
%
)
P
e
r
f
o
r
m
a
n
c
e
M
e
t
r
i
c
s
D
A
P
A
-
I
W
G
B
P
E
R
F
O
R
M
A
N
C
E
-
T
O
N
-
I
O
T
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
15
, N
o.
1
,
F
e
br
ua
r
y
20
26
:
580
-
591
588
T
he
s
ug
ge
s
te
d
D
A
P
A
-
I
W
G
B
a
r
c
hi
t
e
c
t
ur
e
w
a
s
c
om
p
a
r
e
d
to
a
num
be
r
of
c
ur
r
e
nt
a
nd
w
e
ll
-
kno
w
n
in
tr
us
io
n
de
t
e
c
ti
on
m
od
e
ls
i
n
I
oT
-
E
d
ge
c
ont
e
xt
s
i
n
or
de
r
to
f
ur
t
he
r
a
s
s
e
s
s
it
s
e
f
f
ic
a
c
y
a
nd
r
e
s
il
ie
n
c
e
. A
r
e
vi
e
w
of
pe
r
f
or
m
a
n
c
e
m
e
tr
ic
s
f
r
om
th
e
l
it
e
r
a
t
ur
e
,
in
c
lu
di
ng
a
s
a
c
c
ur
a
c
y,
pr
e
c
is
io
n,
r
e
c
a
l
l,
a
n
d
F
1
-
s
c
or
e
,
i
s
s
how
n
i
n
T
a
bl
e
1
.
I
t
i
s
c
l
e
a
r
f
r
om
T
a
b
le
1
th
a
t
th
e
s
u
gge
s
te
d
D
A
P
A
-
I
W
G
B
m
ode
l
pe
r
f
or
m
s
be
tt
e
r
th
a
n
c
ur
r
e
nt
m
e
th
ods
in
e
ve
r
y
s
ig
ni
f
i
c
a
n
t
pe
r
f
or
m
a
nc
e
c
a
t
e
gor
y
.
T
h
e
m
od
e
l
a
c
hi
e
ve
d
a
n
a
c
c
ur
a
c
y
of
99.9
21%
,
a
pr
e
c
i
s
io
n
of
99.92%
, r
e
c
a
ll
of
9
9.85%
,
a
nd
a
n F
1
-
s
c
or
e
of
99.85%
, de
m
ons
tr
a
ti
ng
s
up
e
r
io
r
c
l
a
s
s
if
i
c
a
ti
on c
a
pa
bi
li
ty
.
T
a
bl
e
1. D
A
P
A
-
I
W
G
B
pe
r
f
or
m
a
nc
e
c
om
pa
r
is
on w
it
h e
xi
s
ti
ng
a
ppr
oa
c
he
s
u
s
in
g t
he
T
oN
-
I
oT
da
t
a
s
e
t
R
e
f
e
r
e
nc
e
/
Y
e
a
r
M
ode
l
A
c
c
ur
a
c
y
P
r
e
c
i
s
i
on
R
e
c
a
l
l
F1
-
s
c
or
e
L
oge
s
w
a
r
i
e
t
al
.
[
17]
, 2025
H
F
O
M
C
98.83
98.56
98.81
98.65
S
a
i
ye
d
a
nd A
l
-
A
nba
gi
[
18]
, 2024
G
A
T
D
D
95
95
95
94
C
ui
e
t
al
.
[
19]
,
2024
D
R
N
-
AM
99.24
99.18
99.15
99.16
J
a
ve
e
d
e
t
al
.
[
20]
,
2024
I
D
S
-
S
A
E
L
99.55
99.31
99.24
99.39
Y
a
ng
e
t
al
.
[
21]
, 2024
GAD
-
E
L
A
R
97.67
98.21
97.67
97.84
B
ouz
i
ni
s
e
t
al
.
[
22]
, 2025
S
t
a
t
A
vg
93.38
-
-
61.22
F
a
r
e
s
e
t
al
.
[
23]
, 2025
ST
-
L
S
T
M
-
D
T
L
98.8
98.24
98.67
98.45
L
i
e
t
a
l
.
