TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol. 15, No. 2, August 201
5, pp. 381 ~
389
DOI: 10.115
9
1
/telkomni
ka.
v
15i2.837
5
381
Re
cei
v
ed Ap
ril 17, 2015; Revi
sed
Jun
e
26, 2015; Accepted July 1
3
,
2015
Security for Mobile Agents: Trust Estimate for
Platforms
Raz
ouki Hassan*, Hair Abdellatif
Lab
orator
y of
Mode
lin
g an
d Co
mp
utation (
L
MC), F
a
cult
y
of
Science a
n
d
T
e
chnol
og
y,
Univers
i
t
y
S
u
lt
an Mou
l
a
y
Sli
m
ane, Ben
i
Me
llal, Morocc
o
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: razouki.h
a
ss
an@
gmai
l.com
A
b
st
r
a
ct
T
he mobi
le
ag
ent has
bee
n s
een
as a pr
o
m
ising
di
strib
u
te
d co
mp
uting te
chno
logy. T
he
mo
bil
i
ty
character
i
stic o
f
mobi
le a
g
e
n
t mak
e
s it to travel oft
en in
op
en netw
o
rk. In this scenar
io, i
t
is obvious th
at
the
mo
bil
e
ag
e
n
ts are
vu
lner
a
b
le
to v
a
rio
u
s
security
thr
eat
s. Protecting
fr
ee-ro
a
m
in
g
mobil
e
ag
ents fr
om
ma
lici
ous
host
and fro
m
other
mo
bi
le a
g
e
n
ts has dr
aw
n
mu
ch attenti
on i
n
recent ye
ars. T
he pr
otectio
n
o
f
mo
bil
e
ag
ents
is consid
ere
d
as one
of th
e greatest ch
alle
ng
es of se
curity, becaus
e the platfor
m
of
executi
on h
a
s
access to al
l
the co
mpo
n
e
n
ts of t
he mo
bile
ag
ent. In this pap
er, w
e
prese
n
t a n
e
w
architectur
e
pa
radi
gm of
mob
ile
ag
ents, w
h
ich allow
s
the separ
ation
of the i
m
pl
e
m
e
n
ta
tion tasks of th
e
age
nt an
d its s
e
curity
mec
h
a
n
is
ms. Our ap
proac
h is b
a
se
d on
usin
g tw
o strategies
of a
daptati
on to
ad
apt
the
mo
bil
e
a
g
e
n
t security
at r
unti
m
e, d
e
p
e
n
d
in
g o
n
th
e
se
nsitivity of th
e
services r
e
q
u
ir
ed to
perfor
m
t
h
e
duties of the a
gent an
d t
he d
egre
e
of confid
ence of
the vis
i
ted platfor
m
s.
Ke
y
w
ords
:
mobil
e
a
g
e
n
t, multi-a
gent syst
ems, s
e
curity,
cryptogra
phy,
softw
are co
mpon
ents, dyn
a
mic
ada
ptatio
n.
Copy
right
©
2015 In
stitu
t
e o
f
Ad
van
ced
En
g
i
n
eerin
g and
Scien
ce. All
rig
h
t
s reser
ve
d
.
1. Introduc
tion
The m
obile
a
gent, a
s
a ty
pical
dist
ribut
ed
comp
utin
g technolo
g
y, ha
s attra
c
te
d many
resea
r
chers’
attention a
nd
develop
ed q
u
i
ckly to
sa
tisf
y many di
stri
buted
appli
c
a
t
ions. Th
e a
b
i
lity
of mobile
age
nts to auto
n
o
m
ously mi
gra
t
e from ho
st t
o
ho
st tran
sf
erri
ng thei
r
code a
nd inte
rnal
state en
able
s
them to a
ccomplish ta
sks in di
st
ribute
d
enviro
n
ment
more
ro
bu
stly and effici
en
tly
than traditio
n
a
l app
roa
c
h
e
s
. De
spite of
significant b
enefits, security prese
n
ts
a cruci
a
l poin
t
in
mobile ag
ent system
s and
may hinde
r the
expan
sion
and u
s
e of this paradig
m
[1-2].
The inte
ra
ction of mo
bile
agent
with its platform
will
bring
so
me
secu
rity pro
b
le
ms [3].
Four th
reat categori
e
s a
r
e
identified:
1)
Agent attacks against an e
x
ecution platform.
2)
Platform attacks against a mobile agent.
3)
Agent attacks against another agent in
the same execution platfor
m
.
4)
Exterior entity attacks against an agent or a platform.
These attacks are prim
aril
y focused on the co
mm
unication
capability of the platform to
exploit potent
ial vulnerabilities [4].
The se
cu
rity of mobile agent ha
s b
e
co
m
e
an seriou
s probl
em that needs dee
p
con
s
id
eratio
n
[5], becau
se
the platform of exec
ution
has acce
ss to all the compone
nts of the
mobile agent. Since, a platform
has the responsibility
to execute a mobile agent; it is assumed
that the platform must have
full acce
ss t
o
agent'
s
co
d
e
and data. A platform mig
h
t be malicio
us
and m
a
y try to execute the
cod
e
in
a ma
nner in
wh
i
c
h
it is not a
u
th
orized to
do.
A platform m
a
y
try to change
agent'
s
state,
code o
r
ro
uting duri
ng ag
e
n
ts' exe
c
ution
[6].
