Int
ern
at
i
onal
Journ
al of Ele
ctrical
an
d
Co
mput
er
En
gin
eeri
ng
(IJ
E
C
E)
Vo
l.
11
,
No.
1
,
Febr
uar
y
2021
, pp.
53
6
~
544
IS
S
N:
20
88
-
8708
,
DOI: 1
0.1
1591/
ijece
.
v
11
i
1
.
pp
536
-
54
4
536
Journ
al h
om
e
page
:
http:
//
ij
ece.i
aesc
or
e.c
om
Secured
node det
ec
ti
on te
chniq
ue base
d
on
artifi
cial neur
al
network
for wirel
ess senso
r netw
ork
Bassam H
asa
n
1
,
S
ameer
Al
an
i
2
,
M
oham
med Ay
ad
Saad
3
1
Depa
rtment of
El
e
ct
roni
cs
and
Com
m
unic
at
ion Engi
ne
eri
ng,
Co
ll
eg
e
of
Engi
n
eering,
Univer
siti
Te
n
ag
a
Nasiona
l, Mala
y
sia
2
Depa
rtment of
Com
pute
r
Techn
ic
a
l
Eng
ineeri
ng
,
Al
-
Kit
ab
Univ
e
rsit
y
Coll
ege,
Ir
a
q
3
Depa
rtment of
Medic
a
l
Instrum
ent
a
ti
ons T
ec
hn
i
que
Eng
ine
e
ring
,
Al
-
Kit
ab
Univ
e
rsit
y
,
Ir
aq
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Ma
r
18
, 20
20
Re
vised
Jun
17
, 20
20
Accepte
d Aug
2
1
, 20
20
The
wire
le
ss
sensor
net
work
is
bec
om
ing
the
m
ost
popula
r
net
work
in
the
la
st
r
ec
en
t
y
ea
rs
as
it
ca
n
m
e
asure
the
env
iro
nm
ent
al
conditio
ns
and
send
the
m
to
proc
ess
purposes.
Many
vi
ta
l
challe
ng
es
fac
e
the
dep
lo
y
m
ent
o
f
W
SNs suc
h
as
e
ner
g
y
consum
ption a
nd
se
cur
ity
i
ss
ues.
Vari
ous a
t
ta
cks
coul
d
be
subjects
against
W
SN
s
and
ca
u
se
damag
e
ei
th
er
in
th
e
stabi
lit
y
of
comm
unic
at
ion
or
in
the
destruction
of
the
sensit
ive
data.
Thus
,
t
he
demands
of
int
rusion
det
e
ct
ion
-
b
ase
d
ene
r
g
y
-
e
fficie
nt
t
ec
h
nique
s
rise
dra
m
at
i
ca
l
l
y
as
the
net
work
dep
lo
y
m
ent
b
ec
om
e
s
vast
and
comp
li
c
at
ed
.
Qualnet
sim
ula
ti
on
is
used
to
m
e
asure
th
e
per
form
anc
e
of
th
e
n
etw
orks.
Thi
s
pa
per
a
ims
to
opti
m
iz
e
th
e
en
erg
y
-
base
d
in
tru
sion
det
ection
t
ec
hniqu
e
using
the
artifi
ci
a
l
neur
al
n
et
work
b
y
using
MA
TL
AB
Sim
uli
nk.
The
r
esult
s
show
how
the
opti
m
i
ze
d
m
et
hod
base
d
o
n
the
bio
logi
c
al
ner
vous
s
y
ste
m
s
improves
int
rusion
detec
t
i
on
in
W
SN
.
In
addi
ti
on
to
th
a
t,
the
unse
cur
ed
nodes
are
aff
ecte
d
th
e
ne
twork
per
form
anc
e
neg
at
iv
ely
and
troubl
e
i
t
s
beha
vior.
The
r
egr
ess
an
aly
sis
for
both
m
e
thods
detec
ts
th
e
var
ia
t
ions
whe
n
all
nodes
are
sec
ure
d
and
when
so
m
e
are
unsec
ure
d.
Thu
s,
Node
det
ec
t
io
n
base
d
on
pac
ke
t
de
li
ver
y
rat
io
and
en
erg
y
consum
ption
coul
d
eff
icient
l
y
b
e
implemente
d
in an
ar
ti
fi
cial
n
eur
al
n
et
work.
Ke
yw
or
ds:
ANN
IDS
Qu
al
net
WSN
This
is an
open
acc
ess arti
cl
e
un
der
the
CC
B
Y
-
SA
l
ic
ense
.
Corres
pond
in
g
Aut
h
or
:
Ba
ssam
H
asan,
Dep
a
rtm
ent o
f El
ect
ro
nics
and C
omm
un
ic
ation
En
gin
ee
rin
g
,
Un
i
ver
sit
i Te
na
ga Nasi
onal
,
Sela
ngor,
Ma
la
ysi
a
.
Em
a
il
: bassa
m
hasa
n92@gm
a
il
.co
m
1.
INTROD
U
CTION
Adva
nces
in
el
ect
ronics
and
wireless
com
m
un
ic
at
io
n
te
ch
no
l
og
ie
s
ha
ve
enab
le
d
the
de
velo
pm
ent
of
la
rg
e
-
scal
e
wir
el
ess
sensor
ne
tworks
(
WSN
s
)
that
con
sist
of
distrib
uted,
auto
no
m
ou
s
,
low
-
pow
er
,
lo
w
-
c
os
t,
sm
a
ll
-
siz
e
senso
r
node
s
to
c
ollec
t
inf
or
m
ation
a
nd
co
ope
rati
vely
trans
m
it
data
throu
gh
inf
rastr
uctu
re
-
le
s
s
wireless
net
work
s
as
s
how
n
in
Fig
u
r
e
1
[1
-
3]
.
Secur
it
y
appl
ic
at
ion
s
su
c
h
as
intru
si
on
preven
ti
on
or
det
ect
io
n
in
su
c
h
res
ou
rce
-
c
onstrai
ne
d
re
veal
sig
nificant
chall
en
ges
an
d
th
e
m
ai
n
fo
c
us
of
this
pa
per
.
