Indonesi
an
Journa
l
of El
ect
ri
cal Engineer
ing
an
d
Comp
ut
er
Scie
nce
Vo
l.
1
3
,
No.
3
,
Ma
rch
201
9
, p
p.
892
~
9
01
IS
S
N: 25
02
-
4752, DO
I: 10
.11
591/ijeecs
.v1
3
.i
3
.pp
892
-
9
01
892
Journ
al h
om
e
page
:
http:
//
ia
es
core.c
om/j
ourn
als/i
ndex.
ph
p/ij
eecs
Ring os
cil
lator p
hysically
unclona
ble funct
ion usin
g sequen
tial
ring os
cil
lator p
airs fo
r more
chall
enge
-
re
s
ponse
-
p
airs
Ju
li
us
Han Lo
ong Te
o,
N
oo
r
A
li
a N
or
H
ashi
m, A
z
rul G
ha
z
ali,
F
az
rena Az
le
e Ha
mi
d
Dep
a
rtm
ent o
f El
ect
rical
an
d
Com
m
un
ic
at
io
n
E
ngineeri
ng
,
Univer
siti
Te
n
ag
a
Nasiona
l, Mala
y
sia
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
O
ct
25
, 201
8
Re
vised
Dec
1
6
, 2
018
Accepte
d Dec
3
0
, 201
8
The
ring
oscilla
tor
ph
y
s
ic
a
lly
u
ncl
onab
le
fun
ct
i
on
(ROP
UF
)
is
one
of
th
e
seve
ral
t
y
pes
of
PU
F
tha
t
has
gre
at
pot
ent
i
al
to
be
used
for
sec
uri
t
y
purposes.
An
al
te
rna
ti
ve
ROP
U
F
design
is
proposed
w
it
h
two
m
aj
or
diffe
ren
ce
s.
Firstl
y
,
the
m
emristor
is
inc
lude
d
in
the
ring
oscil
lators
as
it
i
s
cl
ai
m
ed
to
produc
e
a
m
ore
ran
dom
oscil
lati
on
fre
quency
.
Ot
her
rea
sons
ar
e
i
ts
m
emory
-
li
ke
prope
rt
ie
s
a
nd
var
i
abl
e
m
emristance,
relati
v
e
compati
b
il
i
t
y
w
it
h
CMO
S,
and
sm
al
l
si
ze
.
Secondly
,
a
d
iff
ere
nt
m
et
hod
of
gene
r
at
ing
the
response
is
implemente
d
wh
ere
b
y
a
seque
n
c
e
of
sel
ection
of ring osc
illat
or
pa
irs
are
us
ed
to
gen
era
t
e
a
m
ult
iple
bi
t
r
esponse,
r
at
h
er
th
an
usi
ng
onl
y
on
e
ring
oscil
l
at
o
r
pai
r
to
gen
era
t
e
a
single
bit
r
esponse.
Thi
s
m
et
h
od
signifi
ca
n
tly
expa
nds
the
set
of
cha
lleng
e
-
response
pai
rs
.
The
p
roposed
m
emristor
-
base
d
ROP
U
F
show
s
48.
57%,
51.
43%,
and
51
.
43%
for
un
iqu
ene
ss
,
uniformit
y
,
and
bit
-
al
i
asing,
r
espe
c
tivel
y
.
Also,
m
od
el
li
ng
b
y
suppor
t
ve
ct
or
m
a
chi
ne
(SV
M)
on
the
proposed
m
emristor
-
base
d
ROP
U
F
onl
y
show
s
61.
95
%
ac
cur
acy
,
the
reb
y
indica
ti
n
g
strong
resist
an
ce
against
SV
M.
Ke
yw
or
ds:
Mem
ristor
Ph
ysi
cal
ly
u
ncl
on
a
ble
functi
on
Ri
ng
os
ci
ll
at
or
Copyright
©
201
9
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed.
Corres
pond
in
g
Aut
h
or
:
Ju
li
us
Ha
n
L
oo
ng Te
o
,
Dep
a
rtm
ent o
f
Ele
ct
rical
an
d
Com
m
un
ic
at
io
n
E
ngineeri
ng,
Un
i
ver
sit
i Te
na
ga Nasi
onal
,
Jal
an
I
KR
AM
-
UNITE
N,
43
001 Ka
j
a
ng, S
el
ango
r,
Mal
ay
sia
.
Em
a
il
: ju
li
us
te
o@
li
ve
.co
m
.
m
y
1.
INTROD
U
CTION
1.1
.
P
hy
sic
ally
Un
cl
ona
ble
Functi
on
The
P
hysic
al
ly
Un
cl
onable
F
un
ct
io
n
(PUF
)
is
a
ci
rcu
it
th
at
has
been
re
garde
d
as
a
n
a
lt
ern
at
ive
to
current
f
or
sec
ur
it
y
m
easur
es
[1
-
3].
P
UF
s
a
r
e
sai
d
to
ha
ve
it
s
own
uniq
ue
and
i
nh
e
ren
t
prop
e
rty
,
analo
gous
t
o
hu
m
ans
ha
ving
a
uniq
ue
set
of
fin
ge
rprints,
from
wh
ic
h
se
cur
it
y
keys
ca
n
be
extracte
d
and
us
e
d
for
s
ecur
it
y
pur
po
ses
,
s
uc
h
as
the
ide
ntif
ic
at
ion
a
nd
au
thentic
at
ion
of
an
el
ect
ronic
dev
ic
e.
T
his
s
o
-
cal
le
d
uniq
ue
an
d
inh
e
ren
t
pro
pe
rty
of
the
P
UF
is
due
t
o
ra
ndom
and
unco
ntr
ollabl
e
va
riat
ion
s
in
the
m
anu
fa
ct
ur
in
g
process
[4,
5].
Of
te
n,
t
hese
var
ia
ti
ons
are
unwa
nted,
pa
rtic
ularly
in
the
fabrica
ti
on
of
chips
,
but
the
PU
F
s
exp
l
oit
these
va
riat
ion
s
t
o
ha
ve
a
ci
rcu
it
pro
per
ty
that
is
un
iqu
e
from
oth
e
r
ci
rc
uits
of
the
sam
e
ty
pe.
W
it
h
the
secur
it
y key
be
ing
un
i
que to a
nd
i
nh
e
re
nt w
it
hin
the
P
UF
, t
he
PU
F ca
n be
use
d
as a
n
al
te
r
na
ti
ve
to the s
t
ori
ng
of
t
hese
keys
i
n
nonvolati
le
m
e
m
or
y
that
can
be
easi
ly
at
ta
cked
[
6].
Als
o,
ta
m
per
in
g
th
e
PU
F
will
ve
r
y
li
kely
dam
age th
e cir
cuit an
d
it
s sec
ur
it
y key
[7
]
,
re
nd
e
rin
g
i
nv
asi
ve
at
ta
cks
futi
le
.
The
in
put
and
ou
t
pu
t
of
a
P
U
F
are
res
pecti
ve
ly
te
r
m
ed
as
‘ch
al
le
nge
’
an
d
‘
respo
ns
e’
.
A
chall
en
ge
sent
to
a
PU
F
r
esults
in
a
respon
s
e
from
the
PU
F
,
w
hich
is
te
rm
ed
as
challen
ge
-
re
spo
ns
e
pair
(CRP
).
Th
e
set
of
CR
Ps
is
co
ns
ide
red
as
the
un
iq
ue
pro
pe
r
ty
that
disti
ng
ui
sh
es
it
sel
f
fr
om
oth
er
PU
F
s
[1
-
3].
T
o
date,
there
are
a
se
ve
ral
P
UF
ty
pes
bei
ng
r
esearc
he
d
li
ke
the
bista
ble
rin
g
P
UF
[
8],
the
a
rb
it
er
P
UF
[1
-
3],
the
SRAM
PU
F
[9
-
11]
,
cr
os
s
bar
a
rr
ay
P
UF
[
12
-
15]
,
an
d
the
ri
ng
os
ci
ll
at
or
PUF
(R
O
PU
F
)
[
5,
16
-
19]
,
w
hich
is
th
e
ty
pe
of PUF ta
ken into disc
us
sio
n for this
r
esea
rc
h.
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Ring oscil
lat
or p
hysicall
y
un
c
lonab
le
f
unct
io
n usin
g
se
quen
ti
al ring o
sci
ll
ato
r…
(
J
ulius
Han Lo
ong Te
o
)
893
1.2.
Mem
ory
Resist
or
Me
anwhil
e,
th
e
m
e
m
r
ist
or
,
s
hort
for
“m
e
m
or
y
resist
or”,
is
sai
d
to
be
the
f
ourth
f
undam
ental
pas
siv
e
ci
rcu
it
el
e
m
ent;
the
first
three
bein
g
the
resi
stor,
ca
pacit
or,
and
in
duct
or
.
The
idea
of
th
e
m
e
m
ristor
fall
s
on
on
e
of
the
si
x
po
ssi
ble
pair
wise
relat
ion
s
hip
s
am
on
g
four
f
undam
ental
ci
rcu
it
var
ia
bl
es,
nam
ely
cur
re
nt
i,
vo
lt
age
v,
c
ha
r
ge
q,
an
d
flu
x
li
nk
a
ge
Φ
.
C
hua,
in
19
71,
cl
ai
m
ed
that
t
he
re
la
ti
on
sh
i
p
bet
ween
c
harge
and
fl
ux
li
nk
age
t
o
be
m
e
m
rista
nce
[2
0
-
22
]
.
Ch
ua
m
ade
this
po
st
ulati
on
f
or
the
sake
of
com
plete
ness
beca
use
,
at
the
tim
e,
the
relat
i
on
s
hi
p
betwee
n
cha
rg
e
a
nd
fl
ux
li
nkage w
as
the
on
ly
pairwi
se
relat
ion
sh
i
p
le
ft
that
was
no
t
ye
t
firm
l
y unders
t
ood. T
hese
pairw
ise
relat
ion
s
hip
s
are vis
ualiz
ed
in
Fig
ure
1.