The Pro
p
o
s
e
d
app
roa
c
h i
s
to find a ne
w mobile a
gent
para
d
igm a
r
chite
c
ture,
which
ca
n
prote
c
t the mobile ag
ent
via two stra
tegies of
ada
ptation. It takes into acco
unt the dynamic
asp
e
ct
s of the se
curity ne
eds of a mobi
le agent in ea
ch ru
ntime en
vironme
n
t.
The
first
i
s
a static ada
ptation
p
e
rfo
r
me
d
by th
e MSA
S
(Ma
nag
em
ent System
o
f
Agents
Security)based on the
sensit
ivity of the services reque
sted
by the agent,
theMSAS adds
addition
al se
curity co
mpo
nents an
d d
e
termin
es
th
e policy of the dynami
c
adaptatio
n to be
followe
d by the mobile ag
e
n
t during its e
x
ecution.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 15, No. 2, August 2015 : 381 –
389
382
The se
con
d
i
s
a reflexive dynamic
st
ru
ctural
ad
aptat
ion pe
rforme
d by the m
o
b
ile age
nt
itself. Acco
rdi
ng to the de
gree
of confi
den
ce on
th
e
platform visit
ed, the mobil
e
agent
sele
cts
and ad
apts
secu
rity comp
onent
s to the tasks to be p
e
rform
ed by this platform.
This pap
er is organi
zed
a
s
follo
ws: in
se
cti
on
2
we
begin
by id
en
tifying the sta
t
e of a
r
t
and p
r
e
s
e
n
t the differe
nt
approa
che
s
t
o
protect
m
o
bile ag
ents.
Section
3 p
r
ese
n
ts th
e n
e
w
architectu
re
of our syste
m
, identifies the
functio
n
s of the variou
s comp
onent
s and
trust
estimation. Fi
nally a con
c
lu
sion i
s
pre
s
e
n
ted in se
ctio
n 4.
2.
The Security
of Mobile Agents
2.1.
Securit
y
Iss
u
es Rela
ted
to Mobile Ag
ents
Thre
e types o
f
problem
s a
r
i
s
e
with re
gard to se
cu
rity in the co
ncept
of mobile ag
ent: the
se
curity
of th
e ag
ent mi
gration,
the
protection
of th
e platfo
rm
a
gain
s
t ag
ent
s a
n
d
mali
ci
ous
platform
s, an
d the prote
c
tion of the age
nt agai
n
s
t other age
nts an
d
maliciou
s
pla
tforms [7].
2.1.1. Malicious
Agent
It is a pro
g
ra
m or
softwa
r
e that functio
n
s fo
r an
int
r
uder with
a p
u
rpo
s
e
to attacking
th
e
itinera
r
y host
.
It can pose
a severe ri
sk to the host
s
se
rving a p
l
atform to the mobile age
n
t.
Hen
c
e
the
s
e
ho
sts
are
vulnerable
to
attack by
su
ch age
nts executin
g
o
n
t
hem.
Th
ere
are
several kin
d
s of such ag
en
ts and the
s
e can b
e
virus,
worms o
r
spy
i
ng agent
s. The prote
c
tion
of
t
he v
i
sit
e
d
h
o
st
s
ag
ain
s
t
at
t
a
ck
s
ca
rri
e
d
by
m
a
lic
i
o
u
s
m
obile
age
nts i
s
a
p
r
obl
em that i
s
no
w
fairly well controlled.
2.1.2. Malicious
Host
Whe
n
a
mobi
le ag
ent mov
e
s from
one
desi
gnate
d
h
o
st to
anoth
e
r
de
sig
nated
host i
n
the net
work t
o
complete
its ta
sk, th
ere
exists
so
me
desi
gnate
d
h
o
sts who t
r
ie
s to
ham
per
the
integrity, functionality and
confid
entiality of the agent
in ord
e
r to
benefit them
selve
s
in so
me
way. For exa
m
ple they ca
n cha
nge the
agent code
with an intent
ion to get so
me task don
e
on
its behalf a
s
agent move
s to other ho
st
s to harm
the
reputatio
n of the agent o
w
ne
r or to h
a
rm
the other h
o
sts in som
e
way. Also a malicio
us
h
o
st
can try to acce
ss
som
e
u
nautho
rized
data
belon
gs to
a
n
age
nt thus attacks o
n
its confide
n
ti
al
ity. It is more difficult to
prote
c
t a mo
bile
agent from a
malicio
us h
o
st rather than
prot
e
c
ting a h
o
st from a ma
liciou
s
mobil
e
agent.
2.2.
Appro
ach
es
to the Pro
t
ec
tion of th
e M
obile Agen
ts
Several ap
proache
s for th
e prote
c
tion o
f
m
obile age
n
t
have been
prop
osed. Th
ey try to
ensure
the
acce
ss of the
m
obile
a
gent to
ho
sts in
which it may have
co
nfiden
ce
o
r
dete
c
t tho
s
e
that are mali
ciou
s. T
hey a
r
e
prima
r
ily i
n
tende
d to
d
e
tect
atta
cks or ren
der
the
m
ineffective, our
prop
osal is b
a
se
d on prote
c
tion.
Xinwen
Zha
n
g
[8] present
s a
mobile
p
o
licy fra
m
ework to p
r
ote
c
t the inform
a
t
ion and
resou
r
ces im
ported by mo
bile co
de an
d
agents in
ru
ntime
enviro
n
m
ents with
trusted comp
uting
techn
o
logie
s
.
This fram
ewo
r
k in
clud
es
poli
c
y speci
fi
cation
and de
fi
nitio
n
, as well as
impleme
n
tation a
r
chitectu
re in
Java.