[
24]
, 2025
S
M
V
L
98.7
-
-
92.5
C
ha
ndna
ni
e
t
al
.
[
29]
, 2025
F
e
d
-
M
L
D
L
98.1
98.3
98.3
98.3
T
hi
s
r
e
s
e
a
r
c
h
D
A
P
A
-
I
W
G
B
99.921
99.92
99.85
99.85
C
om
pa
r
e
d
to
de
e
p
r
e
c
ur
r
e
nt
m
ode
ls
s
uc
h
a
s
ge
ne
ti
c
a
lg
or
it
hm
e
m
pl
oyi
ng
t
-
T
e
s
t
-
ba
s
e
d
D
D
oS
de
te
c
ti
on
(
G
A
T
D
D
)
[
18]
,
hybr
id
m
ode
ls
li
ke
D
R
N
-
A
M
[
19]
a
nd
I
D
S
-
S
A
E
L
[
20]
,
a
nd
th
e
D
A
P
A
-
I
W
G
B
e
xhi
bi
ts
a
not
a
bl
e
in
c
r
e
a
s
e
in
bot
h
pr
e
c
is
io
n
a
nd
r
e
c
a
ll
,
in
di
c
a
ti
ng
a
r
e
duc
e
d
f
a
ls
e
pos
it
iv
e
r
a
te
a
nd
im
pr
ove
d
s
e
ns
it
iv
it
y
to
a
c
tu
a
l
in
tr
us
io
ns
.
M
ode
l
s
s
uc
h
a
s
hybr
id
f
e
a
tu
r
e
opt
im
iz
a
ti
on
th
r
ough
m
ul
ti
-
s
ta
ge
c
la
s
s
if
ie
r
(
H
F
O
M
C
)
[
17]
a
nd
ST
-
L
S
T
M
-
D
T
L
[
2
3]
pe
r
f
or
m
r
e
a
s
o
na
bl
y
w
e
l
l;
ho
w
e
ve
r
,
th
e
ir
m
e
tr
i
c
s
r
e
m
a
in
m
a
r
g
in
a
ll
y
lo
w
e
r
t
ha
n
D
A
P
A
-
I
W
G
B
, pa
r
ti
c
ul
a
r
l
y
in
F
1
-
s
c
or
e
, w
h
ic
h
in
di
c
a
te
s
a
b
a
l
a
n
c
e
b
e
tw
e
e
n pr
e
c
i
s
i
on a
n
d
r
e
c
a
ll
. T
h
e
G
A
T
D
D
[
18]
m
od
e
l,
a
lt
hou
gh
b
a
s
e
d
o
n
e
n
s
e
m
bl
e
p
r
in
c
i
pl
e
s
,
r
e
c
or
ds
a
l
ow
e
r
p
e
r
f
or
m
a
n
c
e
(
F
1
-
s
c
or
e
of
9
4%
)
,
hi
gh
li
g
ht
i
ng
li
m
it
a
ti
on
s
i
n dyn
a
m
ic
p
a
tt
e
r
n
d
e
t
e
c
ti
o
n
i
n e
v
ol
vi
ng
I
o
T
e
n
vi
r
o
nm
e
nt
s
. M
or
e
o
ve
r
,
t
he
p
e
r
f
or
m
a
nc
e
of
m
od
e
l
s
li
k
e
S
t
a
tA
vg
[
2
2]
a
n
d
S
M
V
L
[
24]
s
ug
ge
s
t
s
s
ig
ni
f
i
c
a
nt
tr
a
d
e
-
of
f
s
in
d
e
t
e
c
ti
o
n
pe
r
f
or
m
a
n
c
e
,
w
i
th
e
it
he
r
i
nc
om
p
le
te
m
e
tr
ic
r
e
po
r
ti
n
g
or
c
om
p
a
r
a
ti
ve
ly
l
ow
e
r
F
1
-
s
c
or
e
s
.