Figure
1
.
Map
ping
of the
pair
wise r
el
at
io
nsh
ip of circ
uit
va
riables
Howe
ver,
the
act
ual
ph
ysi
cal
m
e
m
ristor
was
on
ly
disco
vere
d
in
2008
by
a
te
a
m
of
researc
her
s
in H
P
Labs
[
23,
24
]
.
Their
m
e
m
rist
or
str
uctu
re
si
m
pl
y
con
sist
s
of
on
ly
tw
o
ti
t
anium
dio
xid
e
,
TiO
2
la
ye
rs:
undo
ped
and
dope
d
with
oxyge
n
vaca
ncies,
denoted
as
TiO
2
-
x
w
hich
are
san
dwic
hed
betw
een
plati
nu
m
el
ectr
ode
s
.
The
str
uctu
re
of
t
he
m
e
m
rist
or
is
s
how
n
i
n
Fig
ur
e
2,
i
n
wh
ic
h
the
le
ngth
of
the
dope
d
la
ye
r
is
la
be
ll
ed
w
,
wh
e
reas
t
he
le
ng
t
h
of
t
he
m
e
m
ris
tor
is
la
bel
le
d
D
.
A
s
a
sc
hem
at
ic
,
the
m
e
m
ris
tor
is
re
pre
sented
a
s
tw
o
s
eries
-
connecte
d
vari
able
resist
or
s
,
as
sh
own
in
Figu
re
3.
T
he
ex
pr
essi
on
fo
r
the
eq
ui
valent
m
e
m
r
is
ta
nce,
M
is
sh
own
in
(1
)
,
wh
e
re
by
m
e
m
rista
nce
is
si
m
ply
resist
ance
but
on
ly
sp
eci
fical
ly
fo
r
m
e
m
ristors.
M
ON
and
M
OFF
are
the
m
e
m
rista
nce
a
t
ON
sta
te
or
low
re
sist
ance
sta
te
(LRS)
an
d
OF
F
sta
te
or
hig
h
resist
a
nc
e
sta
te
(LRS),
resp
ect
i
vely
. S
im
ply pu
t, M
ON
an
d
M
O
FF
are the
m
axi
m
u
m
an
d
m
inim
u
m
r
esi
sta
nces,
re
sp
ect
ivel
y.
Figure
2
.
Str
uc
ture of
the
HP
m
e
m
ristor
Figure
3
.
Ci
rcui
t equ
ivale
nt
of the
HP
m
e
m
ri
stor
The
m
e
m
r
ist
a
nce
cha
ng
e
s
over
ti
m
e
with
resp
ect
to
the
po
la
rity
of
the
excit
at
ion
sig
nal,
unti
l
it
appr
oach
es
M
ON
or
M
OFF
.
Furtherm
or
e,
the
m
e
m
rista
nce
in
retai
ned
on
c
e
the
e
xcita
ti
on
si
gn
al
it
is
re
m
ov
ed.
This
m
e
m
or
y
-
li
ke
cha
racteri
st
ic
is
wh
e
re
th
e
nam
e
m
e
m
or
y
resist
or
com
es
from
,
and
it
is
the
m
os
t
prom
i
nen
t
char
act
e
risti
c that disti
ng
uishe
s it
sel
f
f
ro
m
ot
her
circ
uit c
om
po
nen
ts.
The
m
e
m
or
y
and
var
yi
ng
m
e
m
ris
ta
nce
are
du
e
to
the
m
ov
em
ent
of
the
ox
y
gen
vacan
c
ie
s
that
ar
e
fou
nd
in
t
he
dope
d
ti
ta
ni
um
dioxide
la
ye
r
,
TiO
2
-
x
.
T
he
oxygen
vaca
ncies
m
ov
e,
de
pe
nd
ing
on
th
e
po
la
rity
of
the
excit
at
ion
sign
al
,
a
nd
c
ha
ng
e
t
he
thick
ness
of
the
doped
ti
ta
niu
m
dio
xid
e
la
ye
r
,
w,
w
hic
h
co
nse
qu
e
ntly
change
the
ov
erall
m
e
m
r
ist
a
nce
as
ex
press
ed
in
(
1)
ea
rlie
r.
When
the
ex
ci
ta
ti
on
is
rem
ov
e
d,
w
is
no
longer
aff
ect
ed
a
nd is
un
c
ha
ng
e
d, an
d hen
ce
, th
e
m
e
m
rista
nce is ret
ai
ned
[23
-
27]
.
The
c
hange
in
m
e
m
rista
nce
with
res
pect
to
the
excit
at
ion
is
nonlinear
.
I
n
fact,
t
he
I
-
V
plo
t
of
th
e
m
e
m
ristor
exhi
bits
a
hyste
res
is
loop
pi
nch
e
d
at
the
or
igi
n
wh
e
n
e
xc
it
ed
by
a
bipolar
pe
rio
dic
signa
l,
li
ke
a
sinu
s
oid,
as
s
how
n
in
Fig
ure
4.
The
pinc
hed
hyste
resis
loop
reduces
in
area
with
increasin
g
e
xc
it
at
ion
fr
e
qu
e
ncy,
unti
l i
t reduces i
nto
a
strai
ght l
in
e [
23
-
27]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
1
3
, N
o.
3
,
M
a
r
c
h
2019
:
8
9
2
–
9
0
1
894
Figure
4
.
I
-
V
pl
ot of th
e
m
e
mristor at
v
a
rio
us fre
que
ncies
The
rea
sons
f
or
i
ncorpo
rati
ng
t
he
m
e
m
ri
stor
in
t
he
R
OPUF
are
the
m
e
m
or
y
-
li
ke
pro
per
ty
an
d
var
ia
ble
m
e
m
r
esi
sta
nce.
Als
o,
the
m
e
m
risto
r
m
anu
fact
ur
i
ng
te
chnolo
gy
is
sai
d
to
be
re
la
ti
vely
co
m
patible
with
the
m
od
e
rn
CM
O
S
fa
bri
cat
ion
sta
ndar
ds
[
28]
.
In
a
ddit
ion
,
t
he
m
e
m
ris
tor
-
base
d
PU
Fs
a
re
sai
d
to
be
m
or
e
resist
ant
to
m
od
el
li
ng
at
ta
cks
than
purely
CM
OS
-
base
d
PUFs
be
cause
m
e
m
ris
tors
are
bid
ire
ct
ion
al
dev
ic
es
as
c
om
par
ed
to
MOSFE
Ts
w
hic
h
are
un
i
direc
ti
on
al
[29].
F
ur
t
her
m
or
e,
th
e
m
e
m
ristor
l
eng
t
h
is
ty
pical
ly
in
the
te
ns
of
na
nom
et
ers,
w
hich
is
m
uch
sm
al
l
er
tha
n
m
os
t
CM
OS
com
ponen
ts
a
nd
ca
n
reduce
ROPUF
ci
rcu
i
t
area.
Fo
r
th
ese
reasons,
r
esearch
ef
f
or
ts
hav
e
bee
n
m
ade
to
inclu
de
the
m
e
m
r
ist
or
into
diff
e
re
nt ty
pes of
PUFs t
o
e
nhance it
s
perfor
m
ance as a ha
r
dw
a
re s
ec
uri
ty
d
evice
[
28
-
34]
.
1.3
.
Ring Os
ci
ll
at
or Ph
ys
ic
ally
Un
cl
ona
b
le
Fun
cti
on
The
cl
assic
al
r
ing
os
ci
ll
at
or
Ph
ysi
cal
ly
Uncl
on
a
ble
F
un
ct
ion
(RO
PUF)
was
int
rod
uce
d
by
S
uh
a
nd
Dev
a
das
[5
]
,
a
s
show
n
in
Fi
gure
5.
T
he
id
ea
of
t
he
ROP
UF
is
the
c
omparis
on
of
the
fr
e
qu
e
ncy
bet
ween
a
sel
ect
ed
pair
of
identic
al
RO
s
to
gen
e
rate
a
response
bit.
Du
e
to
rand
om
and
unco
ntr
ollable
m
anu
fa
ct
ur
in
g
process
var
ia
ti
on
s
,
the
re
is
a
discre
pan
cy
in
the
delay
s
of
e
ach
in
ver
ti
ng
unit
in
the
RO,
conseq
ue
ntly
c
ausi
ng
discre
pan
cy
i
n t
he
osc
il
la
ti
ng
fr
e
qu
e
ncy am
on
g i
de
ntica
l R
Os.
Gen
e
rall
y,
the
ROPUF
co
nsi
sts
of
k
R
Os,
tw
o
k
-
to
-
1
m
ulti
plexer
s
,
two
c
ounte
r
s,
an
d
on
e
com
par
at
or
.
T
he
ope
rati
on
of
the
ROP
UF
c
an
be
sim
p
li
fie
d
as
fo
ll
ows.
T
he
c
halle
ng
e
t
o
the
P
UF
is
ta
ke
n
as
the
sel
ect
inp
ut
s
to
bo
t
h
k
-
to
-
1
m
ulti
plexers,
w
hich
sel
ect
s
a
pair
of
R
Os
to
be
direc
te
d
to
the
res
pe
ct
ive
counters
.
T
he
counters
c
ount
the
pulse
s
within
a
ce
rtai
n
durati
on,
a
nd
fi
nally
the
com
par
at
o
r
com
par
e
s
both
counter
values
with
one
an
ot
her
to
gen
e
rat
e
a
respo
ns
e
bit.
If
one
co
un
t
value
is
hi
gh
e
r
tha
n
the
oth
er
,
then
t
he resp
on
se b
it
is lo
gic
-
0, an
d vice
ver
s
a.
Ther
e
ar
e
k
C
2
po
ssi
ble
pairs
of
ROs
to
ge
ner
at
e
k
C
2
si
ng
le
-
bit
respo
ns
es.
H
owe
ver,
d
ue
to
the
occurre
nce
of
cor
relat
ed
re
sp
onse
bits,
the
1
-
out
-
of
-
p
m
asking
sc
he
m
e
was
pr
op
os
e
d.