Th
e be
ne
fi
t of t
h
is
enforce
m
ent a
r
chitectu
re i
s
th
at he
can
de
fi
ne an
d impleme
n
t the permi
ssion
class in a
mobile poli
c
y
,
maintaining
the
fl
exibility and
compatibility with current
runtime technologies. However, the
downfall is that the system uses
‘Tru
sted
Co
mputing
Devi
ce
s (TCD)’
a
nd d
epen
ds
on ‘T
ru
sted
Runtime
Env
i
ronm
ent (TRE)’.
Therefore, for this system t
o
work, TCD
and TRE a
r
e
mandato
r
y.
Ibharal
u et al. in [9] have propo
se
d the
use of
a
chai
n of digital en
velope
s with
platform
regi
strie
s
to
sup
port
dyna
mic a
gent’s i
t
iner
a
r
ie
s in
open
network envi
r
onm
e
n
t. This
sche
me
prote
c
ts
and
allows m
obile
age
nts to
ro
am fr
e
e
ly in
open
networks e
n
viron
m
en
t without
bein
g
c
o
mpr
o
mis
ed in
a ma
lic
i
ous
ho
s
t
s
.
T
h
e
ma
in
ad
vanta
ge i
s
that
the
propo
sed
scheme
exhibit
e
d
better perfo
rmance wh
en comp
ared to the results
o
b
tained from
obfuscatio
n
method
s in term
s
of data integri
t
y and se
curit
y
. The main d
r
awba
ck
i
s
th
at the prop
osed sche
me consume
s
a little
more time
visiting pl
atform regi
strie
s
and exe
c
utin
g co
mplex cryptograp
hic f
unctio
n
s th
an
the
obfuscatio
n
method
s. Th
ough th
e dat
a is p
r
ote
c
ted
fr
om ho
st
s, the code i
s
vu
lnera
b
le to
attack
by other mali
ciou
s ag
ents
resi
ding in th
e host.
Nisha
et al. p
r
esented
a
secu
rity sol
u
tion [10]
th
at p
r
otect
s
b
o
th t
he mo
bile
ag
ent itself
and th
e ho
st
resou
r
ces tha
t
encrypt the
data b
e
fore
p
a
ssing
it to m
obile a
gent
a
nd d
e
crypt it
on
the visited
h
o
st
side
s. T
h
e metho
d
of
"comp
u
ting
with en
cryption
functio
n
" h
a
s b
een
u
s
ed.
It
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Sec
u
rity
for
Mobile Agents
: Trus
t Es
ti
mate for Platforms
(Raz
ouk
i
Has
s
an)
383
solve
s
the problem of mali
ciou
s ho
st that can ha
rm
mobile ag
ent
or the inform
ation it contai
ns.
Here also, the itinera
r
ie
s are e
n
crypte
d well in
adv
ance indi
cati
ng the usage
of static itinerary.
No provisio
n is given to se
cure
code fro
m
other mali
cious a
gent
s.
Shibli et
al. [11-1
2
] Propo
sed
a
se
cu
re
syst
e
m
fo
r d
eployment
of mobile
ag
en
ts. Th
e
system provi
des method
ol
ogy
that
sp
a
n
s a
num
ber
of pha
se
s in
agent
’
s
lifetime: it start
s
from
agent cre
a
tion and
end
s with a
gen
t’s executio
n
.
It addresses
classification, validatio
n,
publi
s
hing,
di
scovery, a
d
o
p
tion, auth
e
n
t
ication
and
authori
z
atio
n
of a
gent
s.
This sy
stem
is
based
on se
cure web se
rvice
s
and uses
RBAC
XACML poli
c
ies
a
nd SAML pro
t
ocol.Tho
ugh
the
work auth
enti
c
ate
s
the co
de. The integrity of
the code is not a
s
sure
d, and use
s
asymm
e
tric
encryption re
peatedly in
creases the ex
ecutio
n time of the mobile agent.
Leri
c
he
and
Arca
ngeli [1
3
]
propo
se
d a
model of
the
mobile a
gent
that is self-a
daptive,
by assem
b
lin
g re
usable
compon
ents.
The a
gent
s
a
r
e
config
ured
and
may b
e
re
config
ured
at
runtime
so th
at they are ab
le to resp
ond
to chan
ge
s in
the executio
n environ
men
t
.
The an
alysi
s
, at the end
of this stud
y, showe
d
th
at all approa
che
s
u
s
e th
e sam
e
se
curity m
e
chani
sm to
protect mo
bile
age
nts a
gai
nst different
malicio
us pla
tforms,
witho
u
t
taking
into a
c
count
either the security
req
u
ir
e
m
ent
s of e
a
ch a
gent (th
e
se
nsitivity of the
informatio
n containe
d in the mobile ag
e
n
t) and t
he
credibility of platforms. Inde
ed, each mo
bile
agent req
u
ires different se
curity
m
e
chani
sms
ba
sed o
n
thei
r
servi
c
e
s
a
n
d
ba
sed
on
the
credibility of each visited pl
atform.
3.
A Ne
w
P
e
rsp
ectiv
e of Securit
y
Each m
obile
agent requi
res
differe
nt se
cu
rity me
chani
sms
ba
sed on thei
r
servi
c
e
s
andb
ased o
n
the cre
d
ibilit
y of each visi
ted platfo
rm. These se
cu
rity
mech
ani
sm
s
a
r
e stored
i
n
the MSAS (the Management System
of Agents
Security)
as security compon
ent
s (see Figure
1).