WSN
i
s
beco
m
ing
i
ncrea
sing
ly
popu
la
r
as
it
e
na
bles
se
nsor
node
s
to
m
eas
ur
e
the
surr
oundin
g
en
vir
onm
ent,
com
m
un
ic
at
e
a
nd
process
m
easur
e
d
data
[4
-
6]
.
W
S
N
has
be
en
di
rected
f
r
om
m
ilit
ary
app
li
cat
ion
s
to
va
rio
us
ci
vil
app
li
cat
io
ns
,
es
pecial
ly
i
n
hosti
le
areas
[7]
.
Me
dical
,
i
ndus
tria
l
an
d
s
m
art
ener
gy
a
ppli
cat
ion
s
are
s
ti
ll
in
need
of
e
xten
s
ive
resea
rc
h
due
to
d
i
ff
e
ren
t
chall
en
ges
e
nc
ountere
d
[8
,
9]
.
E
nergy
c
onsu
m
ption
is
on
e
of
the
vital
chall
e
ng
e
s
that
face
WSNs'
researc
h.
N
od
es
ar
e
s
upplied
with
ba
tt
eries
that
cannot
be
rec
ha
r
ged
or
rep
la
ce
d
in
th
e
fiel
d
of
ope
rati
on
[
10
-
12]
.
Ma
nagem
ent
of
WSN'
s
ene
rg
y
helps
to
i
ncr
ease
the
ne
twork
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N:
20
88
-
8708
Secured
no
de dete
ct
ion t
ech
ni
qu
e
base
d on
ar
ti
fi
ci
al n
eu
ral
n
et
work f
or
w
ire
le
ss sen
s
or
.
..
(
Ba
ssam
H
asan
)
537
li
fetim
e.
No
wa
days,
WSN
ha
s
num
ero
us
a
ppli
cat
ion
s
i
n
m
il
it
ary,
healt
h
a
nd
en
vir
onm
ental
areas
due
t
o
ease
of
us
e
a
nd
ha
ving
the
abili
ty
to
withsta
nd
harsh
e
nvir
onm
ental
con
di
ti
on
s
[13
-
15]
.
These
netw
ork
s
are
sel
f
-
a
dm
inist
er
ed
net
works
in
w
hich
node
s
are
sel
f
-
org
anized
to
ha
ve
reli
able
co
m
m
un
ic
at
ion
betwee
n
them
.To
hav
e
secur
e
c
omm
un
ic
at
io
n
am
on
g
var
i
ous
sel
f
-
orga
nized
nodes,
sec
ur
it
y
issues
a
re
of
m
ai
n
con
ce
r
n.
T
here
are
va
rio
us
ty
pes
of
at
ta
c
ks
that
vulne
r
able
to
WSN
and
el
im
i
nat
e
the
com
m
u
nicat
ion
betwee
n
the
node
s.
S
o,
m
any
stud
ie
s
f
ocus
on
detect
ing
the
int
ru
si
on
in
WSN
by
diff
e
re
nt
al
gor
it
h
m
s
and ap
proac
he
s.
Figure
1.
WSN stru
ct
ur
e
The
m
ai
n
pr
oblem
of
intru
sio
n
is
m
iss
-
conn
ect
ing
the
com
m
un
ic
at
ion
between
the
c
onne
ct
ed
nodes
,
wh
ic
h
le
d
to
dro
p
t
he
pac
kets
an
d
re
du
ce
t
he
t
hroug
hput
[16
,
17]
.
D
ue
t
o
t
he
la
ck
of
a
so
li
d
li
ne
of
de
fen
se
li
ke
gateways
or
s
witc
hes
to
m
on
it
or
the
i
nfor
m
at
ion
flo
w,
the
sec
uri
ty
of
WSN
is
a
sign
ific
a
nt
crit
ic
al
pro
blem
,
especial
ly
fo
r
a
pp
l
ic
at
ion
s
w
he
re
co
nf
ide
ntial
it
y
has
pr
im
e
i
m
po
rtance
[
6,
18]
.
It
is
obvi
ou
s
t
o
con
cl
ud
e
that
tradit
ion
al
sec
ur
it
y
so
l
utions
of
wire
d/wi
re
le
ss
netw
orks
would
not
be
feasible
for
W
SN
s.
Th
us
,
dif
fer
e
nt
ty
pes
of
al
go
rithm
s
and
arc
hi
te
ct
ur
e
are
ava
il
able
to
fi
nd
t
he
tr
us
te
d
no
de
an
d
to
f
in
d
s
ecur
e
d
routes.
O
ver
t
he
ye
ars
,
a
la
r
ge
nu
m
ber
of
us
ef
ul
te
ch
niques
hav
e
bee
n
util
iz
ed
to
i
nvest
igate
a
nd
dev
e
l
op
the
pe
rfo
rm
ance
of
WSN.
T
he
recent
st
udy
of
[
19
,
20
]
r
eviews
dif
fer
e
nt
bi
o
-
i
ns
pi
red
te
chn
i
qu
e
s
de
velo
pe
d
for
i
m
pr
ovin
g
the
cy
ber
secu
rity
of
cy
ber
-
ph
ysi
cal
syst
em
s
us
ed
in
WSN
s.
T
he
dra
wb
ac
ks
of
pri
or
bi
o
-
insp
ire
d
a
ppr
oa
ches
im
po
se
d
the
resea
rc
her
s
to
pro
pose
a
ge
ne
ric
bio
-
ins
pire
d
m
od
el
cal
le
d
swar
m
intel
li
gen
ce
f
or
W
S
N
cy
be
rse
cur
it
y
(SI
W
C
).
The
new
sc
he
m
e
sh
ows
hi
gh
perform
ance
with
lo
w
com
plexity
.
A
com
par
at
ive
stud
y
a
nd
s
um
m
arization
of
intr
us
io
n
detect
ion
a
ppro
ac
hes
us
ed
i
n
W
SN
wer
e
rev
ie
wed
i
n
the
w
ork
of
[
21
,
22
]
.