N
ot
a
b
ly
,
F
e
d
-
M
L
D
L
[
29]
a
c
hi
e
ve
s
c
om
pe
ti
ti
ve
r
e
s
ul
ts
,
ye
t
s
ti
ll
unde
r
pe
r
f
or
m
s
r
e
la
ti
ve
to
D
A
P
A
-
I
W
G
B
,
pa
r
ti
c
ul
a
r
ly
in
c
a
pt
ur
in
g
r
a
r
e
a
nd
e
vol
vi
ng a
tt
a
c
k i
ns
ta
nc
e
s
due
t
o i
ts
s
ta
ti
c
f
e
de
r
a
te
d de
s
ig
n.
T
he
s
upe
r
io
r
r
e
s
ul
ts
o
f
D
A
P
A
-
I
W
G
B
c
a
n
be
a
tt
r
ib
u
te
d
to
it
s
:
dyna
m
ic
w
e
i
ght
a
d
ju
s
tm
e
n
t,
w
hi
c
h
r
e
duc
e
s
m
ode
l
bi
a
s
to
w
a
r
d
m
a
jo
r
it
y
c
la
s
s
e
s
,
r
e
a
l
-
t
im
e
t
r
e
e
pr
u
ni
ng
a
nd
a
u
gm
e
nt
a
ti
on
,
e
na
bl
i
ng
a
da
pt
a
bi
l
it
y
to
a
tt
a
c
k
e
vol
ut
io
n.
I
n
a
d
di
ti
on,
s
ta
t
is
ti
c
a
l
m
o
ni
to
r
i
ng
f
a
c
il
it
a
te
s
t
im
e
ly
r
e
tr
a
in
in
g
in
r
e
s
po
ns
e
to
di
s
tr
i
but
i
ona
l
s
hi
f
ts
.
I
n
s
um
m
a
r
y,
th
e
D
A
P
A
-
I
W
G
B
f
r
a
m
e
w
o
r
k
de
m
ons
t
r
a
te
s
s
ta
te
-
of
-
th
e
-
a
r
t
pe
r
f
or
m
a
nc
e
a
c
r
os
s
s
ta
nda
r
d
be
nc
h
m
a
r
ks
a
n
d
va
l
id
a
te
s
i
ts
e
f
f
ic
a
c
y
f
o
r
s
e
c
ur
in
g
I
oT
-
e
dge
-
e
na
bl
e
d
C
P
S
a
ga
in
s
t
a
w
id
e
r
a
nge
o
f
c
ybe
r
t
h
r
e
a
ts
.
U
s
in
g
th
e
U
N
S
W
-
N
B
15
be
nc
hm
a
r
k
da
ta
s
e
t,
a
c
om
pa
r
a
ti
ve
pe
r
f
or
m
a
nc
e
e
va
lu
a
ti
on
w
a
s
c
a
r
r
ie
d
out
to
ve
r
if
y
th
e
r
obus
tn
e
s
s
a
nd
de
te
c
ti
on
a
c
c
ur
a
c
y
of
th
e
s
ugge
s
t
e
d
D
A
P
A
-
I
W
G
B
m
ode
l.
T
he
pe
r
f
or
m
a
nc
e
of
D
A
P
A
-
I
W
G
B
is
s
how
n
in
T
a
bl
e
2
a
lo
ngs
id
e
m
a
ny
ne
w
ly
r
e
le
a
s
e
d
in
tr
us
io
n
de
te
c
ti
on
m
ode
l
s
th
a
t
ha
ve
a
l
s
o
us
e
d
th
e
U
N
S
W
-
N
B
15
d
a
ta
s
e
t
a
s
a
n
e
va
lu
a
ti
on
be
nc
hm
a
r
k.
T
he
s
ugg
e
s
te
d
D
A
P
A
-
I
W
G
B
c
ons
id
e
r
a
bl
y
out
pe
r
f
or
m
s
a
ll
e
va
lu
a
te
d
m
ode
ls
a
c
r
os
s
a
ll
a
s
s
e
s
s
m
e
nt
m
e
tr
ic
s
,
a
s
s
ho
w
n
in
T
a
bl
e
2.