The MSAS
m
u
st
support and update all
existing
security mechani
s
m
s
. Thi
s
approachi
s
based
on usi
ng two
strategi
es
of adaptatio
n to adapt t
he m
obile ag
ent secu
rity at run
t
ime, depen
di
ng
on the sen
s
iti
v
ity of the service
s
re
quire
d to
perform
the duties of t
he agent an
d the degree
of
confid
en
ce of
the visited platforms.
We
have
co
nce
p
tually
structu
r
e
d
the
overall
s
y
s
t
em in
to
thr
ee fu
n
c
tion
a
l
ar
e
a
s
:
a)
Cre
a
tion Are
a
: whe
r
e ag
e
n
ts are creat
ed, validat
ed,
appraised, a
nd publi
s
h
ed;
b) Deploym
e
n
t
Area: add
s
secu
rity comp
onent
s acco
rding to the
sensitivity of the se
rv
ices
required by the
agent, an
d d
e
termin
e the
adaptatio
n
policy to b
e
followe
d by
the mobil
e
agent d
u
rin
g
its
executio
n, and c) Execution Ar
ea: wh
ich co
ntain
s
actual r
untim
e comp
one
nts for (phy
sical
netwo
rk) a
g
e
n
ts. Agent
s traverse th
e n
e
twork
and
p
e
rform
their tasks i
n
the
Execution A
r
ea.
Agent sele
cts a
nd
ada
pts the
compo
nents of
se
curity tasks t
o
ru
n o
n
thi
s
pl
atform. T
h
is
adaptatio
n is
perfo
rmed
by the mobil
e
itself de
pen
di
n
g
on
confid
en
ce level
of the platform
ag
ent
visited.
Figure 1. Gen
e
ral a
r
chitect
u
re of the sy
stem
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 15, No. 2, August 2015 : 381 –
389
384
In order to in
cre
a
se th
e le
vel of securit
y
offered
an
d
provide
a
sa
tisfacto
ry protection
model, we al
so u
s
e traditi
onal p
r
ote
c
tion me
cha
n
isms such a
s
password, cryptograp
hy a
n
d
digital sig
nat
ure. Th
ese mech
ani
sm
s help to pres
e
r
ve the integ
r
ity of the data of our mo
bile
agent, and to
control acce
ss to its re
sou
r
ce
s.
3.1.
Compon
ents
of the Sy
stem
3.1.1.
SSA (Sy
s
tem of Security
Adaptation
)
Systems ba
sed o
n
m
obile
age
nts are
chara
c
te
rized by a very dynamic a
s
p
e
ct
. This is
due
mainly to
the mi
gratio
n of a
gent
s t
o
multiple
system
s with
di
fferent
b
ehavi
o
rs a
nd
se
cu
rity
polici
e
s. Ind
e
ed, whil
e visit
i
ng a n
e
w
sy
stem, the
a
g
ent mu
st ada
pt dynami
c
all
y
to the se
cu
rity
requi
rem
ent.
The d
e
finition
of an
ada
pta
t
ion poli
c
y an
d secu
rity co
mpone
nts
of
the mobil
e
a
g
ent is a
cru
c
ial
ste
p
f
o
r the
effe
ctive implem
ent
ation of
se
cu
rity in a
mobi
le ag
ent
syst
em. The
go
al
of
the SSA is to protect the
mobile agent
via a static
adaptation
(see Figure
2). This adaptati
on is
to transfo
rm the mobile a
gent to a se
cure mobile
agent. TheS
SA adds ad
ditional se
cu
rity
behavio
rs
(secu
r
ity com
pone
nts) an
d dete
r
mi
ne
s its
dynam
ic ad
aptatio
n poli
c
y du
ring
executio
n.
Figure 2. The
adaptation of
a mobile age
nt to theSSA
3.1.1. MACD
(Mana
g
emen
t Age
n
t of
Confiden
ce Degree
)
Secu
rity age
nt requi
re
s dynamic a
s
sessment of
the cre
d
ibility of the ho
st to visit, which
can
in
cre
a
s
e
or
de
cr
ea
se
t
he
se
cu
rity of the m
obile
age
nt. The
mobile
age
nt tru
s
ts th
e h
o
st
whe
n
con
s
id
ering
its de
gree of
confide
n
ce
a
s
ve
ry
f
a
vorabl
e, o
r
i
n
crea
se
the l
e
vel of
se
cu
ri
ty
in
the ca
se of an unfavorab
le confide
n
ce
degre
e
. He
n
c
e, it is important that mobile agent
s can
authenti
c
ate
and ide
n
tify the list of host
s
in their itine
r
ary.
MACD ad
ds
the nece
s
sa
ry information
for t
he estimation the deg
ree of confid
ence by
the mobile a
gent. This e
s
timate is based on
inform
ation stored i
n
the mobile
agent and t
h
e
informatio
n collecte
d
from the environ
m
ent visi
ted. MACD up
date
s
such informa
t
ion following
an
insp
ectio
n
/ obse
r
vation by
the mobile a
gent
duri
ng its execution o
n
the host visi
ted.
3.1.2.