I
n
this
schem
e,
on
ly
one
res
pons
e
bit
is
gen
e
rated
from
a
gr
oup
of
p
ROs,
wh
e
re
the
two
ROs
that
has
the
la
rg
est
dif
fe
ren
ce
in
osc
il
la
ti
on
f
reque
ncy
are
c
om
par
ed
[5
]
.
T
hu
s
,
t
he
act
ual
nu
m
ber
of
res
ponse
bits
is
the
n
cal
culat
e
d
by
k/p.
Wh
il
e
this
sc
hem
e
add
re
sses
the
issue
of
correla
te
d
bits,
it
reduces
th
e
nu
m
ber
of
CR
Ps
at
the
c
os
t
of
increase
d har
dware
over
hea
d.
Much
resea
rch
effo
rt
ha
s
be
en
m
ade
to
i
m
pr
ov
e
t
he
R
OPUF
i
n
te
rm
s
of
perform
a
nce
m
et
rics,
nam
ely
un
iq
ue
ness,
unif
or
m
ity,
bit
-
al
ia
sing,
an
d
reli
abili
ty
;
to
increase
the
nu
m
ber
of
CR
Ps;
or
t
o
si
m
pl
y
reduce
the
siz
e
of
t
he
ci
rc
uit.
On
e
e
xam
p
le
is
the
co
nf
i
gura
ble
ROP
UF
[16,
36
-
38
]
w
he
r
eby
the
ROs
c
an
be
config
ur
e
d
by
sel
ect
ing
the
de
sired
in
ve
rter
s
in
them
.
This
m
et
ho
d
is
done
to
im
pr
ov
e
pe
rfor
m
ance
m
e
tric
s,
par
ti
cula
rly
uniqu
e
ness
a
nd
r
el
ia
bili
ty
,
as
well
as
to
avo
id
us
in
g
the
1
-
out
-
of
-
p
m
as
king
schem
e.
The
m
ai
n
dr
a
w
back,
ho
wev
e
r,
is
the
i
ncr
ease
d
over
he
ad
as
a
2
-
to
-
1
m
ult
iplexer
is
require
d
i
n
e
ve
ry
sta
ge,
or
i
n
oth
er
words, a
fter e
ve
ry in
ver
te
r
in e
ach R
O.
Anothe
r
ROP
UF
exam
ple
is
the
design
tha
t
has
ROs
with
co
nf
i
gurab
le
du
ty
cy
cl
es
[1
7].
The
dut
y
cy
cl
e
of
the
RO
is
co
nf
ig
ur
e
d
by
co
ntr
olli
ng
the
t
ra
ns
ist
or
widt
h
rati
o
an
d
th
e
nu
m
ber
of
sta
ges.
The
c
om
par
iso
n
betwee
n
tw
o
ROs
ca
n
the
n
be
m
ade
from
diff
e
re
nt
node
s
within
t
he
RO
since
the
res
ul
ta
nt
os
ci
ll
at
ion
fr
e
quency
is
no
w
di
ff
ere
nt
du
e
to
al
te
red
du
ty
cy
cl
es.
Th
e
purpose
of
this
desi
gn
is
to
inc
reas
e
the
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Ring oscil
lat
or p
hysicall
y
un
c
lonab
le
f
unct
io
n usin
g
se
quen
ti
al ring o
sci
ll
ato
r…
(
J
ulius
Han Lo
ong Te
o
)
895
nu
m
ber
of
C
RPs.
Howe
ver,
this
desig
n
increases
ha
rdwar
e
over
hea
d
as
it
requir
es
two
m
or
e
m
-
to
-
1
m
ul
ti
plexer
s,
wh
e
re m
is the
nu
m
ber
of stag
es in eac
h
R
O.
The
wav
e
f
or
m
ROP
UF
[18]
us
es
a
noth
e
r
RO
to
sam
ple
the
P
UF
’
s
R
O,
beca
us
e
it
has
highe
r
fr
e
qu
e
ncy
(a
nd
sam
pling
rate)
than
c
onve
ntion
al
cl
oc
ks
.
Thu
s
,
res
ponse
gen
e
rati
on
is
sp
ed
up
a
nd
pow
e
r
consum
ption
i
s
re
du
ce
d.
M
eanwhil
e
the
rand
om
te
le
gr
aph
ic
noise
(
RTN)
i
nduce
d
ROP
UF
im
pr
oves
reli
abili
ty
[
19
]
.
How
e
ve
r,
both RO
PUFs
ha
ve
uniq
uen
e
ss
va
lues that a
re
r
el
at
ively
low
.
Figure
5
.
ROP
UF
by S
uh
a
nd D
e
vad
a
s [5]
In
this
pa
pe
r,
the
m
e
m
r
ist
or
is
introdu
ce
d
i
n
the
inv
e
rtin
g
un
it
s
of
the
ring
os
ci
ll
at
or
s
(ROs
)
as
it
causes
a
m
or
e
rand
om
os
ci
ll
a
ti
ng
f
reque
ncy
[35].
F
ur
t
her
m
or
e
,
a
di
ff
e
ren
t
m
e
tho
d
i
n
ge
ne
rati
ng
a
res
pons
e
is
pro
po
se
d
that
sign
ific
a
ntly
e
xp
a
nds
the
CR
P
set
with
on
l
y
a
s
m
al
l
nu
m
ber
of
ROs
.
It
is
al
so
sh
own
t
hat
the
pro
po
se
d
m
et
ho
d
has
good
pe
rfor
m
ance
m
e
t
rics
an
d
str
ong
resist
ance
to
m
od
el
li
ng
at
ta
cks
by
s
upport
vecto
r
m
achine (S
VM
),
wh
ic
h
is
one
of the
widely
use
d
m
achine le
arn
i
ng alg
or
it
hm
s tod
ay
.
2.
RESEA
R
CH MET
HO
D
2.1
.
Memrist
or
in
Rin
g
O
s
ci
ll
at
or
The
basic
ri
ng
os
ci
ll
at
or
(R
O)
con
sist
s
of
an
odd
num
ber
of
inv
e
rters
or
NO
T
gates
tha
t
are
loo
pe
d
in
a
ri
ng,
as
it
s
nam
e
su
ggest
s.
I
n
this
pa
per,
rathe
r
t
han
usi
ng
i
nv
e
rters
(
us
ua
ll
y
sta
ti
c
CM
OS
)
a
re
re
placed
with
com
m
on
-
so
urce
am
plifi
ers
with
m
e
m
risti
ve
load
,
as
sh
ow
n
in
Fi
gure
6.
The
r
easo
ns
f
or
us
ing
this
config
ur
at
io
n
are
it
s
relat
ively
s
m
a
ll
siz
e
com
par
ed
to
CM
OS
com
ponen
ts
a
nd
m
ore
rando
m
os
ci
ll
at
ion
fr
e
qu
e
ncy
[35].
Figure
6
.
MOS
FET c
omm
on
-
so
urce am
plifie
r wit
h
re
sist
ive load (left
)
a
nd m
e
m
rist
ive load (
rig
ht)
2.2
.
Se
quen
ti
al
Rin
g Osci
ll
ator P
airs
The
ge
ner
at
i
on
of
on
res
ponse
bit
is
m
ade
fr
om
the
co
m
par
is
on
of
on
e
pair
of
R
Os.
In
a
dd
it
io
n,
to
ov
e
rco
m
e
t
he
issue
of
co
rr
el
at
e
d
bits,
s
ever
al
m
et
ho
ds
hav
e
bee
n
pr
opos
e
d
but
at
the
cost
of
inc
reased
hard
war
e
over
head,
wh
il
e
othe
rs do n
ot si
gn
i
ficantl
y im
pr
ove the
p
e
rfo
rm
ance m
et
rics va
lues.
In
this
re
searc
h,
it
is
pr
op
os
e
d
that
the
ge
ne
rati
on
of
the
re
sp
onse
is
do
ne
by
a
sequ
e
nce
of
sel
ect
ion
of
R
O
pai
r.
I
n
oth
e
r
w
ords
,
the
outp
ut
bit
f
ro
m
ever
y
RO
pair,
is
a
rr
a
ng
ed
to
form
one
ROPUF
re
spon
s
e.
In
this r
esea
rch, the n
um
ber
o
f
ROs
us
ed
in the s
i
m
ulati
on
is eigh
t i
.e.
k
= 8
ROs.
T
her
e a
re th
en
8
C
2
= 2
8
RO
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
1
3
, N
o.
3
,
M
a
r
c
h
2019
:
8
9
2
–
9
0
1
896
pairs
an
d
th
us
, 2
8
ou
t
pu
t
bi
ts. Th
ese
28
outp
ut
bits
can
be
arr
a
ng
e
d
in
28!
ways,
or
rou
ghly
3×1
0
29
respon
s
es,
wh
ic
h
is
a
n
e
xtrem
el
y
la
rg
e
num
ber
.
T
he
chall
enge
is
th
en
⌈
log
2
(2
8!)
⌉
=
98
bits.
A
n
inter
facin
g
ci
rc
uitry
receives
the
c
halle
ng
e
a
nd
t
hen
se
quentia
ll
y
sel
ect
s
pairs
of
ROs
f
or
the
ge
ner
at
io
n
of
a
28
-
bit
res
pons
e
.
Ther
e
f
or
e,
with
a
sm
all
nu
m
ber
of
R
Os,
th
e
num
ber
of
C
RPs
is
sig
nific
antly
la
rg
e.
Al
so
,
with
a
m
ulti
ple
-
bit
respo
ns
e
as
c
om
par
ed
to
a
sing
le
-
bit
resp
onse
,
the
di
ff
ic
ulty
in
predict
ing
a
res
pons
e
is
incr
ease
d
expo
nen
ti
a
ll
y f
ro
m
2
to
2
n
.
2.3
.