W
it
h
99.93%
a
c
c
ur
a
c
y,
99.965%
pr
e
c
is
io
n, 99.967%
r
e
c
a
ll
, a
nd 99.961%
F
1
-
s
c
or
e
, i
t
s
ig
ni
f
ic
a
nt
ly
out
pe
r
f
or
m
s
c
ur
r
e
nt
m
e
th
ods
.
T
a
bl
e
2. D
A
P
A
-
I
W
G
B
p
e
r
f
or
m
a
nc
e
c
om
pa
r
is
on w
it
h e
xi
s
ti
ng
a
ppr
oa
c
he
s
u
s
in
g U
N
S
W
-
N
B
15 be
nc
hm
a
r
k
R
e
f
e
r
e
nc
e
/
Y
e
a
r
M
ode
l
A
c
c
ur
a
c
y
P
r
e
c
i
s
i
on
R
e
c
a
l
l
F1
-
s
c
or
e
C
ui
e
t
al
.
[
19]
, 2024
D
R
N
-
AM
89.23
883.83
87.77
88.25
Y
a
ng
e
t
al
.
[
21]
, 2024
GAD
-
E
L
A
R
98.84
98.8
98.84
98.78
L
i
e
t
al
.
[
24]
, 2025
S
M
V
L
97.9
-
-
85.6
L
i
e
t
a
l
.
[
25]
, 2025
S
L
R
F
90.43
87.69
83.12
-
W
u
e
t
al
.
[
26]
, 2025
T
R
A
C
E
R
86.02
-
-
-
B
i
a
n a
nd L
i
u
[
28]
, 2025
G
M
C
W
A
E
81.1
-
-
-
E
l
a
z
i
z
e
t
al
.
[
30]
, 2025
TTF
-
E
E
F
O
98.50
98.30
98.20
98.25
T
hi
s
r
e
s
e
a
r
c
h
D
A
P
A
-
I
W
G
B
99.93
99.965
99.967
99.961
C
om
pa
r
e
d
to
r
e
c
e
nt
D
L
a
ppr
oa
c
he
s
s
uc
h
a
s
G
A
D
-
E
L
A
R
[
21]
a
nd
f
r
a
m
e
w
or
k
ba
s
e
d
on
th
e
t
a
b
tr
a
ns
f
or
m
e
r
-
e
le
c
tr
ic
e
e
l
f
o
r
a
gi
ng
opt
im
iz
a
ti
on
(
T
T
F
-
E
E
F
O
)
[
30]
,
w
hi
c
h
a
ls
o
e
xhi
bi
t
hi
gh
a
c
c
ur
a
c
y
a
nd
ba
la
nc
e
d
de
t
e
c
ti
on
pe
r
f
or
m
a
nc
e
,
th
e
D
A
P
A
-
I
W
G
B
m
ode
l
yi
e
ld
s
a
not
ic
e
a
bl
e
im
pr
ove
m
e
nt
in
pr
e
c
is
io
n
a
nd
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
D
y
nam
ic
at
ta
c
k
pat
te
r
n
-
aw
a
r
e
i
nt
e
ll
ig
e
nt
c
y
b
e
r
-
phy
s
ic
al
…
(
V
is
hal
a I
bas
apur
a L
ak
s
hm
in
a
r
ay
anappa
)
589
r
e
c
a
ll
,
r
e
f
le
c
ti
ng
it
s
e
nha
nc
e
d
a
bi
li
ty
to
di
s
ti
ngui
s
h
be
twe
e
n
be
ni
gn
a
nd
m
a
li
c
io
us
tr
a
f
f
ic
w
it
h
m
in
im
a
l
f
a
ls
e
a
la
r
m
s
.
T
h
e
F
1
-
s
c
or
e
im
pr
ove
m
e
nt
f
ur
th
e
r
de
m
ons
tr
a
te
s
t
he
pr
opos
e
d
m
ode
l'
s
c
a
pa
bi
li
ty
to
m
a
in
ta
in
de
te
c
ti
on
c
ons
i
s
te
nc
y
e
v
e
n
in
th
e
pr
e
s
e
nc
e
of
im
ba
la
n
c
e
d
da
ta
di
s
tr
ib
ut
io
ns
.