MAS (Ma
n
ag
ement Agen
t of Serv
ices)
The MAS is
use
d
to analy
z
e a
nd dete
r
mine the
sen
s
itivity of the
se
rvi
c
e
s
re
qu
ested
by
mobile a
gent
s (e.g. Top S
e
cret, Secret, Confid
ential,
Re
stri
cted, a
nd Uncl
assifi
ed), the
n
sea
r
ch
and filter
ho
sts that provid
e thes
e servi
c
e
s
from
UDDI se
rver
(se
e
Figu
re 1
)
.
The o
b
jective
o
f
s
e
rvic
es
c
l
ass
i
fic
a
tion is
to protec
t
the of
ficer
cond
ucti
ng se
nsitive tasks.
3.1.3.
MAC
(Man
ag
ement Agen
t of Cer
t
ifica
t
es)
The MA
C u
s
e
s
symm
etric an
d a
s
ym
metric
crypto
grap
hy to prevent the b
ehavior
analysi
s
of t
he mobil
e
ag
ent. Symmetric cryptog
r
ap
hy is u
s
ed t
o
en
crypt / d
e
crypt
sen
s
iti
v
e
tasks of the mobile agent. The list of secret
keys is
shared between the MSAS and the host
s
.
Asymmetri
c
cryptog
r
a
phy
is use
d
to ensure
a
u
th
enticatio
n an
d se
curity of commu
nication
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Sec
u
rity
for
Mobile Agents
: Trus
t Es
ti
mate for Platforms
(Raz
ouk
i
Has
s
an)
385
betwe
en the
different
ent
ities of the
system. Fo
r
t
h
is,
it
u
s
e
s
r
e
liable
me
ch
anism
s
su
ch
as
encryption, hashing and
digital sign
ature.
T
he MAC can a
l
so re
cord and che
c
k the
hos
t
s
’certific
a
tes
(ID
c
e
rtificate, validate date ...).
3.1.4. UDDI
Serv
er
UDDI Server
is based on the st
and
ard concept of UD
DI
serv
e
r
s,
a
s
spe
c
if
ied by
OA
S
I
S
[14], i.e. web servi
c
e
s
pub
lishin
g
and
discovery. In
o
u
r sy
stem, UDDI Se
rver
a
c
ts a
s
a regi
stry
contai
ning a
spe
c
ial
categ
o
ry of se
rvice
s
: age
nt
provi
s
ion
web
se
rv
ice
s
. Any business entity (i.
e
.
Agent Facto
r
y) who
want
s to provide a
gents to
the
comm
unity (i.e. end
u
s
ers) must prope
rl
y
expose a we
b servi
c
e inte
rface a
nd correc
tly publi
s
h
this we
b se
rvice interfa
c
e.
3.1.5.
Cer
t
ifica
t
e Authority
Local
Certificate a
u
t
hority local is an entity that
issue
s
digital
ce
rtificate
s
. The di
gital ce
rtificate
certifie
s the
o
w
ne
rship
of a
publi
c
key
b
y
the platfo
rm of the
ce
rti
f
icate. Thi
s
al
lows the
mob
ile
agent to rely upon
sign
ature
s
o
r
a
s
sertions ma
de
b
y
the private
key that correspon
ds to t
h
e
publi
c
key th
at is
ce
rtified. In this m
odel
of tru
s
t
rel
a
tionship
s
, a
CA is a t
r
u
s
ted
third
party th
at is
trusted
by
bot
h the
platform of the
certi
f
icate a
n
d
the
age
nt relying
upo
n th
e
certificate. CA
s
are
cha
r
a
c
teri
stic of many public ke
y infras
truc
ture (PKI) sc
hemes
.
3.2. Protec
tion
Protoc
ol
This
proto
c
ol
aims to
prote
c
t a mo
bile a
gent
code
ag
ainst m
a
licio
u
s
ho
sts. T
he
estimate
the deg
ree of
confide
n
ce o
c
cupie
s
the
center of
o
u
r
works in
ceit reveal
s the trust deg
re
e of
the
target ho
st (see Figu
re 3).
Figure 3. Sce
nario of mo
bil
e
agent exe
c
ution
1)
The a
gent
o
w
ne
r
creates and i
n
itiate
s a
mobil
e
agent to
pe
rform the
tasks
requ
este
d by
the cu
stom
e
r
. The b
ehavi
o
r of the m
o
b
ile age
nt is e
n
crypte
d u
s
in
g the pu
blic
key
of MSAS, which allows it to
guarantee the
confidentiality and
i
n
tegrity of mi
gration bet
ween
owner and MSAS.
2)
The age
nt owner send
s the
mobile age
nt to the MSAS.
3)
The MSAS decrypt
s the behavior of the m
obile agent
using it
s priv
ate key, and t
hen
analyses an
d
dete
r
mine
s t
he
sen
s
itivity of the
se
rvice
s
req
u
e
s
ted
by the
agent,
then
ada
pt t
he
agent’
s
se
cu
rity and determine the sm
a
r
t itinerary.
4)
The MSAS encrypt the sensit
ive tasks
of the mobile agent.
The list of secret keys is
shared between the MSAS and the visited hosts.
5)
The mo
bile
agent mu
st
authenti
c
ate
t
he ho
st in que
stion,
and verify
some
informatio
n n
e
ce
ssary for
the perfo
rma
n
ce of it
s ta
sk. To do
so, the age
nt mu
st obtain
cert
ain
informatio
n from the runti
m
e environ
m
ent (for
exa
m
ple, the identity of the
host visited
the
password, digital certificate...). Then, the mob
ile agent encrypts the information collected usi
n
g
the publi
c
key
of MSAS, to
compare them with those i
t
holds.