Wh
il
e
[2
3
]
integ
rated
ne
ural
netw
ork
wit
h
the
f
uzzy
ap
proac
h
to
secu
re
WSN
by
m
aking
a
uthor
iz
ed
acce
ss
to
the
desire
d
sys
tem
by
exa
m
i
ning
the
net
w
ork
traf
fic
an
d
the
pr
e
vious
r
ecord.
Mor
e
ov
e
r,
the
stud
y
of
[2
4
]
op
ti
m
iz
ed
a
Ligh
t
weig
ht
and
scal
a
ble
intru
si
on
a
ppr
oac
h
us
i
ng
inte
rele
m
ent
dep
e
ndency
m
od
el
s
su
it
able
for
t
he
WSN
en
vironm
ent.
T
his
st
ud
y
is
m
a
i
nly
loo
ki
ng
to
c
om
par
e
thes
e
al
gorithm
s theo
reti
cal
ly
as wel
l as op
ti
m
iz
e
the secu
re
dete
ct
ion
alg
ori
thm
b
y ne
ural
n
et
work te
ch
niqu
e b
ase
d
on
e
nergy
consum
ption
and
pack
et
dro
ppin
g
of
the
instr
uc
te
d
nodes.
B
y
m
utatio
n
sta
ge,
the
m
os
t
e
nergy
dro
pp
e
d
node
,
a
nd
the
m
os
t
pa
cket
droppe
d
will
be
se
pa
rated
from
the
ne
twork
.
T
his
pa
per
is
organ
iz
e
d
a
s
fo
ll
ows.
Sect
io
n
1
intr
oduce
s
the
WSN
al
ong
with
rece
nt
stud
ie
s
.
Sect
i
on
2
prese
nts
the
propose
d
m
et
hod.
Sect
ion
4
pres
ents
the
sim
ula
ti
on
pa
ram
et
ers.
Sect
io
n
5
discu
sses
t
he
sim
ula
ti
on
res
ults.
Co
nclu
ding
rem
ark
s
are
decr
ib
ed
in
secio
n 5.
2.
RESEA
R
CH MET
HO
D
The
so
l
ution
pro
posed
her
e
i
s
based
on
tw
o
m
e
tric
s
to
detect
the
intru
s
ion
on
the
WSN,
w
hich
a
re
the
energy
consum
ption
an
d
the
pac
ket
deliv
ery
rati
o.
They
will
be
passed
to
the
arti
fici
a
l
neu
ral
-
im
m
u
ne
as
two
i
nputs.
It
has
t
he
respo
nsi
bili
ty
,
with
t
he
ru
le
s
data
s
et
,
w
hich
co
nt
ai
ns
the
av
era
ge
powe
r
c
ons
um
pt
ion
and
pack
et
del
ivery
rati
o
f
or
the
distrib
uted
nodes,
to
com
par
e
these
values
with
the
re
al
on
es.
U
pon
th
em
,
intru
si
on
detec
ti
on
co
uld
be
de
ci
ded
.
The
t
w
o
m
et
rics,
energy
co
nsum
ption
an
d
pack
et
l
os
s
(t
raffic
),
w
il
l
b
e
cl
assifi
ed
into
three
s
ub
-
cl
ass
ific
at
ion
s
w
hic
h
are
norm
al
,
m
or
e,
or
high
energy
co
nsu
m
pt
ion
an
d
no
rm
al,
m
od
erate,
or
high
loss
pac
kets,
f
or
th
is
purpose.
T
he
propose
d
ap
proac
h
util
iz
es
un
s
uper
vise
d
bac
k
pro
pag
at
io
n
-
ba
sed
le
arn
i
ng
si
nce
it
will
hav
e
based
on
m
e
asur
e
d
th
resho
ld
values
rat
he
r
than
pre
-
de
fine
d
values
. F
i
gure
2
dep
ic
ts t
he M
et
hodo
l
og
y
overall
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
11
, No
.
1,
Febr
uar
y
2021
:
53
6
-
54
4
538
Figure
2.
O
veral
l
m
et
ho
dolo
gy
The
propose
d
m
et
ho
d
will
be
ope
rated
in
three
se
quentia
l
phases
,
as
sho
wn
i
n
F
ig
ure
3
.
Firstl
y,
the
pha
se
cal
le
d
Gath
erin
g
data.
Dat
a
will
be
gat
he
red.
Bot
h
t
he
powe
r
c
onsu
m
pt
ion
a
nd
pack
e
t
loss
will
be
m
easure
d
a
fter
disse
m
inati
ng
data
betwee
n
tw
o
s
pecif
ic
nodes
.
The
n
data
will
pass
into
the
trai
ning
ph
a
se,
w
hich
i
s
the
sec
ond
phase.
D
ur
i
ng
this
ph
ase
,
ANN
will
be
trai
ne
d
to
ide
ntify
the
norm
al
behavio
r
of
the
pro
posed
WSN.
T
hus,
a
ny
ab
norm
al
beh
avi
or
c
ould
be
detect
e
d.
Fi
nally
,
the
res
ul
ts
of
th
e
trai
ni
n
g
phase
are
cal
culat
ed,
and
A
NN
is
re
ady
to
cl
assify
the
perf
or
m
ance
of
the
WSN.
Hen
ce
,
A
N
N
can
ide
ntify
the
I
D
S
base
d on the e
nergy c
on
s
um
ption
a
nd p
ac
ket
d
el
ive
ry r
at
io
.
Figure
3.
A
NN operati
on
phas
es
[
25
]
3.
SIMULATI
O
N
S
CEN
A
RI
OS
AND
P
A
R
AM
ET
E
RS
The
pro
pose
d
scenari
o
will
con
sist
of
120
nodes
distri
bute
d
rand
om
ly
ov
er
the
s
pecifi
ed
reg
i
on
by
us
in
g
Q
ual
net
si
m
ulator.
10
nodes
a
re
suppose
d
un
der
at
ta
ck
and
t
he
rest
w
ork
no
r
m
al
l
y.
The
pr
opos
e
d
scenari
o
is
il
lustrate
d
in
Fig
ure
4
.
Mo
reove
r
,
30
nodes
a
re
sel
ect
ed
to
be
sink
no
des
in
wh
ic
h
al
l
cal
culat
ion
s
are
ta
ke
placed
a
nd
fi
gure
out.