O
th
e
r
m
ode
l
s
li
ke
.
D
R
N
-
A
M
[
19]
,
S
M
V
L
[
24
]
,
a
nd
S
L
R
F
[
25
]
,
s
uf
f
e
r
f
r
om
lo
w
e
r
a
c
c
ur
a
c
y
a
nd
r
e
c
a
ll
,
in
di
c
a
ti
ng
s
e
ns
it
iv
it
y
to
f
e
a
tu
r
e
va
r
ia
bi
li
ty
a
nd
a
tt
a
c
k
di
ve
r
s
it
y.
A
ddi
ti
ona
ll
y,
s
om
e
m
ode
ls
s
uc
h
a
s
T
R
A
C
E
R
[
26]
,
R
e
M
e
N
e
t
[
27]
,
a
nd
G
M
C
W
A
E
[
28]
r
e
por
t
onl
y
pa
r
ti
a
l
m
e
t
r
ic
s
,
m
a
ki
ng
c
om
pr
e
he
ns
iv
e
c
om
pa
r
is
on
di
f
f
ic
ul
t
but
s
ti
ll
hi
ghl
ig
ht
in
g t
he
ir
l
im
it
a
ti
ons
i
n a
c
hi
e
vi
ng holi
s
ti
c
t
hr
e
a
t
de
te
c
ti
on.
T
he
pe
r
f
or
m
a
nc
e
a
dva
nt
a
ge
s
of
D
A
P
A
-
I
W
G
B
c
a
n
be
a
tt
r
ib
u
te
d
to
it
s
:
i
)
dyna
m
ic
a
tt
a
c
k
pa
tt
e
r
n
a
da
pt
a
ti
on,
w
hi
c
h
e
na
bl
e
s
le
a
r
ni
ng
f
r
om
e
vol
vi
ng
th
r
e
a
ts
;
i
i)
c
ova
r
ia
te
s
hi
f
t
ha
ndl
in
g
th
r
ough
a
da
pt
iv
e
e
ns
e
m
bl
e
upda
ti
ng;
ii
i)
m
in
or
it
y
c
la
s
s
r
e
-
w
e
ig
ht
in
g
a
nd
pr
uni
ng
s
tr
a
te
gi
e
s
,
w
hi
c
h
pr
e
ve
nt
bi
a
s
to
w
a
r
d
m
a
jo
r
it
y
c
la
s
s
e
s
;
a
nd
iv
)
r
e
a
l
-
ti
m
e
w
e
ig
ht
opt
im
iz
a
ti
on
a
nd
r
e
tr
a
in
in
g,
e
nha
nc
in
g
r
obus
tn
e
s
s
a
c
r
os
s
va
r
yi
ng
tr
a
f
f
ic
pa
tt
e
r
ns
.
T
he
s
e
r
e
s
ul
ts
a
f
f
ir
m
th
a
t
th
e
D
A
P
A
-
I
W
G
B
m
ode
l
is
hi
ghl
y
s
ui
ta
bl
e
f
or
r
e
a
l
-
ti
m
e
in
tr
us
io
n
de
te
c
ti
on
in
I
oT
-
e
dge
-
e
na
bl
e
d
C
P
S
e
nvi
r
onm
e
nt
s
,
o
f
f
e
r
in
g
s
upe
r
io
r
pe
r
f
o
r
m
a
nc
e
c
om
pa
r
e
d
to
e
xi
s
ti
ng
M
L
a
nd D
L
-
ba
s
e
d i
nt
r
us
io
n de
te
c
ti
on t
e
c
hni
que
s
on t
he
U
N
S
W
-
N
B
15 da
ta
s
e
t.
4.
C
O
N
C
L
U
S
I
O
N
T
he
r
a
pi
d
c
r
e
a
ti
on
of
th
e
I
oT
ha
s
r
e
vol
ut
io
ni
z
e
d
C
P
S
by
e
na
bl
in
g
r
e
a
l
-
ti
m
e
da
ta
e
xc
ha
nge
be
twe
e
n
s
m
a
r
t
de
vi
c
e
s
,
e
dge
node
s
,
a
nd
c
lo
ud
in
f
r
a
s
tr
uc
tu
r
e
s
.