6)
After authen
tication
s of the host, the
mobile ag
ent estimate
the degree
of
confid
en
ce a
nd extra
c
t the tasks that
need to b
e
e
x
ecuted
on this platfo
rm. Dep
endin
g
o
n
the
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 15, No. 2, August 2015 : 381 –
389
386
adaptatio
n p
o
licy, the ag
ent may de
ci
de to de
cr
ypt
sen
s
itive tasks
and exe
c
ute them on
the
host if the co
nfiden
ce is very favorable
,
add
or repl
ace
se
curity comp
one
nts i
f
the degree
of
confid
en
ce i
s
not favo
rabl
e, or sto
p
a
n
d
leav
e
the
host to
go
to
the n
e
xt whi
l
e notifying t
he
MSAS about this failure.
7)
The m
obile
agent
retu
rn
s to the
ag
ent
owne
r
with
the results
o
b
tained
from
the
different ho
sts.
3.3.
Internal Stru
cture o
f
a Se
cure Mobile
Agen
t
Figure 4
outlines th
e g
ene
ral a
r
chitectu
re of a
secure mobil
e
ag
e
n
t model
ada
pted by
the MSAS. T
h
is architecture defines four ma
in parts in an agent. The interface between t
h
e
agent
and
th
e ru
ntime
pla
tform. The
b
a
si
c level
is the "op
e
rative part"
of the
agent. Th
e m
e
ta
level is non
-functio
nal services of the
agent a
s
se
curity, commu
nicatio
n
, mob
ility. The agent's
memory
cont
ains the info
rmation nee
de
d to perform the wo
rk of th
e mobile ag
e
n
t.
Figure 4. Internal st
ru
cture
of a secure m
obile ag
ent
3.3.1. Interface
This is the compon
ent through which the agent co
mmuni
cate
s and interact
s with the
runtime e
n
vironment. Thi
s
comp
onent
allows aut
h
e
n
ticating pl
atform
s, detecti
ng acce
sse
s
o
f
the external h
o
sts to its resour
ce
s, re
cei
v
ing requ
est
s
and providi
n
g servi
c
e
s
in a suitabl
e form.
The interfa
c
e
contai
ns thre
e comp
one
nts:
1)
Communicati
on use
d
to se
nd message
s to other age
nts and
recei
v
e messag
e
s from
other agents.
2)
Sensor is
use
d
for the perception of
the environment and the acqui
sition of data on
the visited platform. These data are
u
s
eful for
auth
entication an
d the adaptation of the mo
bile
agent.
3)
Actuator is used to execute a sequence
of actions
on a selected component. The
choice of executing compo
nent
is determined by
the
controller.
3.3.2. Meta
L
e
v
e
l
The meta lev
e
l rep
r
e
s
ent
s the non-fu
nctional co
de of
the mobile a
gent de
scribi
ng non-
function
al
services and pe
rformin
g
tra
n
s
form
ation
s
o
n
the b
a
si
c l
e
vel, the met
a
level contai
ns
components created
by
the agent owner and compon
ents added by the MSAS.
1)
Policy assem
b
ly is a non
-functional com
ponent added
by the owne
r's agent, u
s
e
d
to
control the a
ssembly process of comp
onents u
s
ed
by the age
nt, since the
outbreak of
an
assembly to its realization. Th
is policy also initia
lizes the componen
ts with default settings.
2)
Adaptation policy: represents a very im
portant factor to suppo
rt the protectio
n
strategy we
propose for
our mobile a
gent. Indeed,
the adaptation policy is
a set of rule
s tha
t
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Sec
u
rity
for
Mobile Agents
: Trus
t Es
ti
mate for Platforms
(Raz
ouk
i
Has
s
an)
387
defines the relationship between the
secu
rity
components of the agent and the degree o
f
confidence associated with change
s of th
e execution context.
3)
Intelligent itin
erary: con
s
ist
s
of a
set
of
n
odes. A nod
e
is a
co
nnecti
on point between
the compone
nt to be exe
c
uted by the
agent, t
he i
dentity and the co
nfidence degree on
the
platform of execution. This section al
so contai
n
s
th
e estimated t
i
me of the e
x
ecution of e
a
ch
component in order to avoid replay attacks.
4)
Security com
ponents
contain se
curity co
mponents
re
placeable tha
t
the mobile agent
can use during adaptation. T
he components are added by the
MSAS based on the security
needs of the mobile agent. Each component is a
piece of code selected
by the controller
According to the degree of confidence of each platform.
5)
Controller pla
y
s the role
of
coordinator
b
e
tween the
components of
the mobile a
gent.
Provides the proper funct
i
oning of the
whol
e agen
t system and can also
detect any p
oor
execution of
the mobile agent in the platform
in question. The controller comp
ares the actu
al
execution time of each task with the estimated
time
. If the controlle
r discovers an exceeding in
the prescribe
d
time,
it can
assume this
as a
misuse
of the mobile
agent and therefore, the agen
t
must stop its execution in the conce
r
ned
host in order
to migrate
to the next
host
in it
s it
inerar
y.
The controller sends notifications to
MSAS, following
a detected attack.
3.3.3. Basic
Lev
e
l
The b
a
si
c l
e
vel co
ntain
s
a fun
c
tion
al com
pon
e
n
t libra
ry of mobile a
g
e
n
t, each
comp
one
nt with a u
n
iqu
e
i
dentifier
and
carrie
s
out
a
spe
c
ific task,
the a
s
sembly
of comp
one
nts
that impleme
n
t the fun
c
tionality of ou
r mobil
e
ag
e
n
t, combi
nati
on an
d sequ
enci
ng of th
ese
comp
one
nts to first be stu
d
i
ed and
spe
c
i
f
ied by the assembly poli
c
y
.
