ip protection for fir filter fpga implementation ersity of ...ip protection for fir filter fpga...
TRANSCRIPT
1
IP P
rote
ctio
n fo
r FIR
Filt
er
FPG
A Im
plem
enta
tion
Wei
Dai
Dep
t of E
CE,
Uni
vers
ity o
f Win
dsor
Supe
rvis
or :
Dr.H
.K.K
wan
& D
r H.W
uM
ay.2
004
2
Out
line
1.In
trodu
ctio
n to
Dig
ital W
ater
mar
king
Te
chno
logy
for I
P Pr
otec
tion
2.R
evie
w o
f IP
Prot
ectio
n Te
chno
logi
es fo
r AS
IC a
nd F
PGA
Des
ign
3.R
evie
w o
f Wat
erm
arki
ng S
chem
es fo
r Fi
lter D
esig
n 4.
Prop
osed
New
Wat
erm
arki
ng M
etho
ds fo
r FP
GA
Impl
emen
tatio
n of
FIR
Filt
er
3
1. In
trod
uctio
n
�W
hat i
s th
e si
gnifi
canc
e of
IP p
rote
ctio
n?�
Wha
t is
a di
gita
l wat
erm
ark?
�W
hat i
s w
ater
mar
king
tech
nolo
gy?
�H
ow a
dig
ital w
ater
mar
k is
em
bedd
ed a
nd
dete
cted
?
4
1.1.
Sign
ifica
nce
of IP
Pro
tect
ion
�C
opyr
ight
pro
tect
ion
is n
eede
d fo
r va
rious
di
gita
l pro
duct
s�
Dig
ital s
till i
mag
es�
Dig
ital a
udio
pro
duct
s (i.
e., m
usic
CD
)�
Dig
ital v
ideo
pro
duct
s (i.
e., D
VD m
ovie
s)�
Softw
are
sour
ce c
ode
�AS
IC a
nd F
PGA
IP c
ores
�El
ectro
nic
publ
icat
ions
5
1.2.
Wha
t is
a Di
gita
l Wat
erm
ark
and
Wat
erm
arki
ng T
echn
olog
y?
�A
digi
tal
wat
erm
ark
is a
pie
ce o
f in
form
atio
n hi
dden
in
side
th
e IP
th
at
show
s IP
ow
ner�s
id
entit
y.
�W
ater
mar
king
re
fers
to
th
e te
chno
logy
to
em
bed
a di
gita
l wat
erm
ark
into
a d
igita
l IP
and
to
dete
ct it
from
a d
igita
l IP
with
wat
erm
ark.
6
Info
rmat
ion
Hid
ing
Cov
ert
Cha
nnel
sSt
egan
ogra
phy
Anon
ymity
Cop
yrig
ht m
arki
ng
Ling
ustic
st
egan
ogra
phy
Tech
nica
l st
egan
ogra
phy
Rob
ust
copy
right
mar
king
Frag
ile
wat
erm
arki
ng
Fing
erpr
intin
gW
ater
mar
king
Impe
rcep
tible
w
ater
mar
king
Visi
ble
wat
erm
arki
ng
Info
rmat
ion
Hid
ing
and
Wat
erm
arki
ng
7
�C
riter
ia fo
r a w
ater
mar
king
tech
nolo
gy�
Easy
to e
mbe
d an
d de
tect
a w
ater
mar
k�
Diff
icul
t to
rem
ove
the
wat
erm
ark
�Lo
w d
esig
n an
d im
plem
enta
tion
over
head
or c
ost
�W
ater
mar
king
tech
nolo
gy h
as a
pplic
atio
ns fo
r �
Cop
yrig
ht p
rote
ctio
n�
Fing
erpr
intin
g�
Dat
a au
then
ticat
ion
�D
ata
hidi
ng
8
1.3.
Wat
erm
ark
Embe
ddin
g an
d D
etec
tion:
Sche
me
for E
mbe
ddin
g a
Wat
erm
ark
Embe
ddin
gal
gorit
hm
Wat
erm
ark
Key
Orig
inal
dig
ital I
PM
arke
ddi
gita
l IP
9
Sche
me
for W
ater
mar
kD
etec
tion
Det
ectio
n al
gorit
hm
Mar
k an
d/or
orig
inal
dig
ital I
P
Key
Test
-dig
ital I
P
Wat
erm
ark
or
conf
iden
ce m
easu
re
11
2.1.
