Veronica Eyo Sharvari Joshi

beansproutscompleteSoftware and s/w Development

Dec 13, 2013 (3 years and 6 months ago)

97 views

Veronica Eyo

Sharvari Joshi



The nodes are left unattended


An Adversary can physically compromise the nodes



Solution?


Authentication and cryptographic mechanisms alone
can not solve this problem



The trust system was used for security in the Wireless
sensor network




These systems identify malicious nodes,
computes a trust rating of the node and
exclude the nodes from the network.


First trust based model designed and
developed for sensor networks.



It makes use of the watchdog mechanism to
collect data and monitor different events
going on in the node, to build
reputation(R
ij
)of the node and then get the
trust rating (T
ij
)of the node.



T
ij =
P
j + 1


P
j +
N
j + 2


P
j =
Positive outcome

N
j =
Negative outcome


RFSN can not distinguish between negative and
positive events in the node.



Addressed the uncertainty issue but
still cooperated with the malicious
nodes.


Has one value of trust rating for
different events.










Event based trust framework is proposed to
detect malicious sensor nodes


A new protocol ESTN is developed


A new direction in trust system for wireless
sensor network is proposed



The trust rating is dependent on different
events in the node



Each event on the sensor node has a different
trust rating



A sensor node has several trust ratings stored
in its neighbor nodes


The agent node



Classifies all the events and then builds a
reputation table of the nodes



A trust table is then computed from the
reputation table and broadcast to all the
nodes.


Event E happened in sensor nodes.


Let E={e
1

,e
2

,e
3

,....e
n

}



Event function.


F={F(e
i

)| ∀e
i

∈E,F(e
i
) ≥1,F(e
i

)∈N}



Positive outcome p
i


Negative outcome n
i


<p ,n > is binary event for a certain event e
i

of sensor node.



Define reputation space of event e
i


RS(e
i

)={<p
i

,n
i
>|t
i
=p
i
+n
i
;p
i
=F(e
i

) or n
i
=F(e
i

)
∀e
i
∈E}


P<p
i

n
i
> (x)=
(p
i
+n
i
+1)! X
i

^(p
i
) (1
-
X
i
)^n
i


p
i
! n
i
!




Reputation:
opinion of one entity about
another


Trust:
Expectation of one entity about
another.


Reputation space to Trust space


Let T
i

(<p
i

,n
i

>)=(pt
i

,nt
i

,ut
i

) be the
transformation from binary event <p
i

,
n
i

> to
trust rating (
pt
i

,
nt
i
,ut
i

)



while True


For all the nodes s and t in the agent node radio


range


For event e1 to en


Agnet node gets the binary event <pi ,ni >


Agent computes the trust rating Tsti;


If (Tsti) < a certain value


Break;


End if


End for


The agent broadcasts the trust rating Tsti;


End for


If the time is the begin of window time


The agent broadcasts all the trust rating Tsti;


End If


End While


Modules used:


Wireless sensor networks (x
A
, x
B
…x
N
)


Agent nodes (x
i
)


Intruder nodes (x
m
)


Traffic data (

)


Events generator




Need to design a special Agent node


Higher processing power for the agent node


Complex architecture


Protection of agent node


Protect agent node from discovery.


Broadcast trust rating to the remote
monitoring facility.



This trust model can be used in large scale
wireless networks




It provides a more accurate guarantee along
with cryptographic mechanisms in detecting
malicious nodes of different events in sensor
networks





Questions?