Sim
ulat
i
on
Setup
The
pro
pose
d
m
e
m
ristor
-
bas
ed
ROP
UF
,
as
sh
ow
n
in
Fig
ur
e
7,
us
e
s
k
=
8
ROs,
w
he
re
each
R
O
consi
sts
of
fi
ve
sta
ges,
an
d
e
ach
sta
ge
is
th
e
com
m
on
-
so
urce
am
plifie
r
with
m
e
m
r
ist
ive
load
as
desc
r
ibed
in
Sect
ion
I
II
-
A.
The
ROs
a
re
s
ent
to
a
pair
of
8
-
to
-
1
m
ulti
pl
exer
s
.
Eac
h
m
ulti
plexer
outp
ut
is
sent
to
a
c
ounter
.
Both c
ounter
outp
uts ar
e
the
n sent to
the c
omparat
or
f
or
t
he resp
onse
bit ge
ner
at
io
n.
Figure
7
.
Pro
pose
d
m
e
m
risto
r
-
base
d
RO
PUF
Ci
rcu
it
sim
ula
ti
on
s
wer
e
pe
rfor
m
ed
on
t
wo
SP
ICE
si
m
ula
tors,
nam
el
y
LTspice
I
V
by
Linea
r
Tech
no
l
og
y
C
orp
or
at
io
n
an
d
Desig
n
Ar
c
hitec
t
-
IC
(DA
-
IC
)
by
Me
ntor
G
raphics
I
ncorpor
at
e
d.
F
or
LT
sp
ic
e
IV
,
Sil
Terr
a
180nm
CM
OS
process
with
1.8
V
sup
ply
vo
lt
age
was
us
e
d.
As
f
or
D
A
-
IC,
Sil
Ter
ra
130nm
CM
OS
process
w
it
h 1.2
V
s
up
ply v
oltage
wa
s u
se
d.
The
m
e
m
ristor
m
od
el
us
ed
in
the
sim
ulati
on
is
pr
e
pa
red
by
Bi
olek
et
al
.
a
s
a
SP
ICE
s
ubci
rcu
it
f
or
transient
a
naly
sis
[39].
The
m
e
m
ristor
SP
IC
E
subci
rc
uit
w
as
ada
pted
w
he
re
the
m
e
m
ris
tor
de
vice
par
a
m
et
ers
are
s
how
n
in
Table
1.
T
o
si
m
ula
te
the
ra
ndom
and
unc
ontr
ollable
va
ri
at
ion
s
i
n
the
m
anu
fact
ur
i
ng
process
,
20% Mo
nte Ca
rlo variat
io
n w
as in
j
ect
ed
i
nto t
he
init
ia
l m
e
m
resist
ance,
M
INIT
, an
d l
en
gth o
f
m
e
m
ristor,
D
.
Table
1
. Mem
ristor
Pa
ram
et
ers
Me
m
r
isto
r
p
ar
a
m
e
ter
Valu
e
Res
istan
ce a
t O
N s
tate
M
ON
1
0
0
kΩ
Res
istan
ce a
t OF
F
state
M
OFF
1
6
kΩ
Initial resistan
ce
M
I
NI
T
1
1
k
Ω (
±2
0
%)
Leng
th
of
m
e
m
rist
o
r
D
1
0
n
m
(
±2
0
%)
Migratio
n
co
ef
f
icien
t
µ
1
0
f
m
2
/
(V∙
s)
Bo
u
n
d
ary
con
trol p
ara
m
et
er
p
10
2.4
.
R
OP
UF
Eva
lu
at
i
on
Cr
iteri
a
Detect
ion
of
bi
as
:
T
her
e
are
a
fe
w
aspects
us
e
d
t
o
detec
t
bias
[40
-
42]
,
but
th
ree
c
om
m
on
ly
us
ed
m
et
rics
are
ta
ken
into
discu
ss
ion
,
nam
el
y
un
iqu
e
ness,
unif
orm
i
ty
,
and
bit
-
al
ia
sing
,
wh
ic
h
hav
e
bee
n
de
r
ived
by
Ma
it
i
et
al
.
[42].
T
he
c
om
pu
ta
ti
on
of
these
perf
orm
ance
m
et
rics
wa
s
pe
rfor
m
ed
us
in
g
M
A
TLAB.
To
cal
c
ulate
th
ese m
et
rics, the
f
ollow
i
ng p
a
r
a
m
et
ers
are
ne
eded.
a)
x
–
the
num
ber o
f
P
UF
ci
rcu
it
s in
t
he
set
|
b)
n
–
the
num
ber o
f bit
s in
the
P
UF
res
pons
e
c)
HD(R
i
,R
j
)
–
t
he
H
am
m
ing
dis
ta
nce b
et
wee
n t
wo
res
pons
es
, R
i
and R
j
, w
hich
a
re
o
f
t
he
sa
m
e leng
th
In
ge
ne
ral,
Ha
m
m
ing
distance
between
tw
o
string
s
is
the
nu
m
ber
of
pos
it
ion
s
that
have
diff
e
rent
sy
m
bo
ls
or
c
ha
racters.
It
can
on
ly
be
ap
plied
betwee
n
tw
o
string
s
that
ar
e
of
the
sam
e
l
eng
t
h.
I
n
the
c
on
te
xt
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Ring oscil
lat
or p
hysicall
y
un
c
lonab
le
f
unct
io
n usin
g
se
quen
ti
al ring o
sci
ll
ato
r…
(
J
ulius
Han Lo
ong Te
o
)
897
of
this
researc
h,
Ham
m
ing
di
sta
n
ce
bet
wee
n
tw
o
res
ponse
sti
ng
s
is
s
pe
ci
fical
ly
the
num
ber
of
bit
-
posit
ions
that d
o no
t
ha
ve
the sam
e b
it
.
Un
i
qu
e
ness
es
tim
a
te
s
the
abili
ty
of
a
PU
F
ty
pe
to
un
i
qu
e
ly
disti
ng
uish
on
e
ci
rc
uit
fro
m
ano
ther
.
In
oth
er
resear
ch
li
te
ratur
e,
un
i
qu
e
ness
is
te
rm
ed
a
s
inter
-
c
hip
va
riat
io
n
or
inter
-
c
hip
Ham
m
ing
Dista
nce
(inter
-
H
D).
U
ni
qu
e
ness
is
cal
culat
ed
by
ave
rag
i
ng
al
l
Ha
m
m
ing
distances
of
al
l
po
ssi
ble
pairs
of
re
sp
onse
s
for
the
sam
e
app
li
ed
c
halle
ng
e
.
The
e
xpr
ession
for
cal
culat
ing
un
i
queness
is
show
n
in
(
2).
The
ideal
value
is
50%.
Un
iq
ue
ness
=
1
(
2
)
∑
∑
(
,
)
=
+
1
−
1
=
1
×
100%
(2)
Un
i
form
i
ty
m
e
asur
e
s
the
pro
portio
n
of
lo
gic
-
0s
an
d
lo
gic
-
1s
in
a
respo
ns
e
of
eac
h
P
UF
.
This
m
et
ric
ind
ic
at
es
w
het
her
t
her
e
is
bias
within
t
he
re
sp
onse
.
F
or
t
he
sam
e
app
li
ed
chall
en
ge,
le
t
r
i,j
be
the
j
th
bit
of
t
he
i
th
resp
onse
(R
i
)
that
is
n
bits
l
ong,
the
n
the
m
at
he
m
at
ic
a
l
e
xpressi
on
f
or
cal
culat
ing
unif
or
m
it
y
of
the
i
th
PU
F
ci
rcu
it
is give
n
by (
3).
The
id
eal
value
is
al
s
o
50%;
a resu
lt
that
is
le
ss
tha
n
50%
in
dicat
e
s
a h
ig
her
pro
portio
n
of
lo
gic
-
0s
whereas
a
res
ult
that
is
m
or
e
t
ha
n
50%
in
dicat
e
s
a
higher
pro
portio
n
of
lo
gic
-
1s
for
t
hat
res
ponse
.
The
ove
rall
unifor
m
it
y
of
the
PU
F
ty
pe
ca
n
be
obta
ine
d
by
si
m
ply
aver
ag
ing
eac
h
unif
orm
i
ty
value
over
the
nu
m
ber
of r
es
ponse
s, x.
Un
if
orm
it
y
=
1
∑
,
=
1
×
100%
(3)
Bit
-
al
ia
sing
est
i
m
a
te
s
the
ten
de
ncy
of
the
PU
Fs
to
generate
identic
al
responses,
w
hich
is
a
n
unwa
nted
e
ff
e
ct
.
I
n
ot
her
w
ords,
bit
-
al
ia
sing
m
easur
es
the
af
finity
of
a
bit
posit
io
n
in
a
re
spo
ns
e
towa
r
ds
ei
ther
log
ic
-
0
or
log
ic
-
1.
By
le
tt
ing
r
i,j
be
the
j
th
bit
of
the
i
th
resp
onse
(R
i
)
that
is
n
bits
lon
g,
the
m
a
them
at
ic
al
expressi
on
f
or
cal
culat
ing
th
e
bit
-
al
ia
sin
g
a
t
the
j
th
bit
posit
ion
is
s
how
n
in
(
4).
Li
ke
un
i
form
i
ty
,
the
idea
l
value
is
50%
wh
e
re
a
res
ult
that
is
le
ss
tha
n
50%
in
dicat
es
a
hig
he
r
pro
portio
n
of
log
ic
-
0s
an
d
a
res
ult
that
is
gr
eat
er
tha
n
50
%
ind
ic
at
es
a
hig
he
r
pro
portio
n
on
log
ic
-
1s
f
or
that
bit
po
sit
ion
.
T
he
overal
l
bit
-
al
ia
sing
of
the
PU
F
ty
pe
ca
n be
ob
ta
ine
d by
si
m
ply avera
gin
g ea
ch
b
it
-
al
i
asi
ng v
al
ue ove
r
the
num
ber
of r
es
pons
e
b
it
s,
n.