Non
-wat
erm
arki
ng m
etho
ds fo
r
ASIC
and
FPG
A de
sign
IP p
rote
ctio
n:
�IP
ow
ners
try
to p
rote
ct th
eir c
opyr
ight
by
encr
yptin
gVH
DL/
Veril
og s
ourc
e co
des
befo
re
send
ing
them
to th
e us
ers.
�
The
encr
ypte
d co
des
are
load
ed in
to
auth
oriz
ed s
imul
ator
sor
syn
thes
is to
ols.
�Th
e so
urce
cod
es a
re in
visi
ble
to th
e sy
stem
de
sign
er w
ho u
ses
the
IP b
lock
s.
12
Non
-wat
erm
arki
ng m
etho
ds a
re n
otse
cure
en
ough
:�
The
CAD
to
olm
aint
ains
the
saf
ety
of t
he
copy
right
. �
In
prac
tice,
th
is
can
ofte
n be
bro
ken
by
atta
ckin
g th
e C
AD t
ool (
sim
ulat
ors,
syn
thes
is
tool
) dire
ctly
. �
Bette
r m
etho
ds a
re n
eede
d fo
r pr
otec
tion
of
the
ASIC
and
FPG
A IP
.
13
2.2.
Wat
erm
arki
ng m
etho
ds fo
r VLS
I/FPG
A
�Ad
vant
ages
of w
ater
mar
king
met
hods
�VL
SI/F
PGA
IP w
ater
mar
king
can
hel
p de
ter
thef
tand
cou
nter
feiti
ng.
�Th
e em
bedd
ed w
ater
mar
k se
rves
as
evid
ence
of
ow
ners
hip.
�A
wat
erm
arki
ng
met
hod
incl
udes
tw
o ph
ases
�W
ater
mar
k sy
nthe
sis
�W
ater
mar
k de
tect
ion
14
FPG
A an
d AS
IC W
ater
mar
king
Sch
emes
1.Fi
nger
prin
ting
for F
PGA
IP p
rote
ctio
n�
This
met
hod
uses
the
unus
ed L
UT
(look
-up
tabl
e) b
its to
em
bed
the
sign
atur
e bi
ts.
2.C
hang
e of
FPG
A bi
t-str
eam
dat
a�
The
met
hod
subs
titut
es w
ater
mar
k bi
ts f
or
som
e of
the
bits
in
the
conf
igur
atio
n bi
t-st
ream
th
at
cont
rols
m
ultip
lexe
rs
for
the
unus
ed C
LB o
utpu
ts.
15
3.
Hie
rarc
hica
l wat
erm
arki
ng�
It us
es a
n un
ique
map
ping
of
topo
logi
cal
info
rmat
ion
onto
a s
eque
nce
of s
ymbo
ls.
4.
Prot
ocol
s fo
r IP
prot
ectio
n�
Hid
e w
ater
mar
k da
ta a
t th
e co
mbi
natio
nal
logi
c sy
nthe
sis
leve
l.
5.
Fing
er-m
arki
ng�
Hid
e w
ater
mar
k at
th
e la
yout
le
vel
by
mod
ifyin
g th
e tra
nsis
tor W
/L a
nd th
e nu
mbe
r of
fing
ers.
16
3. R
evie
w o
f Filt
er D
esig
n W
ater
mar
king
�Th
ree
tech
nolo
gies
hav
e be
en p
ropo
sed:
�M
agni
tude
mod
ifica
tion
�Fi
lter t
ap�s
equ
al-re
plac
emen
t�
Win
dow
ing
func
tion
wat
erm
arki
ng
17
3.1.
Mag
nitu
de M
odifi
catio
n
�St
ep 1
: Pre
pare
wat
erm
ark
code
(i.e
., 7-
bits
)�
Step
2: S
epar
ate
the
filte
r sto
p or
pas
s-ba
nd
to s
ever
al e
qual
wid
th z
ones
(i.e
., se
ven
zo
nes)
�St
ep 3
: Mod
ify th
e fil
ter m
agni
tude
resp
onse
ac
cord
ing
to th
e w
ater
mar
k b
its:
�If
the
bit i
s 1,
decr
ease
the
filte
r mag
nitu
de
resp
onse
by
1 dB
.�
If th
e bi
t is
0,in
crea
seth
e fil
ter m
agni
tude
re
spon
se b
y 1
dB
.