3.3.4. Agen
t
Memo
r
y
Agent memo
ry contain
s
st
atic and dyn
a
mic dat
a tra
n
sp
orted fro
m
the origina
l
host or
host
s
visited
by the mobile
agent.
1)
Static data:
this is information that
is transported
from the original host. This
information contains data
that do
not change such
as the identity
of the creator and the
digital
signature.
2)
Dynamic data
:
this is information collect
ed from the
runtime environment after e
a
ch
migration. Su
ch As the partial results of calculat
ion of e
a
ch platform,
the data obtained during the
execution of
the agent.
3.4.
Trust Estim
a
tion
The qua
ntitative dimensi
o
ns of tru
s
t are
ba
sed o
n
the quantita
t
ive dimensio
n of its
cog
n
itive con
s
tituents. To
determi
ne the
host tru
s
t, we must identif
y the followin
g
para
m
eters:
1)
Parameters
that mak
e
a tr
ansaction tru
s
tworthy.
2)
Parameters that determine a leve
l of tru
s
t of e
v
ery costumer.
3)
Parameters that determine a set of
costumers to which the host belongs.
4)
Software an
d hard
w
are
parameters that
may affect perce
ption of trust and
transaction fulfillment.
5)
Reputation (may not exist)
of hosts pro
v
ided by the
agent’s owne
r or a third party
and witch related to
the history of host’s
transa
c
tions.
The tru
s
t deg
ree e
s
timatio
n
of the visited ho
st
is cal
c
ulate
d
usin
g
the colle
ctio
n of th
e
values of certain pa
ramet
e
rs sta
r
ting f
r
om th
e envi
r
onm
ent. The
trust
deg
ree
T is calculat
ed
according
to i
m
porta
nce Ij, wei
ght
Wj of
the p
a
ramet
e
r j
and
fa
cto
r
Sj
whi
c
h i
s
equal
to 1
in
the
ca
se of su
ccess (confo
r
m
i
ty of
information), and e
qual to 0 in the case of
a failure (n
on-
confo
r
mity of
informatio
n).
The trust
de
gree
e
s
timation is pe
rform
ed a
c
cording
to the follo
wi
ng
formula (1):
(
1
)
Each p
a
ra
me
ter J ha
s p
r
e
defined valu
e
s
of
its impo
rtance and its
weig
ht. These values
are
sto
r
ed
int
o
the m
obile
agent m
e
mo
ry. With an
ai
m of de
cidi
ng
on
an
adeq
u
a
te re
actio
n
, the
value of T is
comp
ared wit
h
the limits o
f
the va
riou
s trust e
s
timati
on interval
s (see T
able
1).
If
the tru
s
t valu
e bel
ong
s to
the go
od inte
rval (e.g
., [81
-
100]
) the
ho
st is tru
s
ted,
the ag
ent
ca
n
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 15, No. 2, August 2015 : 381 –
389
388
execute safel
y
. On the other ha
nd, if the obtaine
d trust value is con
s
ide
r
ed to be low, it is
possibl
e to find the exa
c
t cau
s
e of thi
s
failure
by se
ekin
g amo
n
g
the paramet
ers
whi
c
h
ha
d a
factor S
equ
a
l
to ze
ro, an
d
then the m
obi
le age
nt sel
e
cts a
nd
adapt
s security co
mpone
nts to t
h
e
tasks to
be
p
e
rform
ed
by the ho
st. Give
n the imp
o
rta
n
ce
of some
para
m
eters, t
hey ca
n (i
n
case
of failure
) la
rgely influen
ce the
choi
ce
of the
a
c
tion
to be u
nde
rta
k
en. T
he val
ues
assig
ned
to
the attribute
s
(the
weig
ht
and the
impo
rtance)
of
ea
ch p
a
ramete
r define it
s im
pact in
the fi
nal
deci
s
io
n.
Table 1. Example of estim
a
tion inte
rval
s with their
re
lated feedb
ack
Interval of
trust
estimation
Feedback
commentar
y
0
→
20
Stopping
The mo
bile agen
t
w
ill stop and e
x
it the cu
rren
t
host after no
tif
y
in
g the re
ason for
w
h
ich he
took
the decision and go to the ne
xt ho
st.
21
→
80
Reducing
The mobile agent selects and adapts security
components to the tasks to be perform
ed b
y
this host.
81
→
100
Performing
The agent can
ru
n safely
4. Conclu
sion
Our stu
d
y showed that the se
cu
rity
of the mobile agent req
u
i
res the ad
di
tion of
a
dynamic assessment
of t
he
credibility
of the host
visited and
det
ermination of
the sensitivity
of
the se
rvice
s
reque
sted by the mobile a
g
ent. A
cco
rdin
gto our
kno
w
l
edge, no a
p
p
r
oa
ch ha
s u
s
ed
the servi
c
e
sen
s
itivity and adapta
b
ility to so
lve this proble
m
. We have p
r
opo
se
d a n
e
w
architectu
re f
o
r mo
bile ag
ents a
nd ide
n
t
ified the
functions of different sy
stem compon
ents. This
prop
osal is
b
a
se
d on two
strategi
es
of adaptatio
n. The first i
s
a
static ada
ptation pe
rform
e
d
by
the MSAS. The second i
s
a reflexive dynamic
st
ruct
ural adaptat
i
on perform
e
d by the mobil
e
agent itself.