Bit
-
al
ia
sing
=
1
∑
,
=
1
×
100%
(4)
Resi
stan
ce
to
SV
M
:
A
no
t
her
c
rite
rion
to
e
val
uate
the
PUF
i
s
re
gardin
g
it
s
abili
ty
to
resist
m
od
el
li
ng
at
ta
cks,
pa
rtic
ularly
by
m
ac
hin
e
le
ar
ning
al
gorithm
s.
In
this
resea
rch,
the
m
achine
le
arn
i
ng
al
gorith
m
that
was
ta
ke
n
int
o
co
ns
i
der
a
ti
on
is
SV
M
.
Th
e
trai
ning
an
d
te
sti
ng
of
th
e
CR
Ps
wer
e
perform
ed
us
ing
t
he
LIBSV
M
pac
ka
ge
in
MA
TL
AB,
w
hich
is
m
ade
avail
able
on
li
ne
[43].
A
lt
ho
ug
h
the
C
RP
set
of
the
pro
posed
ROPUF
is
sig
nificantl
y
la
rge,
trai
ning
a
nd
te
sti
ng
t
he
whole
set
w
ould
ta
ke
a
c
on
side
rab
ly
lo
ng
ti
m
e.
Hen
ce
,
due
to
this
com
pu
ta
tio
nal
co
ns
trai
nt
,
on
ly
1024
C
RPs
are
us
e
d,
wh
ic
h
we
re
ta
ken
from
the
first
te
n
chall
enge
bits.
The
trai
ni
ng
s
et
siz
es
us
e
d
a
re
20
%
an
d
50
%
of
the
CR
P
subset.
The
n,
the
rest
of
the
CR
P
su
bse
t
is
us
ed
fo
r
te
st
ing.
S
i
m
i
la
rly
,
the
CR
P
pairs
to
be
trai
ne
d
are
cho
se
n
at
ran
dom
.
The
ex
pected
m
od
el
li
ng
accuracy
is
50%
,
wh
ic
h
is
the
prob
a
bili
ty
of
ob
ta
ining
one
out
of
t
wo
e
qual
ly
po
ssi
ble
outc
om
es,
li
ke
a fai
r
c
oin
toss. In t
he
con
te
xt of th
is
rese
arch, the
tw
o o
utcom
es are s
i
m
pl
y l
og
ic
-
0
a
nd lo
gic
-
1.
3.
RESU
LT
S
A
ND AN
ALYSIS
3.1.
Det
ec
tio
n of
Bias
Table
2
s
how
s
the
resu
lt
s
of
the
pro
po
s
ed
m
e
m
r
ist
or
-
base
d
ROP
UF
in
te
r
m
s
of
un
i
qu
e
ness
,
un
i
form
i
ty
,
and
bit
-
al
ia
sin
g,
in
com
par
iso
n
with
a
fe
w
othe
r
ROP
UF
des
ign
in
t
he
li
te
ratur
e.
Ta
ke
no
te
that
so
m
e
inform
ation
,
par
ti
cularl
y
on
perform
a
nce
m
e
tric
val
ues
are
not
app
li
cable
or
not
avail
able
in
the
resp
ect
i
ve
li
te
r
at
ur
e,
a
nd
th
us
are
la
belle
d
‘N
/A’
.
This
is
because
their
pri
m
ary
aspect
of
e
valuati
on
of
their
ROPUFs is
on
un
i
qu
e
ness
or i
nter
-
H
D.
The
res
ults
for
bo
th
180nm
a
t
1.
8
V
a
nd
130n
m
at
1.
2V
s
how
pe
rfor
m
ance
m
e
tric
values
that
are
ver
y
cl
os
e
to
on
e
an
o
th
er,
ind
ic
at
ing
c
onsist
ency
re
ga
rd
le
ss
of
the
ci
rcu
it
si
m
ulato
r
,
CM
OS
pr
ocess,
and
sup
ply
volt
age
us
e
d.
M
or
e
im
po
rta
ntly
,
the
pe
rfor
m
ance
m
et
ric
values
f
or
both
ci
rcu
it
sim
ulato
rs
a
re
ver
y
cl
os
e
t
o
t
he
i
deal
50%
value.
T
hus,
t
he
pro
po
s
ed
m
e
m
ristor
-
bas
ed
R
OPUF
show
go
od
perform
ance,
i
m
plyi
ng
that i
t i
s a w
el
l
-
func
ti
on
in
g
P
UF.
In
co
ntrast
to
m
os
t
of
the
othe
r
ROP
UF
des
ign
s
,
the
pro
posed
m
e
m
r
ist
or
-
base
d
ROP
UF
sh
ow
bette
r
perform
ance
m
et
ric
resu
lt
s
than
[5,
16,
18,
19]
especi
al
ly
in
te
r
m
s
of
uniq
uen
es
s.
In
c
o
ntrast
to
[
17]
,
the
resu
lt
s
are
com
par
able
to
ano
the
r.
Most
of
the
oth
e
r
r
esearch
e
ffor
ts
i
m
pr
ov
e
certa
in
aspects
but
forgo
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
1
3
, N
o.
3
,
M
a
r
c
h
2019
:
8
9
2
–
9
0
1
898
oth
e
rs,
s
uc
h
as
i
m
pr
ovin
g
pe
rfor
m
ance
sp
e
ed
an
d
reli
abili
ty
against
volt
age
cha
nges
but
at
the
ex
p
e
ns
e
of
reducin
g u
niqu
eness [
18,
19
]
.
The
pri
m
ary
a
dv
a
ntage
of
th
is
desi
gn
is
th
at
it
on
ly
us
es
ve
ry
fe
w
RO
s
to
pro
duce
a
ve
ry
la
r
ge
nu
m
ber
of CRPs. T
he
idea is t
hat a seq
ue
nce
o
f
al
l t
he
possi
ble p
ai
rs of
R
Os
is use
d
to
ge
ner
at
e a m
ultip
le
-
bit
respo
ns
e,
rathe
r
tha
n
us
in
g
one
pair
of
ROs
for
a
sin
gle
-
bit
respo
ns
e.
Ne
ver
t
heless,
it
is
al
so
possible
t
o
us
e
on
ly
a
s
ubset
of
the
possible
pairs
of
ROs
,
e
sp
eci
al
ly
wh
e
n
the
num
ber
of
ROs
us
e
d
is
la
rg
e
.
Wh
il
e
the
r
e
are
oth
e
r
m
et
ho
ds
that
hav
e
bee
n
pro
po
se
d
to
i
nc
rease
the
CR
P
set
and
im
pr
ov
e
uniq
u
ene
s
s,
it
is
no
t
as
ef
fecti
ve
as the m
et
ho
d
pro
po
se
d
i
n
thi
s p
a
per.
Table
2
. Per
for
m
ance
Me
tric
Re
su
lt
s
ROPUF
Nu
m
b
e
r
o
f
ROs
Nu
m
b
e
r
o
f
CR
Ps
Perf
o
r
m
an
ce
m
et
ri
c (
%)
Un
iq
u
en
ess
Un
if
o
r
m
it
y
Bit
-
aliasin
g
Clas
sical [
5
]
1024
128
4
6
.15
N/A
N/A
Gao
et
al
.
[
1
6
]
32
16
4
6
.79
N/A
N/A
Ag
u
stin
et
al.
[
1
7
]
128
2048
4
9
.41
4
8
.13
5
2
.10
Tana
m
o
to
[
1
8
]
24
N/A
4
2
.20
4
6
.74
N/A
Yo
sh
in
ag
a et
al.
[
1
9
]
256
128
48
N/A
N/A
Prop
o
sed
(
1
8
0
n
m
,
1
.8V)
8
3
×1
0
29
4
8
.57
5
1
.43
5
1
.43
Prop
o
sed
(
1
3
0
n
m
,
1
.2V)
8
3
×1
0
29
5
1
.36
4
9
.49
4
8
.81
3.2.
Resist
ance
to SV
M
Table
3
s
how
s
the
m
od
el
lin
g
acc
uracy
by
SV
M
on
the
pro
posed
m
e
m
ristor
-
bas
ed
RO
P
U
F
.
Althou
gh
the
m
od
el
li
ng
acc
ur
acy
is
not
ve
ry
cl
os
e
to
t
he
ideal
50%,
the
values
are
acce
ptable
as
the
S
VM
is
sti
ll
un
able
to
accuratel
y
pre
dict
the
res
ponse
s
f
or
both
20
%
an
d
50
%
tra
ining
siz
es.
In
add
it
io
n,
t
his
r
esult
is
com
par
able
to
a
previ
ou
s
rese
arch
ef
fort
on
t
he
RO
PUF
th
a
t
was
al
s
o
te
ste
d
with
S
VM,
wh
e
re
by
the
hi
gh
e
st
pr
e
dicti
on
acc
ur
acy
obta
ine
d
was
on
ly
60.
9%
[
44
]
.
I
n
fa
ct
,
the
resu
lt
s
are
m
uch
bette
r
in
con
t
rast
to
the
cl
assic
al
arb
it
er
PU
F
,
a
diff
e
r
ent
ty
pe
of
PUF,
w
hich
sho
w
ed
up
to
99%
m
od
el
ing
accu
racy
[45
-
48
]
,
a
nd
th
us
ind
ic
at
e
high
vulne
rab
il
it
y
to
m
od
el
li
ng
at
tack
s.
T
her
e
fore,
the
resu
lt
s
indi
cat
e
that
the
pr
op
os
ed
m
e
m
ri
stor
-
base
d
RO
PUF
is resist
ant to
a
tt
acks b
y
SV
M
.
Table
3
. SVM
Mod
el
li
ng
Acc
ur
acy
Tr
ain
in
g
size (%
)
20
50
Mod
ellin
g
accurac
y
(
%
)
6
1
.95
5
8
.59
4.
CONCL
US
I
O
N
I
n
s
hort,
tw
o
m
ai
n
changes
on
th
e
ROP
U
F
are
pro
pose
d.
T
he
first
ch
ang
e
is
the
inc
lusio
n
of
t
he
m
e
m
ristor
in
the
in
ver
ti
ng
unit
s
of
the
R
O.