18
3.1.
Mag
nitu
de M
odifi
catio
n (C
on�t)
�St
ep 4
: Use
the
mod
ified
filte
r mag
nitu
de re
spon
se a
s th
e de
sign
con
stra
ins
inpu
t to
the
desi
gn to
ol.
�St
ep 5
: Obt
ain
the
filte
r coe
ffici
ents
.
20
Adva
ntag
es o
f thi
s m
etho
d:�
Add
wat
erm
ark
at a
lgor
ithm
leve
l, th
e hi
ghes
t lev
el
for f
ilter
des
ign
�H
ard
to re
mov
e fro
m lo
wer
leve
l of f
ilter
�s
impl
emen
tatio
n, li
ke lo
gic,
layo
ut o
r circ
uit l
evel
Dis
adva
ntag
es o
f thi
s m
etho
d:�
Incr
ease
the
com
plex
ityof
filte
r des
ign.
�In
crea
se th
e or
der o
f filt
er, s
ince
we
intro
duce
new
rip
ple
cons
train
to th
e fil
ter m
agni
tude
resp
onse
�In
crea
se th
e ha
rdw
are
cost
by +
7%
21
3.2.
Filte
r wat
erm
arki
ng b
y fil
ter t
ap�s
equ
al-
repl
acem
ent
�Pr
oced
ure:
�St
ep 1
: Pre
pare
the
wat
erm
ark
(7-b
its).
�St
ep 2
: Des
ign
the
filte
r with
the
orig
inal
pe
rform
ance
spe
cific
atio
n.�
Step
3: R
epla
ce th
e fil
ter t
aps
by u
sing
equ
al
filte
r stru
ctur
e re
plac
emen
t.�
Ther
e ar
e th
ree
equa
l fun
ctio
n fil
ter s
truct
ures
, A, B
and
C.
�W
hen
the
wat
erm
arki
ng b
it is
0,us
e B
to im
plem
ent
this
tap.
�W
hen
the
wat
erm
arki
ng b
it is
1,u
se C
to im
plem
ent
this
tap.
23
Adva
ntag
es o
f thi
s m
etho
d:�
Wat
erm
arki
ng a
t alg
orith
mle
vel,
hard
to re
mov
e at
the
low
er le
vels
�N
o de
grad
atio
n of
filte
r per
form
ance
Dis
adva
ntag
es o
f thi
s m
etho
d:�
Incr
ease
har
dwar
e co
st d
ram
atic
ally
by
+33%
�M
ake
the
filte
r stru
ctur
e no
t uni
form
, and
in
crea
se th
e de
sign
tim
e fo
r im
plem
enta
tion
w
ith A
SIC
or F
PGA
24
3.3.
Win
dow
ing
func
tion
wat
erm
arki
ng
�Pr
oced
ure:
�St
ep 1
:�
Supp
ose
W(n
) is
the
orig
inal
win
dow
func
tion,
w
here
1<=
n <
=N.
�Ad
d ra
ndom
noi
se to
W(n
) to
obta
in W
m(n
)W
m(n
)=W
(n)+
a * r
(n),1
<=n<
=N,
�Se
lect
a a
s 0.
001
�r(n
) is
a ra
ndom
seq
uenc
e w
ith z
ero
mea
n
25
�St
ep 2
: Wm
(n),
1<=
n <
= i-1
Wc(
n) =
W
m(n
)+b*
c(n-
i+1)
, i
<= n
<=i
+P-1
Wm
(n),
i+
P<=
n <=
N/2
Wc(
N+1
-n),
N
/2+1
<= n
<=N
.
�Se
lect
b=0
.000
1.�
The
star
ting
bit o
f the
P-b
it w
ater
mar
k co
de
sequ
ence
c(n
-i+1)
is b
it i o
f Wm
(n).
�Th
e se
quen
ce W
c(n)
, n=1
,�,N
, is
the
new
w
indo
w fu
nctio
n w
hich
con
tain
s th
e w
ater
mar
k in
form
atio
n.