C
on
sta
nt
bit
rate
(CBR
)
is
co
ns
ide
re
d
a
s
a
c
onnecti
on
betw
een
sen
de
rs
a
nd
receiver
s, w
hich
a
re r
a
ndom
ly selec
te
d.
The
sel
ect
ed
pa
ram
et
ers
f
or the
sim
ula
ti
on
are
il
lustra
te
d i
n Tab
le
1
.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N:
20
88
-
8708
Secured
no
de dete
ct
ion t
ech
ni
qu
e
base
d on
ar
ti
fi
ci
al n
eu
ral
n
et
work f
or
w
ire
le
ss sen
s
or
.
..
(
Ba
ssam
H
asan
)
539
Figure
4
.
Th
e
pro
po
se
d
sim
ulati
on
sce
nar
i
o
Table
1
.
T
he
sim
ula
ti
on
para
m
et
ers
of the
pro
posed
sce
nari
o
Para
m
eter
Valu
e
Nu
m
b
e
r
o
f
no
d
es
120
Ter
r
ain
1000
-
1
0
0
Si
m
u
latio
n
ti
m
e
3
0
0
sec
Tr
af
f
ic app
licatio
n
CBR
Ite
m
to
sen
d
2
0
0
bytes
Interval:
1
sec
Mob
ility Mo
d
el
Ran
d
o
m
W
a
y
p
o
in
t
Pau
se ti
m
e
1
0
sec
W
ire
less
Ch
an
n
el Fr
eq
u
en
cy
2
.4 GHz
4.
RESU
LT
S
A
ND
DI
SCUS
S
ION
S
This
sect
ion
outl
ines
an
d
dis
cusses
the
m
ain
fin
ding
of
w
ork.
Determ
ini
ng
the
i
ntrusio
n
in
W
S
N
is
a
vital
proce
ss
ei
ther
in
tim
e
-
consum
ing
to
detect
or
in
the
accu
racy
of
de
te
ct
ion
.
I
n
order
to
detect
in
trusion
nodes
i
n
W
S
N
,
The
e
nergy
c
on
s
um
ption
a
nd
num
ber
of
pa
ckets
delive
re
d
are
t
he
tw
o
m
ai
n
crit
eria
fr
om
wh
ic
h
detect
io
n
of
i
ntr
us
io
n
will
be
dep
e
nd
ent.
T
he
res
ults
div
i
de
d
int
o
two
m
ai
n
sect
or
s
.
Fir
stl
y,
no
rm
a
l
analy
sis wh
e
n
so
m
e o
f
the sele
ct
ed
nodes
ar
e expose
d
to be un
sec
ur
e
d
by
elim
inati
ng
th
e secur
it
y
m
ec
han
ism
ov
e
r
them
.
The
sel
ect
ed
node
s
are
nodes
nu
m
ber
(
11,
14,
20,
22,
25,
30
)
.
The
sec
ond
sect
or
fo
c
us
on
i
m
ple
m
enti
ng
an
a
rtific
ia
l ne
ur
al
netw
ork s
yst
e
m
u
sing
M
ATL
AB sim
ul
at
or
.
4
.
1.
Perf
orm
an
ce
e
valua
tio
n
4
.
1.1
.
Pac
ket
deli
very
r
ati
o
PD
R
is
ho
w
m
uch
the
pac
kets
receiv
ed
to
the
pac
kets
sent.
F
ro
m
Figure
5
,
it
is
obvious
t
hat
the
unsec
ur
e
d
nodes
im
pact
the
nu
m
ber
s
o
f
receive
d
pac
ke
ts.
At
nodes
2
a
nd
22,
seve
re
dr
op
s
in
pa
ckets
receive
d,
4.2%
and
8.3%
of
pa
ckets
sen
d
ar
e
receive
d,
re
s
pecti
vely
.
Be
sides,
the
unsec
ur
e
d
co
ndit
ion
m
akes
sign
ific
a
nt
va
r
ia
ti
on
s
in
t
he
pack
et
delive
r
y
rati
o,
s
o
it
is
us
e
d
in
desi
gn
i
ng
the
AN
N
in
orde
r
to
detect
the
unsec
ur
e
d
nodes
i
n
WSN
.
F
or
e
xam
ple,
the
hi
ghest
di
ff
e
ren
ce
of
P
D
R
is
occ
urred
at
node
2
by
91
.
66%
and
the
lo
wes
t
PD
R
at
-
216
%
at
node
15.
M
or
e
over,
the
P
DR
val
u
es
at
unsecu
re
d
no
des
a
re
90.
83
%,
33.33%
,
20.
83
%,
8.3
%,
16.
66%,
a
nd
70.83
%
f
or
node
s
11
,
14,
20,
22,
25
,
30,
res
pecti
ve
ly
.
The
var
ia
ti
on
s
in
PD
R
at
thes
e
nodes
w
hen
a
ll
nodes
a
re
s
ecur
e
d
a
nd
s
om
e
are
unsec
ur
e
d
a
re
55.2%,
42.
85%,
58.
88%,
71.42%
,
33.
33
%,
an
d
10.
52
%.
T
hat
m
ea
ns
the
sec
ur
it
y
exists
the
ne
twork
pe
rform
ance,
es
pecial
ly
at
the last
three
node
s.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
11
, No
.
1,
Febr
uar
y
2021
:
53
6
-
54
4
540
Figure
5
.
P
DR
for bo
t
h node c
rite
ria
4.1.
2.
E
nerg
y
cons
u
mpti
on
Fr
om
Figure
6
,
it
is
cl
ear
that t
he
unsec
ur
e
d
nodes
im
pact
the
e
nergy
c
on
s
um
ption
of
wir
el
ess
sen
s
or
nodes
.
At
node
5
an
d
20,
t
he
hi
gh
e
st
dif
fer
e
nce
i
n
e
ne
rg
y
c
on
s
um
pti
on
with
32.59
%
an
d
the
lo
west
diff
e
re
nce
in
energy
co
ns
um
pt
ion
with
-
115.1
7%.