T
his
is
bec
ause
the
m
e
mristor
is
sm
all
er
tha
n
m
os
t
CM
O
S
com
po
ne
nts,
t
her
e
by
r
e
duci
ng
ci
rcu
it
siz
e
an
d
po
wer
c
on
s
um
ption
.
F
ur
t
her
m
or
e,
it
is
cl
aim
ed
that
the
m
e
m
ristor
is
r
el
at
ively
com
p
at
ible
with
C
MOS
fabrica
ti
on
sta
ndar
ds.
The
se
co
nd
c
ha
ng
e
is
the
m
e
thod
of
gen
e
rati
ng
the
respo
ns
e
of
t
he
P
UF
.
In
ot
her
R
OPUF
de
sign
s
,
one
pa
ir
of
R
O
ge
ne
rates
one
res
ponse
bit;
bu
t
in
the
pro
po
s
ed
m
e
m
rist
or
-
ba
sed
RO
P
UF
,
on
e
se
que
nce
of
RO
pai
rs
ge
ner
at
es
one
m
ulti
bit
resp
onse
.
This
m
et
ho
d
sign
i
ficantl
y
expands
the
CR
P
s
et
li
st
witho
ut
extensi
ve
ci
rcui
t
ov
er
hea
d.
T
he
resu
lt
s
in
te
r
m
s
of
un
i
qu
e
ness
,
u
ni
fo
rm
it
y,
and
bi
t
-
al
ia
sing
sho
w
that
the
pr
opose
d
m
e
m
rist
or
-
ba
sed
R
OPUF
ha
ve
li
tt
le
bias
in
the
res
pons
e
.
Also
,
S
VM
is
unable
t
o
ac
cur
at
el
y
m
od
el
the
be
ha
vior
of
the
pro
posed
m
e
m
ristor
-
base
d
ROPUF,
t
her
e
by
the
R
OP
U
F
is
resist
ant
t
o
at
ta
cks
by
S
VM.
I
n
c
onc
lu
sion,
the
pro
pose
d
m
e
m
ris
tor
-
base
d
ROPUF is a
sim
ple ye
t robu
s
t PUF desi
gn.
As
f
or
f
uture
r
esearch
pla
ns
, th
e
pro
pose
d
m
e
m
ristor
-
bas
ed
RO
PUF
is
to
un
derg
o
f
urt
her
te
sti
ng a
nd
validat
io
n,
s
uc
h
as v
oltage
a
nd
te
m
per
at
ur
e
r
el
ia
bili
ty
,
ran
dom
ness
te
st
li
ke
the
N
IS
T
te
st
su
it
e,
an
d
m
odel
ing
at
ta
cks
us
i
ng
ot
her
m
achine
l
earn
i
ng
al
gorithm
s.
Even
t
ually
,
the
RO
PUF
is
to
be
fa
br
ic
a
te
d
an
d
im
plem
ented
as an
actual
physi
cal
h
ardwa
r
e d
e
vice f
or se
cur
it
y p
urp
os
es
.
ACKN
OWLE
DGE
MENTS
This
resea
rch
i
s
su
pp
or
te
d
by
the
F
undam
e
ntal
Re
search
Gr
a
nt
Schem
e
(F
RGS
)
unde
r
the
pro
j
ect
cod
e
FRG
S/1/
2015/TK
04/U
NI
T
EN/
02
/
2,
wh
ic
h
ha
s
bee
n
awa
r
ded
by
the
Mi
nistry
of
Higher
Ed
uc
at
ion
,
Ma
la
ysi
a,
and
by
th
e
Un
i
ve
rsiti
Tena
ga
Nasi
on
al
(UNI
T
EN
)
In
te
r
nal
G
ra
nt
un
der
the
pro
j
e
ct
code
J5
10
0507
61. Also,
m
any thanks t
o
al
l coll
ea
gu
e
s in
UNI
TE
N
f
or thei
r
insi
gh
t a
nd s
uppor
t.
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Ring oscil
lat
or p
hysicall
y
un
c
lonab
le
f
unct
io
n usin
g
se
quen
ti
al ring o
sci
ll
ato
r…
(
J
ulius
Han Lo
ong Te
o
)
899
REFERE
NCE
S
[
1
]
B.
Gass
end,
D.
Cla
rke
,
M.
v
.
Dijk,
and
S.
Deva
das,
"S
ilic
on
ph
y
si
ca
l
r
an
dom
func
ti
ons,"
pre
sente
d
at
t
he
Proce
edi
ngs of
t
he
9th
ACM
con
fer
ence
on
Com
pute
r and
com
muni
cations
sec
u
ri
t
y
,
W
ashingt
on
,
DC,
US
A,
2002.
[
2
]
J.
W
.
L
ee
,
D.
L
im,
B.
Gass
end
,
G.
E
.
Suh,
M.
Van
Dijk,
and
S
.
Deva
d
as,
"A
t
ec
hniqu
e
to
build
a
se
cre
t
ke
y
i
n
int
egr
at
ed
ci
rcu
i
ts
for
ide
nti
f
ic
a
t
ion
and
aut
h
ent
i
ca
t
ion
appl
i
catio
ns,"
in
VLSI
Circui
ts,
2004
.
Digest
of
Technical
Pape
rs
.
2004
Sy
mpos
ium
on
,
20
04,
pp
.
176
-
179
.
[
3
]
D.
Li
m
,
J.
W
.
Le
e
,
B.
Gass
end
,
G.
E.
Suh,
M.
v.
Dijk,
and
S.
Deva
das,
"Extr
ac
t
ing
sec
re
t
ke
y
s
from
int
egr
a
t
ed
ci
rcu
it
s,"
I
EE
E
T
rans
act
ions o
n
V
ery
Lar
ge
S
cal
e
Inte
gration
(
VLSI)
Syste
ms
,
vo
l
.
13,
pp
.
1200
-
12
05,
2005
.
[
4
]
U.
Ruhrm
ai
r
and
D.
E.
Holcomb,
"P
UFs
at
a
gla
nce,"
pre
sent
ed
at
the
Proce
ed
i
ngs
of
the
conf
ere
nce
on
Design
,
Autom
at
ion
&
T
est
in
Europ
e, Dre
sden,
Germ
any
,
2014
.
[
5
]
G.
E.
Suh
and
S.
Deva
das,
"P
hy
sic
al
uncl
on
able
func
ti
ons
for
devi
c
e
aut
hen
ti
c
at
ion
and
sec
r
et
ke
y
gene
r
at
ion
,
"
pre
sente
d
a
t
th
e P
roc
ee
dings of
t
he
44th
annu
al
Design
Autom
ation
Confer
ence,
San
Diego, Calif
ornia
,
2007.
[
6
]
R.
J.
Anderson,
Sec
urity
Engi
n
ee
ring:
A
guid
e
to
buil
ding
d
ep
endabl
e
and
dis
tribut
ed
systems
,
John
W
il
e
y
an
d
Sons
,
2001.
[
7
]
L.
Fei
te
n
,
J.
Oes
te
rl
e,
T.
Mar
ti
n
,
M.
Sauer
,
and
B
.
Be
cke
r
,
"S
y
ste
m
ic
Freque
nc
y
Bia
ses
in
Ring
Os
ci
ll
at
o
r
PU
Fs
on
FP
GAs,
"
IEE
E
Tr
ansacti
ons on Multi
-
S
cal
e
Computing
S
yste
ms
,
vol.
PP
,
pp.
1
-
1,
2016.
[
8
]
Q.
Chen,
G.
Csaba,
P.
Lugl
i
,
U.
Schli
cht
m
ann
,
a
nd
U.
Ruhrm
ai
r,
"The
Bista
ble
Ring
PU
F:
A
ne
w
arc
hit
e
ct
ure
f
or
strong
Ph
y
sical
Unclona
bl
e
Functi
ons,"
in
Har
d
ware
-
Or
ie
nte
d
S
ec
urit
y
and
Tr
ust
(
HO
S
T
)
,
2011
IEE
E
In
te
rnatio
nal
Symposium on
,
2
011,
pp
.
134
-
14
1.
[
9
]
J.
W
.
Jang
and
S.
Ghos
h,
"D
esi
gn
and
ana
l
y
s
is
of
novel
SR
A
M
P
UF
s
with
embedde
d
la
tc
h
f
or
robustness,"
in
Six
t
ee
nth
Int
ernati
onal
S
ymposi
um on
Quality
E
le
c
tronic
Design
,
2015
,
pp
.
298
-
302.
[
1
0
]
L.
Kus
te
rs,
T.
Ig
nat
enko
,
F.
M.
J.
W
il
le
m
s,
R.
Ma
es,
E
.
v.
d
.
Sluis,
and
G.
Sel
imis
,
"S
ec
urity
of
h
elper
data
sche
m
e
s
for
SR
AM
-
PU
F
in
m
ult
ipl
e
enr
oll
m
ent
sce
nar
io
s,"
in
2017
IEE
E
Inte
rnationa
l
Symposium
on
I
nformation
Theory
(
ISIT
)
,
2017,
pp.
1803
-
1807.
[
1
1
]
E.
I.
Vat
aj
e
lu,
G.
D.
Nata
le,
a
nd
P.
Prinet
to,
"Towards
a
highly
re
liable
SR
AM
-
base
d
PUF
s,"
in
2016
Des
ign,
Aut
o
mation
&
T
est
in
Europe
Co
nfe
renc
e
&
Ex
hibi
ti
on
(
DATE)
,
2
016,
pp
.
273
-
27
6.
[
1
2
]
G.
S.
Rose
and
C.
A
.
Me
ade
,
"P
erf
orm
anc
e
ana
l
y
sis
of
a
m
emristive
cro
s
sbar
PU
F
design,
"
in
2015
52
nd
ACM/
EDAC/
I
E
EE
Design
Au
to
mation
Conf
ere
n
ce
(
DAC)
,
2015,
pp.
1
-
6.