26
�Ad
vant
ages
of t
his
met
hod:
�Em
bedd
ing
wat
erm
ark
at a
lgor
ithm
leve
l whi
ch is
ha
rd to
rem
ove
at a
low
er le
ver
�Si
mpl
e an
d di
rect
wat
erm
ark
embe
ddin
g sc
hem
e
�D
isad
vant
ages
of t
his
met
hod:
�In
crea
se d
esig
n co
mpl
exity
27
4. N
ew P
ropo
sals
for F
IR F
ilter
W
ater
mar
king
with
FPG
A Im
plem
enta
tion
�Pr
opos
al o
ne:
�Em
bedd
ing
wat
erm
ark
at F
IR fi
lter
coef
ficie
nts�
LSB
�Pr
opos
al tw
o:
�FP
GA
RAM
cel
l loc
atio
ns� w
ater
mar
king
28
4.1.
Prop
osal
One
: FI
R F
ilter
Coe
ffici
ents
� LSB
Wat
erm
arki
ng
�Le
t a F
IR fi
lter d
esig
n be
giv
en b
y
Y(k)
=A0*
x(k)
+A1*
x(k-
1)+�
�+A
n-1*
x(k-
N-1
),w
here
k=0
,1,�
�,N
-1.
A0,A
1,�
�An
-1 a
re fi
lter c
oeffi
cien
ts.
�A
wat
erm
ark,
for i
nsta
nce,
is g
iven
by
1000
1010
�W
ater
mar
k em
bedd
ing
proc
ess
is to
repl
ace
the
filte
r coe
ffici
ents
� LSB
s w
ith th
e w
ater
mar
k bi
ts.
29
011
11-1
110-
0101
-011
0A7
:111
1-11
10-0
101-
0110
100
00-0
001-
1111
-101
1A6
:000
0-00
01-1
111-
1010
000
00-1
000-
1101
-111
0A5
:000
0-10
00-1
101-
1111
100
00-0
011-
0111
-100
1A4
:000
0-00
11-0
111-
1000
000
00-0
000-
1110
-101
0A3
:000
0-00
00-1
110-
1010
011
11-1
111-
1010
-100
0A2
:111
1-11
11-1
010-
1001
011
11-1
111-
0100
-110
0A1
:111
1-11
11-0
100-
1101
111
11-1
111-
1100
-010
1A0
:111
1-11
11-1
100-
0100
Wat
erm
ark
bits
Mod
ified
Filt
er
Coe
ffici
ents
Orig
inal
Filt
er C
oeffi
cien
ts
30
Mag
nitu
de re
spon
se s
imul
atio
n re
sults
:FI
R fi
lter c
oeffi
cien
ts� L
SB w
ater
mar
king
(left:
bef
ore
wat
erm
arki
ng
right
: afte
r wat
erm
arki
ng)
00.
10.
20.
30.
40.
50.
60.
70.
80.
91
-90
-80
-70
-60
-50
-40
-30
-20
-10010
Norm
alize
d Fr
eque
ncy
( ×π r
ad/s
ample
)
Magnitude (dB)
Magn
itude
Res
pons
e
00.
10.
20.
30.
40.5
0.6
0.7
0.80.
91
-90
-80
-70
-60
-50
-40
-30
-20
-10010
Norm
alize
d Fre
quen
cy (×π
rad/
samp
le)
Magnitude (dB)
Magn
itude
Res
pons
e
Nor
maliz
ed Fr
eque
ncy:
0.0
0391
M
agnit
ude
(dB)
: 4.5
1
Nor
maliz
ed F
requ
ency
: 0.58
8 M
agnit
ude (
dB):
-29.
4 N
orma
lized
Freq
uenc
y: 0.
781
Mag
nitud
e (dB
): -2
9.3
31
4.2.
Prop
osal
Tw
o:
FPG
A R
AM C
ell L
ocat
ions
� Wat
erm
arki
ng
�Le
t a F
IR fi
lter d
esig
n be
giv
en b
yY(
k)=A
0*X(
k)+A
1*X(
k-1)
+��
+An-
1*X(
k-N
-1),
whe
re k
=0,1
,�.N
-1.
�Su
ppos
e a
wat
erm
ark
cons
ists
of 8
bits
as
10
0010
10.
�Le
t the
RAM
cel
l to
stor
e A
i be
RAM
_i,
i=0,
1,�
,7.
�A
RAM
cel
l on
FPG
A ci
rcui
ts c
an b
e id
entif
ied
by it
s co
ordi
nate
s (a
, b).
�W
ater
mar
k bi
ts e
mbe
ddin
g ru
le:
�If
the
bits
are
00,
we
choo
se b
oth
a an
d b
as e
ven.