Be
s
ides,
the
unse
cur
e
d
co
ndit
ion
m
akes
sign
ific
a
nt
var
ia
ti
ons
in
energy
co
nsu
m
pt
ion
,
s
o
it
is
u
sed
as
a
second
in
put
in
desig
ning
the
ANN
to
detect
the
unsec
ur
e
d
nodes
i
n
WSN
.
Mor
eo
ver,
th
e
energy
c
on
s
um
pt
ion
values
at
un
sec
ured
node
s
are
12.
107
m
w,
19.88
5
m
w,
22.023
m
w,
5.415
m
w,
11
.
913
m
w,
and
9.825
m
w
fo
r
no
des
11,
14,
20,
22,
25,
30,
res
pect
ive
ly
.
The
va
riat
ion
s
in
energy
con
s
um
ption
at
these
nodes
w
hen
al
l
no
de
s
are
secur
e
d
an
d
som
e
are
un
secu
r
ed
are
-
60.
16%,
-
44.55%,
-
115.1
7%,
31.
56%,
28.
07
%,
a
nd
5.1
%
.
That
m
eans
th
e
secu
rity
existe
nce
a
ff
ect
s
ne
twor
k
perform
ance,
e
sp
eci
al
ly
at the last
thr
ee
no
de
s.
Figure
6
.
En
e
r
gy co
nsum
ption
for bo
t
h node
crite
ri
a
4.2
.
Ar
tifici
al
neural
netw
or
k based
on P
D
R
The
desi
gned
ANN
co
ns
ist
s
of
be
forem
entioned
from
th
ree
phases,
ga
therin
g
data,
t
rainin
g,
a
nd
resu
lt
s.
T
he
de
te
ct
ion
le
arn
in
g
process
is
ba
sed
on
the
res
ul
ts
of
50
%
of
the
secu
red
node
s,
wh
ic
h
is
div
ide
d
as
f
ollow
s:
20
%
trai
ning,
15
%
validat
io
n,
and
15%
te
st.
The
cl
assifi
cat
ion
process
is
base
d
on
t
he
r
esults
wh
e
n
s
om
e
of
the
nodes
a
re
unsecu
re
d
f
or
both
pac
ket
delivery
rati
os
an
d
e
ner
gy
c
on
s
um
pt
ion
va
lues.
Figure
7
s
how
s
the
ANN
m
od
el
with
20
hi
dd
e
n
la
ye
rs.
Fi
gure
8
s
hows
t
he
A
NN
perform
ance
fo
r
the
three
op
e
rati
ons,
t
rain,
validat
io
n,
a
nd
te
st
over
t
he
total
num
ber
of
ep
oc
hs
,
w
hi
ch
is
5.
I
n
ge
ner
al
,
t
he
pro
pose
d
ANN
m
od
el
per
f
or
m
s
fast
si
nce
th
e
best
va
li
dation
occ
ur
s
at
epo
ch
num
ber
2
with
m
ea
n
square
er
ror
(MSE)
equ
al
t
o 1.1×
105,
5.44, a
nd 0.066 f
or test
, v
al
idati
on
,
and t
rain A
NN
proc
esses.
The
res
ults
of
the
A
NN
m
odel
and
how
it
can
detect
the
i
ntr
us
io
n
no
de
after
trai
ni
ng
i
s
dep
ic
te
d
in
F
igure
9
based
on
P
DR
var
ia
ti
on
.
F
ro
m
the
reg
re
ssio
n
an
al
ysi
s
and
the
lowe
r
rig
ht
-
sid
e
figure
sho
ws
how
ANN
i
den
ti
fie
s
unsec
ur
e
d
nodes
wh
ic
h
a
re
far
a
way
f
r
om
the
slo
pe
li
ne
.
The
n
9
no
des
m
ay
be
unsec
ured
i
n
this case a
nd det
ect
ed
by t
he pr
opos
e
d A
NN.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N:
20
88
-
8708
Secured
no
de dete
ct
ion t
ech
ni
qu
e
base
d on
ar
ti
fi
ci
al n
eu
ral
n
et
work f
or
w
ire
le
ss sen
s
or
.
..
(
Ba
ssam
H
asan
)
541
Figure
7.
A
NN m
od
el
b
ase
d o
n
P
DR
Figure
8.
A
NN p
e
rfor
m
ance
wh
e
n PDR
val
ues
are in
put
Figure
9
.
ANN m
od
el
f
or int
r
us
io
n detec
ti
on f
r
om
PD
R
var
ia
ti
on
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
11
, No
.
1,
Febr
uar
y
2021
:
53
6
-
54
4
542
5.
AR
TIF
ICIAL
N
EU
R
AL
NETWOR
K BA
S
ED O
N
E
NERGY
CONS
U
MPTIO
N
The
dro
p
i
n
e
ne
rg
y
c
onsu
m
ption
is
use
d
to
detect
the
pro
ba
bili
ty
of
i
ns
ec
ur
it
y
in
WSN.
Fig
ur
e
1
0
(
see
a
ppen
d
ix
)
sho
ws
ANN
resu
lt
s
w
hen
e
nergy
c
on
s
um
ption
values
a
re
the
in
put
of
the
pro
posed
ANN
.
Figure 1
1
(
see
app
e
nd
ix
)
sho
ws
the
ANN p
erfor
m
ance f
or
the thr
ee
oper
at
ion
s,
trai
n,
va
li
da
ti
on
, a
nd test
o
ve
r
the total
n
um
ber
o
f
ep
oc
hs
, which
is 5.
In g
e
ner
al
, th
e pr
opos
e
d
ANN
m
od
el
p
er
form
s sl
ow
e
r
than
P
D
R si
nce
the
best
valida
ti
on
occ
ur
s
at
epo
c
h
num
ber
5
with
m
ean
s
q
ua
re
erro
r
(m
se)
eq
ual
to
0.78
×
10
-
5,
0.5
3
×
10
-
5,
and
0.48
×
10
-
5
fo
r
te
st,
vali
dation,
an
d
tr
ai
n
ANN
pro
c
esses.
H
ow
e
ve
r,
com
par
in
g
to
A
NN
-
ba
sed
-
PD
R,
ANN
-
base
d
-
e
ne
rg
y
c
onsu
m
ption
s
hows
a
be
tt
er
perform
ance.