[
1
3
]
M.
Uddin,
M.
B.
Majumder,
G.
S.
Rose,
K.
Beckm
ann,
H.
Mane
m
,
Z.
Alamgir
,
et
al.
,
"Te
chni
q
ues
for
Im
prove
d
Rel
ia
b
il
i
t
y
in
Mem
risti
ve
Cross
bar
PU
F
Circ
uit
s,"
in
2016
IEE
E
Computer
Soci
e
ty
Annual
Sy
mpos
ium
on
VL
SI
(
ISVLSI
)
,
2016,
pp.
212
-
217
.
[
1
4
]
M.
Uddin,
M.
B.
Majumder,
a
nd
G
.
S.
Rose,
"Robus
tne
ss
A
naly
s
is
of
a
Mem
risti
ve
Cross
bar
PU
F
Against
Modeli
ng
At
ta
ck
s,"
IEEE
Tr
ansacti
ons on
Nanotechnology,
vo
l. 1
6,
pp
.
396
-
405
,
2017.
[
1
5
]
G.
S.
Rose,
M.
B.
Majumder,
a
nd
M.
Uddin,
"
Expl
oiting
Mem
risti
ve
Cross
bar
Mem
orie
s
as
Dual
-
Us
e
Secur
i
t
y
Prim
it
ive
s
in
IoT
Devic
es,
"
i
n
2017
IEE
E
Computer
Soci
e
ty
Annual
Sym
posium
on
VLS
I
(
ISVLSI
)
,
201
7,
pp.
615
-
620
.
[
1
6
]
M.
Gao,
K.
La
i
,
and
G.
Qu
,
"A
highly
fle
x
ib
le
ring
osc
il
l
at
o
r
PU
F,"
in
2014
51st
ACM/
E
DAC/IE
E
E
Desi
gn
Aut
omation
Con
fe
renc
e
(
DAC)
,
2014,
pp
.
1
-
6.
[
1
7
]
J.
Agus
ti
n
and
M.
L.
Lope
z
-
Val
le
jo
,
"A
te
m
per
a
ture
-
ind
epe
nden
t
PU
F
with
a
con
figura
bl
e
dut
y
c
y
c
le
of
CMO
S
ri
ng
oscil
lators,"
in
2
016
IEEE
In
te
rn
ati
onal
S
ymposi
um on
Circuits a
nd
Syste
ms
(
ISCAS)
,
2016,
pp
.
2
471
-
2474.
[
1
8
]
T.
Ta
n
amoto,
S.
Yasuda,
S.
Ta
ka
y
a
,
and
S.
Fujit
a
,
"P
hy
si
call
y
Un
cl
onab
le
Functi
o
n
using
Init
ia
l
W
ave
form
of
Ri
ng
Os
ci
ll
at
o
rs,"
IE
E
E
Tr
ansacti
ons
on
Circui
ts
and
Syste
ms
II: Expr
ess B
rief
s,
vol
.
P
P,
pp.
1
-
1,
2016.
[
1
9
]
M.
Yos
hina
ga,
H.
Aw
ano,
M.
Hirom
oto,
and
T.
Sato
,
"P
h
y
si
c
al
l
y
un
cl
onab
le
func
ti
on
using
RTN
-
induc
ed
d
e
l
a
y
fluc
tu
at
ion
in
r
i
ng
oscil
lators,"
in
2016
IEE
E
I
nte
rnational
S
y
mpos
ium
on
Cir
cui
ts
and
System
s
(
ISCAS)
,
20
16,
pp.
2619
-
2622
.
[
2
0
]
L.
O
.
Chua
,
"M
e
m
ristor
-
th
e
m
issing c
ir
cui
t
e
le
m
ent
,
"
IEEE
Tr
ans.
Circuit Theory
,
vol
.
18
,
pp
.
507
-
519,
/
/
1971
.
[
2
1
]
L.
O
.
Chua
and
S.
M.
Kang
,
"M
emristive
d
evice
s a
nd
s
y
s
te
m
s,"
Proc.
I
EEE,
vol. 64, pp. 209
-
223
,
/
/
1976
.
[
2
2
]
L.
Chu
a, "Re
sist
anc
e
sw
it
chi
n
g
m
emories
are
m
emristors,"
Appli
ed
Ph
ysic
s
A,
vo
l.
102
,
pp
.
765
-
7
83,
2011
.
[
2
3
]
R.
S.
W
il
liam
s,
"H
ow We
Found T
he
Miss
ing
Me
m
ristor,
"
IEEE
S
pec
trum,
vo
l. 45, pp. 28
-
35,
2008
.
[
2
4
]
D.
B.
Strukov,
G.
S.
Snider,
D.
R.
Stewar
t,
and
R.
S.
W
il
li
ams
,
"The
m
issing
me
m
ristor
found,
"
Nature,
vol.
45
3
,
pp.
80
-
83
,
05/01
/pri
nt
2008.
[
2
5
]
A.
G.
Radwa
n
and
M.
E.
Fouda,
"M
emristor:
M
odel
s,
T
y
pes,
an
d
Applic
at
ions
,
"
in
On
the
Mathe
matic
a
l
Mode
ling
of
Me
mr
istor,
M
emc
apacitor,
an
d
Me
minduc
tor
,
ed:
Spring
er,
20
15,
pp
.
13
-
49
.
[
2
6
]
P.
Maz
um
der
,
S
.
-
M.
Kang,
and
R.
W
ase
r,
"M
e
m
ristors:
devi
ces
,
m
odel
s,
and
appl
i
ca
t
ions,"
P
roce
edi
ngs
of
th
e
IEE
E
,
vo
l. 100,
pp.
1911
-
1919
,
2012.
[
2
7
]
T.
Prodrom
aki
s
and
C.
Touma
zo
u,
"A
rev
ie
w
on
m
emristive
device
s
and
app
licati
ons,"
in
El
e
ct
ron
ic
s,
Cir
cui
ts,
and
S
y
stems
(ICECS),
2010
17th
I
EEE
Int
ern
ationa
l Confere
nc
e
on
,
2010,
pp
.
934
-
9
37.
[
2
8
]
G.
S.
R
ose,
N.
McDonal
d,
L.
K.
Yan,
B.
W
y
soc
ki,
and
K.
Xu
,
"
Foundati
ons
of
m
emristor
base
d
PU
F
arc
hit
e
ct
ur
es,
"
in
2013
I
EEE/ACM Int
ernat
iona
l
Symposium on
Nanoscale
Architectures (
NANOAR
CH)
,
2013,
p
p.
52
-
57
.
[
2
9
]
J.
Mathe
w,
R.
S.
Chakra
bort
y
,
D.
P.
Sahoo,
Y.
Yang,
and
D.
K.
Pradha
n,
"A
novel
m
emristor
base
d
ph
y
sic
al
l
y
uncl
onab
le
fu
nction,
" Integr. VL
SI J.,
vo
l. 51, pp. 37
-
45,
2015.
[
3
0
]
G.
S.
Rose,
N.
McDonal
d,
L
.
K.
Yan,
and
B.
W
y
socki
,
"A
write
-
ti
m
e
base
d
m
emristive
PU
F
fo
r
har
dware
sec
ur
i
t
y
appl
i
ca
t
ions,"
in
2013
IEE
E/A
C
M
Inte
rna
tional
Confer
en
c
e
on
Com
pute
r
-
Aided
Design
(ICCAD
),
20
13,
pp.
830
-
833
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
1
3
, N
o.
3
,
M
a
r
c
h
2019
:
8
9
2
–
9
0
1
900
[
3
1
]
M.
Uddin,
M.
B
.
Majumder
,
G.
S.
Rose,
K.
Be
c
km
ann,
H.
Man
em,
Z.
Alamgir,
et
al.,
"Te
chni
q
ues
for
Im
prove
d
Rel
ia
b
il
i
t
y
in
Mem
risti
ve
Cross
bar
PU
F
Circ
uits
,
"
in
2016
IE
E
E
Com
pute
r
So
ci
e
t
y
Annual
S
ym
posium
on
VL
SI
(ISV
LSI),
2016,
pp.
212
-
217
.
[
3
2
]
J.
Rajendra
n
,
G.
S.
Rose,
R
.
Karr
i,
and
M.
Potkon
ja
k,
"N
ano
-
PP
U
F:
A
m
emristor
-
base
d
sec
ur
ity
pr
imiti
ve
,
"
in
201
2
IEE
E
Computer
Soci
e
ty
Annual S
ympo
sium on
VL
SI
,
2012
,
pp
.
84
-
87.
[
3
3
]
A.
Maz
ad
y
,
M.
T.
Rahman,
D.
Forte,
and
M.
Anw
ar,
"M
emristor
puf
—
a
sec
uri
t
y
primit
ive
:
The
or
y
an
d
expe
riment
,
"
I
E
EE
Journal
on
E
merging
and
S
elec
t
ed
Topics
in Circuit
s and
Sys
te
ms
,
vol
.
5
,
pp
.
222
-
229,
2015
.
[
3
4
]
Koebe
rl,
P.
,
et
a
l
.
(2013).
Mem
ristor
PU
F
s:
A
n
ew
gene
ration
of
m
emory
-
base
d
Phy
si
ca
l
l
y
Unc
l
onabl
e
Functi
on
s.
Design,
Autom
a
ti
on
&
T
est in
E
urope
Confer
ence
&
Exhi
b
it
ion
(
DA
TE
),
2013
.
[
3
5
]
N.
A.
N.
Hashi
m
,
J.
H.
L
.
T
eo
,
M.
S.
H
amid,
and
F.
A.
H
amid,
"A
naly
sis
of
Mem
ristor
-
base
d
Ring
Os
ci
llat
or
Random
Nu
m
ber
Gene
rat
or
,
"
pre
sente
d
a
t
the
1
2th
IEE
E
Int
ern
at
ion
al
Confer
en
ce
on
Sem
ic
onduct
or
Elec
tron
ics
(ICSE),
Kua
la
L
um
pur,
2016.