�If
the
bits
are
01,
we
choo
se a
as
even
and
b as
odd
.�
If th
e bi
ts a
re 1
0, w
e ch
oose
a a
s od
dan
d b
as e
ven.
�If
the
bits
are
11,
we
choo
se b
oth
a an
d b
as o
dd.
33
0(3
,0)
1R
AM_3
0(3
,2)
1R
AM_2
0(4
,4)
0R
AM_1
0(1
, 4)
1R
AM_0
New
Lo
catio
n of
th
e ch
osen
R
AM c
ell
Wat
erm
arki
ng
bits
RAM
_i fo
r st
orin
g A_
i
34
FIR
filte
r pla
ce a
nd ro
ute
resu
lts:
left:
befo
re w
ater
mar
king
right
:afte
r wat
erm
arki
ng
489/
1536
106.
474
MHz
Afte
r wat
erm
arki
ng
488/
1536
107.
654
MHz
Befo
re w
ater
mar
king
Har
dwar
e co
st
(slic
e)M
ax F
requ
ency
35
Adva
ntag
es o
f the
pro
pose
d m
etho
ds:
�Si
mpl
e w
ater
mar
k em
bedd
ing
and
extra
ctio
n�
Wat
erm
ark
is in
visi
ble
to th
e po
tent
ial
atta
cker
.�
Embe
ddin
g w
ater
mar
k at
bot
h al
gorit
hm a
nd
layo
ut le
vel
�In
crea
sing
sec
ure
stre
ngth
and
the
time
for
reve
rse
engi
neer
ing
Dis
adva
ntag
es o
f the
pro
pose
d m
etho
ds:
�Fo
r the
firs
t pro
pose
d m
etho
d, fi
lter
perfo
rman
ce m
ay s
light
ly w
orse
n du
e to
the
mod
ifica
tion
of fi
lter c
oeffi
cien
ts� L
SB.
�Fo
r the
sec
ond
prop
osal
, max
freq
uenc
y an
d ha
rdw
are
usag
e m
ay b
e sl
ight
ly w
orse
than
th
e or
igin
al d
esig
n.
36
Com
pari
son
of P
ropo
sal O
ne w
ith C
urre
nt M
etho
ds(a
t alg
orith
m le
vel)
Mag
nitu
de
Mod
ifica
tion
Tap
's
Equ
al
Rep
lace
men
tW
indo
win
g Fu
nctio
n W
ater
mar
king
Prop
osal
One
(FIR
Fi
lter
LSB
W
ater
mar
king
)
Filte
r Pe
rfor
man
ce
Deg
rada
tion
Med
ium
Smal
lSm
all
Smal
l
Har
dwar
e U
sage
Incr
ease
+7%
+29%
N/A
0%
Des
ign
Ove
rhea
dH
igh
Med
ium
Low
Low
Ext
ract
ion
Cos
tL
owM
ediu
mL
owM
ediu
m
Prob
abili
ty
of
Coi
ncid
ence
Low
Low
Low
Low
Secu
rity
Med
ium
Med
ium
Low
Med
ium
37
Usi
ng
Spar
e LU
T
Bit-
stre
amM
odifi
ca-
tion
Hie
rarc
hica
l W
ater
mar
k-in
g
Wat
erm
arki
ng
by
Usi
ng
Prot
ocol
s
Fing
er-m
ark-
ing
Prop
osal
Tw
o
Add
Lev
elLa
yout
Layo
utR
TLR
TLLa
yout
Layo
ut
Embe
ddin
g C
ost
Med
ium
Med
ium
Med
ium
Med
ium
Med
Med
Des
ign
Ove
rhea
dM
ediu
mM
ediu
mM
ediu
mM
ediu
mH
igh
Low
Extra
ctio
n C
ost
Med
ium
Hig
hH
igh
Hig
hLo
wM
ed
Prob
abili
ty
of
Coi
ncid
ence
Low
Low
Low
Low
Med
ium
Low
Secu
rity
Low
Hig
hH
igh
Med
ium
Low
Hig
h
App
lied
Are
aFP
GA
FP
GA
ASI
C
(dig
ital)
ASI
C
(d
igita
l)A
SIC
(m
ix-
sign
al)
FPG
A
Com
pari
son
of P
ropo
sal T
wo
with
Cur
rent
Met
hods
(at p
hysi
cal/R
TL
des
ign
leve
l)