The
re
gr
es
s
ion
a
naly
sis
of
ANN
-
base
d
-
e
nergy
consum
ption
wh
ic
h
is dep
ic
t
ed
in Figure 1
2
(
see appen
d
ix
)
sh
ows ho
w
A
NN
detect
s the v
ariat
ion
wh
e
n
so
m
e
nodes
are u
ns
e
cur
e
d. The
lo
w
er
-
rig
ht side
f
i
gure ill
us
trat
es
these
var
ia
ti
on
am
ou
nt co
m
pa
rin
g
to
f
it
values.
6.
CONCL
US
I
O
N
The
W
S
N
is
the
m
os
t
po
pu
l
ar
netw
ork
in
the
la
st
recent
ye
ars
as
it
can
m
easur
e
the
env
i
ronm
ent
al
conditi
ons
an
d
sen
d
them
to
process
pur
pos
es.
V
a
rio
us
at
t
acks
c
ou
l
d
be
su
bject
s
a
gain
st
W
S
N
s
a
nd
cause
dam
age
ei
ther
in
the
sta
bili
ty
of
c
omm
un
ic
at
ion
o
r
in
t
he
de
structio
n
of
t
he
sensiti
ve
data.
So,
the d
em
and
s o
f
intru
si
on
de
te
ct
ion
-
base
d
e
ne
rg
y
-
e
ff
ic
ie
nt
te
chn
i
qu
e
s
rise
dr
am
at
ic
ally
as
the
netw
ork
dep
l
oym
ent
beco
m
e
s
wide
a
nd
com
plica
te
d.
T
his
pap
e
r
int
rod
uc
ed
a
n
opti
m
iz
e
d
en
er
gy
-
based
intru
si
on
dete
ct
ion
te
ch
niqu
e
us
in
g
a
neural
net
w
ork
by
Ma
tl
ab
si
m
ulator.
T
he
resu
l
ts
s
how
in
the
case
of
so
m
e
nodes
tha
t
are
sig
nif
ic
ant
insecu
re
var
ia
t
ion
i
n
value
s
a
re
detect
ed
,
w
hich
m
ean
s
un
secur
e
d
no
des
aff
ect
t
he
perf
or
m
ance
of
t
he
W
S
N
.
The
sec
ond
se
ssion
of
the
re
su
lt
il
lustrate
s
the
re
gr
e
ssio
n
analy
sis
for
t
he
pro
posed
A
NN
-
ba
sed
,
bo
t
h
P
DR
a
nd
e
nergy
co
ns
um
ption.
Ov
erall
the
te
chn
i
qu
e
pro
du
ce
s
good
re
su
lt
s
f
or
bo
t
h
scena
rios.
It
can
be
c
on
c
lud
e
d
that
the
A
NN
base
d
-
P
DR
i
s
faster
th
an
ANN
-
base
d
-
e
nergy
co
nsum
ption,
but
bo
t
h
of
them
detect
s
the v
a
riat
ions
of the
value
.
APP
E
ND
I
X
Figure
1
0
.
A
N
N
m
od
el
-
based o
n
e
nergy c
onsu
m
ption
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N:
20
88
-
8708
Secured
no
de dete
ct
ion t
ech
ni
qu
e
base
d on
ar
ti
fi
ci
al n
eu
ral
n
et
work f
or
w
ire
le
ss sen
s
or
.
..
(
Ba
ssam
H
asan
)
543
Figure
1
1
.
A
N
N per
form
ance when ene
r
gy
consu
m
ption
va
lues are
in
pu
t
Figure
1
2
. A
N
N
m
od
el
for i
nt
ru
sio
n detec
ti
on fro
m
e
nergy c
on
s
umpti
on v
a
riat
ion
REFERE
NCE
S
[1]
M.
Carl
os
-
Manc
il
la,
e
t
a
l.
,
“
W
ire
le
ss
sensor
net
works
form
at
ion:
Approac
hes
and
te
chn
ique
s,
”
Jou
rnal
of
Sensors
,
vol.
2016
,
pp
.
1
-
18,
2016
.
[2]
S.
Alani
,
et
a
l.
,
“
A
new
ene
rg
y
consum
pti
on
te
c
hnique
for
m
obil
e
Ad
-
Hoc
net
works
,
”
Int
ernati
o
nal
J
ournal
of
El
e
ct
r
ic
a
l
and
C
omput
er
Eng
in
e
ering
(
IJE
CE)
, v
ol.
9
,
no
.
5
,
pp
.
4
147
-
4153,
2019
.
[3]
J.
Alamri,
et
a
l.
,
“
Perform
anc
e
Eva
luation
of
Tw
o
Mobile
Ad
-
hoc
Network
Routi
ng
Protocol
s :
Ad
-
hoc
o
n
-
Dem
and
Distanc
e
Vec
tor,
Dy
n
amic
Sourc
e
Routi
ng,
”
Int
e
rnational
J
ournal
of
Adv
anc
ed
Sci
en
ce
and
Tec
hnol
ogy
,
vol
.
29,
no.
5
,
pp
.
9915
-
9920,
2020
.
[4]
S.
A.
Rashid,
e
t
al
.
,
“
Predic
t
ion
base
d
eff
i
ci
en
t
m
ult
i
-
hop
c
luste
ri
ng
appr
oac
h
wit
h
ada
pti
v
e
re
lay
node
select
ion
f
o
r
VA
NET,
”
J
ournal
of
Comm
un
ications
,
vo
l. 15, n
o.
4
,
pp
.
332
-
34
4,
2020
.
[5]
S.
Alani
,
et
a
l.
,
“
A
study
rev
ie
w
on
m
obil
e
ad
-
h
oc
net
work:
Cha
rac
t
eri
sti
cs,
app
l
ic
a
ti
ons,
ch
al
l
en
ges
and
routi
ng
protoc
ols
class
ifi
cation,
”
Int
erna
ti
onal
J
ournal
o
f
Adv
anc
ed
Sc
i
e
nce
and
Technol
ogy
,
vol
.
28,
no
.
1,
pp.
394
-
405
,
2019
.