[
3
6
]
X.
Xin,
J.
-
P.
Kaps,
and
K.
Gaj,
"A
Configura
ble
Ring
-
Os
ci
ll
a
tor
-
Based
PU
F
for
Xili
nx
FP
GA
s,"
pre
sente
d
a
t
the
Proce
edi
ngs of
t
he
2011
14
th Eu
rom
ic
ro
Confer
e
nce
on
Digi
ta
l
Sy
stem Design
,
2
011.
[
3
7
]
A.
Mait
i
and
P.
Schaumont,
"I
m
proving
the
q
ual
ity
o
f
a
ph
y
s
ic
a
l
uncl
on
able
func
ti
on
using
conf
igura
b
le
rin
g
oscil
lators,"
in
2
009
Inte
rnat
iona
l
Conf
ere
nce on Fi
el
d
Progr
amm
able
Logi
c
and
Appl
ic
a
ti
ons
,
20
09,
pp
.
703
-
707
.
[
3
8
]
M.
Choudhur
y
,
N.
Pundir,
M.
Niamat,
and
M.
Mus
ta
pa,
"A
nal
y
si
s
of
a
novel
stag
e
conf
igur
abl
e
R
OP
UF
design,
"
i
n
2017
IEEE
60
th Inte
rnational
Mi
dwest
Symposiu
m on
Circuits an
d
Syste
ms
(
MWSCAS)
,
2017,
pp
.
942
-
945.
[
3
9
]
Z.
Biolek,
V.
Biol
kova,
an
d
D.
Biol
ek,
"S
PICE
m
od
el
of
m
emristor
with
nonli
n
ea
r
dopant
dr
i
ft,
"
Radi
oengi
n
ee
rin
g,
2009
.
[
4
0
]
M.
Majz
oobi
,
F.
Kous
hanf
ar,
and
M.
Potkonj
ak,
"T
esti
ng
t
e
chni
ques
for
ha
rdware
sec
uri
t
y
,
"
in
2008
IEEE
Inte
rna
ti
ona
l Te
s
t
Confer
en
c
e, 20
08,
pp
.
1
-
10
.
[
4
1
]
Y.
Hori,
T.
Yos
hida
,
T
.
Kat
ashi
ta
,
and
A
.
Sato
h,
"Q
uant
i
tative
and
st
at
ist
ic
a
l
per
form
anc
e
ev
al
ua
ti
on
of
arb
i
t
er
ph
y
sic
al
un
cl
on
abl
e
func
t
ions
o
n
FP
GAs,
"
in
2
010
Inte
rn
at
ion
a
l
Confer
enc
e
on
Rec
onf
igura
bl
e
Com
puti
ng
and
FP
GAs,
2010,
p
p.
298
-
303
.
[
4
2
]
A.
Maiti,
V.
Gu
nre
dd
y
,
and
P.
Schaumont,
"A
S
y
stematic
Me
t
hod
to
Eva
lu
at
e
and
Com
par
e
t
he
Perform
ance
of
Ph
y
sical
Unc
lon
abl
e
Functi
ons,"
IACR
eP
rint
,
vo
l
.
657
,
2011
.
[
4
3
]
C.
-
C.
Chang
and
C.
-
J.
Li
n,
"LIB
SV
M:
A
Li
bra
r
y
for
support
vec
tor
m
ac
hine
s,"
ACM
Tr
an
s.
Inte
lligent
Syste
ms
and
Technol
ogy
,
vo
l. 2, pp. 27:
1
-
27:2
7,
2011
.
[
4
4
]
M.
Mus
ta
pa,
"P
UF
base
d
FPGAs
for
har
dware
sec
uri
t
y
an
d
trust,
"
Doctor
of
Philosoph
y
,
Univ
ersity
o
f
Tol
edo
,
2015
.
[
4
5
]
J.
Delva
ux
and
I.
Verba
uwhed
e,
"S
ide
cha
nne
l
m
odel
ing
at
t
ac
k
s
on
65nm
arb
it
er
PU
Fs
expl
oit
i
ng
CMO
S
devi
ce
noise,
"
in
Hardware
-
Or
ie
nte
d
Sec
urity
and
Tr
ust
(HO
ST
)
,
2013
IEE
E
Int
ernati
onal
Sym
posium
on
,
20
13,
pp.
137
-
142
.
[
4
6
]
U.
Rührm
ai
r,
F.
Sehnke,
J.
Sö
lt
er
,
G
.
Dror,
S.
Deva
das,
and
J.
Schm
idhube
r,
"M
odel
ing
attac
ks
on
ph
y
si
c
al
uncl
onab
le
fun
c
ti
ons,"
in
Proceedi
ngs
of
th
e
1
7th
ACM
con
fe
r
enc
e
on
Computer
and
communic
ati
ons
se
curity
,
2010,
pp
.
237
-
2
49.
[
4
7
]
G.
Hos
podar
,
R.
Mae
s,
and
I.
Verba
uwhede
,
"M
ac
hine
l
ea
rning
a
t
ta
cks
on
65nm
Arbite
r
PU
Fs
:
A
cc
ura
te
m
odel
in
g
poses
stric
t
bou
nds
on
usabilit
y,
"
in
2012
I
EEE
Int
ernati
onal
Workshop
on
In
formation
Foren
sics
and
Se
curity
(
WIF
S)
,
2012,
p
p.
37
-
42
.
[
4
8
]
U.
Chat
terje
e,
R.
S.
Chakra
bort
y
,
H.
Kapoor,
and
D.
Mukhopadh
y
a
y
,
"The
or
y
a
nd
Applic
a
ti
on
of
Delay
Constrai
n
ts
in
Arbit
er
PU
F,"
ACM
Tr
ans. Embed.
Comput
.
S
y
st.,
vo
l. 15, pp. 1
-
20,
2016
.
BIOGR
AP
HI
ES OF
A
UTH
ORS
Julius
Han
Loong
Te
o
rec
e
ived
the
B.
Eng
in
El
ectrical
and
El
ectroni
cs
en
gine
er
ing
from
Univer
siti
T
ena
g
a
Nasiona
l,
Mal
a
y
si
a
in
2016
and
subs
eque
ntly
t
he
M.E
ng
degr
e
e
in
2018
in
the
sam
e
unive
rsit
y.
He
was
a
r
ese
arc
h
assistan
t
with
the
E
lectr
oni
cs
and
C
om
m
unic
at
ion
Engi
ne
eri
ng
De
par
tment
in
Uni
ver
siti
Te
n
aga
Nasiona
l
.
His
r
ese
arc
h
int
er
ests
in
c
lude
m
emristor
a
pplication
and
I
C
design.
Noor Ali
a
Nor
Hashim
was born
i
n
Kuala L
um
pur
,
Mal
a
y
s
ia i
n
19
86.
She
recei
v
ed
her
B
.
Eng
in
el
e
ct
ri
ca
l
and
e
lectr
oni
cs
engi
n
eering
from
Unive
rsiti
Te
nag
a
Nas
iona
l
,
Mal
a
y
sia
i
n
2009.
She
is
al
so c
urr
ent
l
y
pu
rsuing
the M.E
n
g
degr
e
e in
th
e
s
ame
unive
rsi
t
y
.
She
is
cur
ren
tly
with
the
Elec
tro
nic
s
and
Com
muni
cation
Engi
n
ee
ring
Depa
r
tme
nt
in
Univer
siti
Te
nag
a
Nasiona
l
as
a
rese
ar
ch
a
ss
ista
nt.
Her
res
ea
rch
invol
ves
m
emristors
and
ran
dom
num
ber
gene
ra
tors.
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Ring oscil
lat
or p
hysicall
y
un
c
lonab
le
f
unct
io
n usin
g
se
quen
ti
al ring o
sci
ll
ato
r…
(
J
ulius
Han Lo
ong Te
o
)
901
Azrul
Ghaz
ali
re
ce
iv
ed
the
B.
En
g
in
el
ec
t
rical
e
ngine
er
ing
from
Vande
rbil
t
Uni
ver
sit
y
,
US
A
in
1998
and the M.
Sc
in
Mi
cro
e
le
c
t
ronic
s from
Uni
ver
siti
Keba
ngsa
an
Mal
a
y
s
ia,
Ma
lay
s
ia i
n
2003
.
He
is
cur
r
ent
l
y
a
senior
le
c
ture
r
i
n
the
Elec
tron
ic
s
and
Com
m
unicati
on
Engi
n
ee
rin
g
Depa
rtment
in
Univer
siti
T
e
naga
Nasiona
l
,
Malay
s
ia.
His
rese
ac
h
in
te
r
ests
inc
lude
IC
desi
gn,
VLSI,
and
m
ic
roe
lectr
oni
cs
.
Fazr
ena
Azlee
Ham
id
rec
e
ive
d
the
B
.
T
ec
h
d
ip
loma
in
eng
inee
ring
from
Cove
ntr
y
Techni
ca
l
Coll
ege,
UK
in
1996,
foll
owed
with
the
B
.
Eng
a
nd
Ph.D
degr
ee
s
in
elec
tron
ic
s
e
ngine
er
ing
from
Univer
sit
y
of
Southampton,
UK
in
1999
and
2004,
respe
c
ti
ve
l
y
.
She
is
working
as
a
senior
le
c
ture
r
wi
th
th
e
El
e
ct
ron
ic
s
and
Com
m
unic
at
ion
Engi
ne
eri
ng
De
par
tment
in
Uni
ver
siti
Te
n
aga
Nasiona
l,
Ma
lays
ia
.
Her
rese
arch
is
cur
ren
tly
fu
nded
b
y
th
e
Min
istr
y
of
Highe
r
Educ
a
ti
on.
Her
rese
arc
h
intere
st
s
inc
lude
IC
d
esign
and
opt
i
m
iz
at
ion
as
wel
l
as
m
emristor
m
odel
li
ng
and
appl
i
ca
t
ions f
or hardware
sec
uri
t
y
.
Evaluation Warning : The document was created with Spire.PDF for Python.