Referen
ces
[1]
M Alvaro, R Sergi.
Co
mpo
nent-Bas
ed D
e
vel
o
p
m
e
n
t of Secure Mob
i
le Age
n
ts Applicati
o
n
. I
n
Procee
din
g
s of
Multi-Age
n
t S
y
stems
and Ap
p
licatio
ns V. 200
7: 113-1
22.
[2]
G Samaras. M
obil
e
ag
ents:
W
hat ab
out th
em?
Did
th
e
y
deliv
er
w
h
at
th
e
y
pr
omise
d
?
Are the
y
h
e
re t
o
sta
y
?. Univers
i
t
y
of C
y
prus. 2
004: 29
4-2
95.
[3]
G Karjoth,
DB
Lan
ge, M Osh
i
ma.
A Sec
u
rity
Model for A
g
let
s
. In IEEE Inter
net Co
mputing.
2009; 1(
4):
68-7
7
.
[4]
S Z
h
ido
ng, W
Xi
ao
pin
g
.
A T
r
usted Co
mp
uting T
e
c
h
n
o
l
ogy En
abl
ed
Mobil
e
Ag
ent
System
. I
n
Procee
din
g
s of
Internati
ona
l
Confer
ence
o
n
Comp
uter
Sci
ence
an
d Soft
w
a
re E
n
g
i
ne
eri
ng. 2
008:
56
7-
570.
[5]
R WANG, T
H
U
, XX
. Research
in t
o
mob
i
l
e
age
nt secur
i
t
y
.
Journa
l of C
h
ong
qin
g
U
n
iver
sity of Posts
and T
e
l
e
co
mmunic
a
tions.
2
0
0
4
; 16(3): 81-8
6
.
[6]
N L
u
kasz, P
M
a
rcin,
R Mic
h
a
l
. Mobi
le
ag
ent
securit
y
.
1st
T
homas
e
d
itio
n. Informatio
n
as
suranc
e
an
d
computer sec
u
rit
y
. 20
06; 10
2-
123.
[7]
P Bella vist
a, A Corrad
i
, C F
r
ederic
i, R Mo
ntanar
i,
D T
i
baldi. Secur
i
t
y
for
Mobil
e
Age
n
ts: Issues an
d
Chal
le
nges. In
: I Mahgoub,
M Il
y
a
s.
Editors
. Invited Ch
apter in th
e Book Ha
nd
bo
ok of Mobil
e
Comp
uting. C
RC Press. 200
4.
[8]
X Z
h
a
ng, F
Pa
risi-Presicc
e, R
Sand
hu.
To
wa
rd
s Re
mo
te
Po
l
i
cy E
n
force
m
e
n
t for Ru
nti
m
e Pr
otectio
n
of Mobil
e
Cod
e
Using T
r
usted
Co
mp
uting
. Internati
ona
l W
o
rkshop o
n
Secu
rit
y
(IW
SEC). 2006.
[9]
FT
Ibharalu, A
B
Sofolu
w
e
, A
T
Akin
w
a
l
e
. A
relia
bl
e protect
i
on
architectur
e
for mob
ile
a
gents i
n
o
p
e
n
net
w
o
rk s
y
ste
m
.
Internation
a
l
jour
nal of co
mputer ap
plic
atio
ns.
2011; 1
7
(7)
:
6-14.
[1
0
]
P N
i
s
h
a
,
K Sun
i
l
,
B Ashu
. Secu
ri
ty
on
mo
b
ile
ag
en
t ba
se
d cra
w
le
r.
Inter
n
ation
a
l jo
urna
l of
co
mp
ute
r
app
licati
ons.
2
010; I(IS): 5-11.
[11]
A Shibli, I Y
ousaf, S Muftic.
MagicNET: Security system fo
r pr
otec
tion of mobi
le
agents.
In
Proceedings of
IEEE Internationa
l Conferenc
e
on A
d
vanced Informa
tion Net
w
orking
a
nd Applications.
201
0: 123
3-12
40.
[12]
MA Shi
b
li, S
M
u
ftic, A Giambr
uno, A
Li
o
y
.
M
agicNET: Security system fo
r development,
v
a
lidation and
ado
ptio
n of
mobil
e
a
g
e
n
ts.
In Proce
edi
ngs
of 3d Inter
nat
ion
a
l C
onfer
en
ce on
Net
w
ork
and s
y
ste
m
securit
y
. 200
9:
389-
396.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Sec
u
rity
for
Mobile Agents
: Trus
t Es
ti
mate for Platforms
(Raz
ouk
i
Has
s
an)
389
[13]
S Lerich
e
, J Arcang
eli. F
l
e
x
i
b
le arc
h
it
ectur
e
s of adaptiv
e
agents:
the a
gent ap
pro
a
ch
.
International
jour
nal of grid computing an
d m
u
lti agent system
s (IJGCMAS).
2010; 1(1)
: 55-75.
[14]
T
Bell
w
o
od,
et al. UD
DI Ver
s
ion
3.0. UD
D
I
Spec T
e
chni
cal C
o
mmittee
Specific
atio
n. 200
2.
w
e
b:
http://uddi.org/
pubs/u
dd
i-v3.0
0-pu
blis
he
d-20
020
71
9.htm.
Evaluation Warning : The document was created with Spire.PDF for Python.