[6]
A.
Djimli,
e
t
a
l.
,
“
Ene
rg
y
-
eff
icient
MA
C
prot
ocol
s
for
wir
ele
ss
sensor
net
works
:
a
surve
y
,
”
TEL
KOMNIKA
(
Tele
communic
ati
on
Comput
ing
,
El
e
ct
ron
ic
s and
Control
)
,
vol. 17
,
no
.
5
,
p
p
.
2301
-
2312
,
2019
.
[7]
A.
M.
Fahad,
et
al
.
,
“
Dete
c
ti
on
of
Bla
ck
Hol
e
Atta
cks
in
Mobi
le
Ad
Hoc
Net
works
via
HS
A
-
CBDS
Method,
”
in
Int
ernati
onal
Confe
renc
e
on
I
nte
lligent Comp
uti
ng
and
Optim
izati
on,
,
pp
.
46
-
55
,
201
8
.
[8]
S.
Su
and
S.
Wa
ng,
“
A
Sim
ple
Monitori
ng
Net
work
Sy
stem
of
W
ire
le
ss
Senso
r
Network,
”
Bul
letin
of
El
e
ct
r
ica
l
Eng
ineering
and
Informatic
s
(
BEEI)
,
vol. 1, no. 4, pp. 251
-
254,
20
12.
[9]
M.
A.
S
aa
d,
et
a
l.
,
“
Perform
anc
e
Eva
luation
Im
prove
m
ent
of
Energ
y
Consum
pti
on
in
Ad
-
Hoc
W
i
rel
ess
Network
,
”
Int.
J. A
d
v. Sc
i
.
Technol
.
,
vo
l. 29
,
no
.
3
,
pp
.
4128
-
4137,
2020
.
[10]
S.
A.
Hussei
n
and
D.
P.
Dahnil
,
“
A
New
Hy
bri
d
Te
chni
qu
e
to
Im
prove
the
Path
Sele
ct
ion
in
R
educ
ing
En
erg
y
Consum
pti
on
in
Mobile
AD
-
HO
C
Networks,
”
Int
ernati
onal
Jo
urnal
of
Appl
ied
Eng
ine
ering
Res
earc
h
,
vo
l.
12,
no.
3
,
pp
.
277
-
2
82,
2017
.
[11]
A.
M.
Fah
ad,
et
al.
,
“
Ns
2
base
d
per
form
ance
co
m
par
ison
stud
y
bet
wee
n
dsr
and
aodv
pro
toc
ols,
”
Int
ernati
ona
l
J
ournal
of
Ad
v
a
nce
d
Tr
ends
in
Comput
er
Sci
en
ce
and
Eng
ine
er
ing
,
vo
l. 8, no. 1.4,
pp
.
379
-
393
,
2019.
[12]
M.
P.
Beh
am
a
nd
S.
M.
M.
R
oom
i,
“
A
rev
iew
of
face
r
ec
og
nit
ion
m
et
hods
,
”
Int
ernati
onal
J
ournal
of
Pat
t
e
r
n
Re
cogn
it
ion
and
Arti
f
icial
In
tell
i
genc
e
,
vo
l. 27, n
o.
4
,
pp
.
135600
5_1
-
1356005_35,
2013
.
[13]
O.
S.
Al
-
hee
t
y
,
et
al.
,
“
A
comprehe
nsive
surve
y
:
Bene
fi
ts,
Servi
ce
s,
Re
ce
nt
wor
ks,
Chal
le
ng
es,
Secur
ity
and
Us
e
ca
ses for
SD
N
-
VA
NET,
”
I
EE
E
Ac
c
ess
,
vol
.
8
,
p
p.
91028
-
91047
,
2020.
[14]
Y.
C.
W
ong,
e
t
al.
,
“
Low
po
wer
wake
-
up
r
e
ce
iv
er
base
d
on
ult
rasound
co
m
m
unic
at
ion
fo
r
wire
le
ss
sens
o
r
net
work,
”
Bu
ll
e
t
in
of
Elec
tr
ic
al
Eng
ineering
and
Informatic
s
,
vol
.
9
,
no
.
1
,
pp
.
21
-
29,
2020
.
[15]
M.
A.
Saad,
e
t
al
.
,
“
Spect
rum
sensing
and
ene
rg
y
d
etec
t
ion
in
cogni
ti
v
e
n
et
works
,
”
In
don
es
ian
J
ournal
of
El
e
ct
r
ic
a
l
Eng
in
ee
ring a
nd
Comput
er
Sc
i
ence
(
IJ
EE
CS)
,
vol
.
17
,
no.
1
,
pp
.
465
-
4
72,
20
20
.
[16]
M.
Pradha
n,
et
a
l.
,
“
Intrusion
de
t
ec
t
ion
s
y
stem
(I
DS
)
and
the
ir
t
ypes,
”
Se
curing
t
he
Inte
rnet
o
f
Things
:
Conce
pts
,
Tools
,
and
App
licati
ons
,
20
20
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
11
, No
.
1,
Febr
uar
y
2021
:
53
6
-
54
4
544
[17]
S.
La
q
ti
b
,
e
t
al
.
,
“
A
technical
r
evi
ew
and
compara
t
ive
anal
y
s
i
s
of
m
ac
hin
e
l
e
arn
ing
te
chn
iqu
es
for
in
trusion
det
e
ct
ion
s
y
s
te
m
s
in
MA
NET,
”
Int
ernati
onal
J
o
urnal
of
El
e
ct
r
i
cal
and
Comput
er
Eng
ine
ering
(
IJE
CE)
,
vol.
10
,
no.
3
,
pp
.
2701
-
2709,
2020
.
[18]
R.
J
y
o
thi
a
nd
N.
G.
Choll
i,
“
An
eff
ic
i
ent
ap
proa
ch
for
se
cu
red
comm
unic
ation
in
wire
le
ss
sensor
net
works
,
”
Int
ernati
onal
J
o
urnal
of
El
e
ct
r
i
c
al
and
Comput
er
Eng
in
ee
ring
(
IJE
CE)
,
vol
.
10
,
n
o.
2
,
pp
.
1641
-
1
647,
2020
.