D
e
s
pi
te
th
e
s
e
a
dva
n
c
e
m
e
nt
s
,
s
e
c
ur
in
g
s
uc
h
he
te
r
oge
ne
ous
a
nd
dyna
m
ic
e
nvi
r
onm
e
nt
s
r
e
m
a
in
s
a
pr
e
s
s
in
g
c
ha
ll
e
nge
due
to
th
e
e
vol
vi
ng
na
tu
r
e
of
c
ybe
r
th
r
e
a
ts
a
nd
th
e
li
m
it
a
ti
ons
o
f
c
onve
nt
io
na
l
de
te
c
ti
on
m
e
th
ods
.
T
o
a
ddr
e
s
s
th
is
is
s
ue
,
th
is
pa
pe
r
pr
opos
e
d
th
e
D
A
P
A
-
I
W
G
B
m
ode
l,
w
hi
c
h
in
te
gr
a
te
s
a
s
ta
ti
s
ti
c
a
l
m
oni
to
r
in
g m
e
c
ha
ni
s
m
to
dyna
m
ic
a
ll
y
d
e
te
c
t
a
nd
a
d
a
pt
to
s
hi
f
ts
in
a
tt
a
c
k
be
ha
vi
or
.
T
he
m
ode
l
e
m
pl
oys
a
da
pt
iv
e
r
e
-
w
e
ig
ht
in
g,
r
e
a
l
-
ti
m
e
tr
e
e
pr
uni
ng,
a
nd
c
ova
r
ia
te
s
hi
f
t
ha
ndl
in
g
to
m
a
in
ta
in
hi
gh
de
te
c
ti
on
f
id
e
li
ty
in
c
om
pl
e
x
I
oT
-
e
dge
s
c
e
na
r
io
s
.
E
xpe
r
im
e
nt
a
l
e
va
lu
a
ti
ons
us
in
g
two
be
nc
hm
a
r
k
da
ta
s
e
ts
,
U
N
S
W
-
N
B
15
a
nd
T
oN
-
I
oT
,
d
e
m
ons
tr
a
te
d
th
e
e
f
f
e
c
ti
ve
ne
s
s
of
th
e
pr
opos
e
d
a
ppr
oa
c
h.
T
he
D
A
P
A
-
I
W
G
B
m
ode
l
a
c
hi
e
ve
d
s
upe
r
io
r
pe
r
f
or
m
a
nc
e
,
w
it
h
a
c
c
ur
a
c
y
l
e
ve
ls
of
99.93%
a
nd
99.921%
,
r
e
s
pe
c
ti
ve
ly
,
w
hi
le
m
a
in
ta
in
in
g
hi
gh
pr
e
c
is
io
n
a
nd
lo
w
f
a
ls
e
pos
it
iv
e
r
a
te
s
.
T
h
e
s
e
r
e
s
ul
t
s
va
li
da
te
th
e
m
ode
l’
s
c
a
pa
bi
li
ty
to
ha
ndl
e
c
la
s
s
im
ba
la
nc
e
a
nd
c
onc
e
pt
dr
if
t,
e
ns
ur
in
g
r
e
li
a
bl
e
in
tr
us
io
n
d
e
te
c
ti
on
in
dyna
m
ic
e
dge
-
ba
s
e
d
C
P
S
e
nvi
r
onm
e
nt
s
.
A
s
pa
r
t
of
f
ut
ur
e
r
e
s
e
a
r
c
h,
w
e
in
te
nd
to
de
v
e
lo
p
a
m
or
e
ge
ne
r
a
li
z
e
d
e
ns
e
m
bl
e
-
ba
s
e
d
in
tr
us
io
n
de
te
c
ti
on
f
r
a
m
e
w
or
k
th
a
t
in
c
or
por
a
te
s
hybr
id
a
r
ti
f
ic
ia
l
in
te
ll
ig
e
nc
e
te
c
hni
que
s
,
s
uc
h
a
s
de
e
p
r
e
in
f
or
c
e
m
e
nt
le
a
r
ni
ng,
f
e
de
r
a
te
d
le
a
r
ni
ng,
a
n
d
gr
a
ph
-
ba
s
e
d
m
ode
ls
.