[19]
S.
Bit
am,
et
al
.
,
“
Bio
-
inspire
d
c
y
b
erse
cur
ity
fo
r
wire
le
ss
sensor
net
works
,
”
IE
EE
Comm
un
ic
at
ions
Mag
azine
,
vol.
54
,
no
.
6
,
pp
.
68
-
74
,
2016
.
[20]
A.
M.
Fahad,
A.
A.
Ah
m
ed,
A
.
H.
Alghusham
i,
and
S.
Alani
,
“
D
et
e
ct
ion
of
Blac
k
Hole
Atta
cks
in
Mobile
Ad
H
o
c
Networks
via
HS
A
-
CBD
S
Method
,
”
in
Springer
Nature
Swit
zerl
and,
Springer
Inte
rnational
Publishing
,
vol.
866
,
pp.
46
–
55
,
2019
.
[21]
A.
Mahboub,
e
t
al
.
,
“
An
ene
rg
y
-
e
fficie
nt
cl
ust
eri
ng
proto
col
u
sing
fuz
z
y
log
ic
and
net
work
s
egmenta
t
ion
for
het
ero
g
ene
ous
W
SN
,
”
Int
ernati
onal
J
ournal
of
El
e
ct
r
ic
a
l
and
Comput
er
Eng
i
nee
ring
(
IJE
CE
)
,
vol.
9,
no.
5,
pp.
4192
-
4203
,
2019.
[22]
A.
S.
Al
-
ahmad,
H.
Kahta
n,
and
S.
Alani
,
“
Inte
l
li
ge
nt
Com
puti
n
g
&a
m
p;
Optim
iz
a
ti
on
,
Int
el
l
.
C
om
put.
&
amp;
Optim.,
vo
l. 866
,
no
.
Novem
ber
2018,
pp
.
267
–
2
76,
2019
.
[23]
M.
R.
Hos
sain,
et
al
.
,
“
Network
flow
opti
m
iz
a
ti
on
b
y
Gen
et
i
c
Algorit
hm
and
loa
d
flow
ana
l
y
sis
b
y
Newton
Raphson
m
et
hod
in
power
s
y
st
em,
”
2nd
In
t
ern
ati
onal
Con
f
ere
nce
on
.
E
lectr
ical
Eng
in
ee
ring
and
Inf
orm
ati
on
Comm
un
ic
ati
on
Technol
ogy
(
iCE
Ei
CT
2015
)
,
pp.
1
-
5,
2015
.
[24]
N.
A.
Alra
je
h,
e
t
al
.
,
“
Artifi
cial
neur
al
ne
twork
base
d
det
e
ct
ion
of
ene
rg
y
exha
u
stion
at
tacks
in
wire
le
ss
sensor
net
works
ca
p
able
of
en
erg
y
har
vesti
ng,
”
Ad
-
Ho
c
and
Sens
or
W
irel
ess
Net
works
,
vol
.
22,
no
.
1
-
2,
pp.
109
-
133
,
2014.
[25]
O.
Avci,
et
a
l.
,
“
Convolut
iona
l
neur
al
ne
tworks
for
rea
l
-
ti
m
e
an
d
wire
le
ss
damage
det
e
ct
ion
,
”
D
ynamic
s
of
Civ
i
l
Struct
ures
,
vo
l.
2,
pp
.
129
-
136
,
2020.
BIOGR
AP
HI
ES OF
A
UTH
ORS
Bass
am
Hasan
rec
e
ive
d
B.
S
de
gre
e
in
Com
put
er
Engi
ne
eri
ng
Te
chno
log
y
f
ro
m
AL
-
MAAREF
Univer
sit
y
Col
l
ege
(IRAQ
)
2
013
-
2014.
He
Master
of
En
gine
er
ing
(Com
m
unic
at
ion
and
Com
pute
r)
form
U
NIV
ERSITI
KEBAN
G
SAAN
M
ALAY
SI
A,
The
Nati
on
al
Univer
sit
y
o
f
Malay
s
ia
in
201
8.
He
is
cur
r
ent
l
y
a
Ph.D.
studen
t
at
Engi
ne
eri
ng
of
comm
unic
at
i
on
in
Univer
sit
i
Te
nag
a
Nasion
al
(Ma
lay
sia)
.
His
rese
ar
ch
area
in
cl
udes
W
SN
,
VA
N
ET
and
wire
l
ess
comm
unic
at
ions
.
Samee
r
Alani
was
born
in
Ir
aq
in
1989
.
He
r
ecei
ved
a
B
.
S.
d
eg
ree
in
compute
r
engi
ne
eri
ng
a
nd
M.Sc.
degr
ee
in
wire
l
ess
comm
unic
at
ion
and
Com
pute
r
net
working
te
chno
log
y
f
rom
The
Nati
on
al
Un
ive
rsit
y
of
Malay
sia
(UK
M)
in
2017.
He
is
cur
re
ntly
pursuing
th
e
Ph.D.
degr
e
e
i
n
wire
le
ss
comm
u
nic
a
ti
on
and
ne
tworking.
His
rese
arc
h
in
te
r
est
s
inc
lude
anten
na
appl
i
ca
t
ions,
wire
le
ss
comm
unic
a
ti
on
and
n
etw
orking
technol
og
y
.
Mohamme
d
Ayad
Saad
rec
ei
ve
d
his
Bs
Degre
e
in
Com
pute
r
and
Com
m
unic
at
ion
(2011
-
2015)
in
Ira
q.
He
ea
rne
d
his
Master
'
s
Degre
e
in
En
gine
er
Te
l
ec
om
m
unic
at
ion
and
Com
pute
r
from
Univer
sit
y
Keb
angsa
an
Ma
lay
s
ia
(UK
M).
He
is
cur
r
entl
y
p
ursuing
his
Ph.
D.
in
Univ
ers
i
t
y
Keba
ngsaa
n
Malay
si
a.
His
r
ese
arc
h
ar
ea
i
ncl
udes
informati
on
technolog
y
and
wir
eless
comm
unic
at
ion,
VA
NET
and
W
SN
.
Evaluation Warning : The document was created with Spire.PDF for Python.