T
hi
s
f
r
a
m
e
w
or
k
w
il
l
e
xpl
ic
it
ly
a
ddr
e
s
s
is
s
ue
s
of
da
ta
he
te
r
oge
ne
it
y,
c
ova
r
ia
te
s
hi
f
t,
a
nd
c
la
s
s
im
ba
la
nc
e
,
a
nd
w
il
l
be
e
va
lu
a
te
d
us
in
g
m
or
e
r
e
a
li
s
ti
c
a
nd
la
r
ge
-
s
c
a
le
da
ta
s
e
t
s
dr
a
w
n
f
r
om
I
oT
,
in
te
r
ne
t
of
ve
hi
c
le
s
(
I
oV
)
,
a
nd
C
P
S
dom
a
in
s
.
S
uc
h
a
dva
nc
e
m
e
nt
s
a
im
to
e
nha
n
c
e
th
e
s
c
a
la
bi
li
ty
,
r
e
s
il
ie
nc
e
,
a
nd
a
d
a
pt
a
bi
li
ty
of
I
D
S
in
ne
xt
-
ge
ne
r
a
ti
on
in
te
ll
ig
e
nt
i
nf
r
a
s
tr
uc
tu
r
e
s
A
C
K
N
O
WL
E
D
G
M
E
N
T
S
W
e
w
oul
d
li
ke
to
th
a
nk
our
s
upe
r
vi
s
or
f
or
hi
s
a
s
s
is
ta
nc
e
,
c
ons
ta
nt
f
e
e
dba
c
k,
a
nd
m
ot
iv
a
ti
on
f
or
th
e
e
nt
ir
e
dur
a
ti
on of
t
he
r
e
s
e
a
r
c
h.
F
U
N
D
I
N
G
I
N
F
O
R
M
A
T
I
O
N
N
o f
undi
ng i
s
r
a
is
e
d f
or
t
hi
s
r
e
s
e
a
r
c
h.
A
U
T
H
O
R
C
O
N
T
R
I
B
U
T
I
O
N
S
S
T
A
T
E
M
E
N
T
T
hi
s
jo
ur
na
l
us
e
s
th
e
C
ont
r
ib
ut
or
R
ol
e
s
T
a
xonomy
(
C
R
e
di
T
)
to
r
e
c
ogni
z
e
in
di
vi
dua
l
a
ut
hor
c
ont
r
ib
ut
io
ns
, r
e
duc
e
a
ut
hor
s
hi
p di
s
put
e
s
,
a
nd f
a
c
il
it
a
te
c
ol
la
bo
r
a
ti
on.
N
am
e
o
f
A
u
t
h
or
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
V
is
ha
la
I
ba
s
a
pur
a
L
a
ks
hm
in
a
r
a
ya
na
ppa
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
K
e
m
pa
ha
num
a
ia
h
M
.
R
a
vi
kum
a
r
✓
✓
✓
✓
✓
✓
✓
✓
✓
C
:
C
onc
e
pt
ua
l
i
z
a
t
i
on
M
:
M
e
t
hodol
ogy
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
da
t
i
on
Fo
:
Fo
r
m
a
l
a
na
l
ys
i
s
I
:
I
nve
s
t
i
ga
t
i
on
R
:
R
e
s
our
c
e
s
D
:
D
a
t
a
C
ur
a
t
i
on
O
:
W
r
i
t
i
ng
-
O
r
i
gi
na
l
D
r
a
f
t
E
:
W
r
i
t
i
ng
-
R
e
vi
e
w
&
E
di
t
i
ng
Vi
:
Vi
s
ua
l
i
z
a
t
i
on
Su
:
Su
pe
r
vi
s
i
on
P
:
P
r
oj
e
c
t
a
dm
i
ni
s
t
r
a
t
i
on
Fu
:
Fu
ndi
ng a
c
qui
s
i
t
i
on
Evaluation Warning : The document was created with Spire.PDF for Python.