Wireless Sensor Network, 2010, 2, 373-380
doi:10.4236/wsn.2010.24049 Published Online May 2010 (http://www.SciRP.org/journal/wsn)
Research on Beta Trust Model of Wireless Sensor
Networks Based on Energy Load Balancing
Danwei Chen1, Xizhou Yu1, Xi ang hui Dong2
1College of Computer, Nanjing University of Posts and Telecommunications, Nanjing, China
2ZTE Corporation, Shenzhen, China
Received February 27, 2010; revised March 18, 2010; accepted March 19, 2010
Abstract
This paper proposed beta trust model based on energy load balancing combines the recent achievements of
the trust models in distributed networks, together with the characteristics of wireless sensor networks. The
inter-node trust relation is established after an overall evaluation of node trust value based on the monitor
results of the node packets forwarding behavior conducted by inter-node collaboration. Due to the node en-
ergy limitation in wireless sensor networks, energy load balancing mechanism is applied to prolong the node
survival time. And the redundant routing protocol involves the presented trust model to develop the novel trust
routing protocol of beta trust model based on energy load balancing. Simulation performance demonstrates that
the beta trust model based on energy load balancing outperforms current schemes in energy consumption.
Keywords: Wireless Sensor Networks, Beta Trust Model, Trust Routing Protocol, Network Security, Trust
Evaluation
1. Introduction
There are various attacking threats in Wireless Sensor
Networks (WSNs) [1], which can be classified into rout-
ing protocol loopholes related attack, such as Sybil attack,
false routing information attack, selective forwarding
attack, Sinkhole attack, Wormhole attacks and fraudulent
confirmation attack; and broadcast authentication loop-
holes related attack, such as HELLO flood attack, DOS
attack and DDOS attack. Donggan Liu et al. proposed
μTESLA [2] and solved the broadcast authentication
loopholes related attack in most application scenario.
The former type of attack, however, remains an open
question for developing a generalized model for common
applications. Center-less topology in WSNs determines
that the nodes can not be authenticated uniformly by a
third party. And the trust management system can be
utilized in distributed network, aiming at establishing a
trust network participated with trust nodes. Therefore, it
is significant to develop a trust management system
suitable for WSNs.
2. Basic Requirements for Trust System in WSNs
Trust nodes in WSNs shall constitute a trust network
which can resist attack from hostile nodes effectively,
and save node energy and network bandwidth to ensure
network existence time and effectiveness. The designed
trust model shall meet the following requirements:
1) Moderate protocol algorithm complexity. The time
and space complexity of the protocol algorithm shall
meet the process speed and storage space requirements.
And digital signature and public key pair system are not
suitable for WSNs.
2) Moderate protocol communication. The energy
consumption from inter-node communication outgrows
that from calculation within the node.
3) Trust evaluation effectiveness. The trust model
shall evaluate the node trust value effectively and reflect
the dynamic node trust value in time by monitoring the
node behavior. Therefore, it is a challenge to reach a
moderate balance between 2) and 3) so as to design an
applicable trust system in WSNs.
4) Security in WSNs. The designed system shall be
able to resist malicious attack, identify attack behavior
from hostile node and take effective resisting actions.
3. Trust Evaluation Model
Blaze M. [3] firstly introduced the trust management for

D. W. CHEN ET AL. 374
the security in distributed system to solve the restriction
using traditional encryption system. The early trust
management engines are KeyNote [4] and RT framework
[5]. Then trust management-based trust systems are ap-
plied in e-commerce, ad hoc network and peer network.
The network nodes in the WSNs are task-oriented, and
functions as 1) transmitting nodes to collect the desig-
nated relevant information and report to the base station,
or 2) forwarding nodes to forward data packet from other
nodes to the base station. Considering encryption and
digests are applied to guarantee the system confidential-
ity and integrity, the node trust values are based on the
inter-node exchange behavior. Unlike the current trust
systems, such as KeyNote, RT framework, eBay [6],
CONFIDANT [7] and peer network, our model com-
bines the direct and the indirect trust value to give an
overall evaluation.
We established the trust model based on the node data
forwarding statistics and the energy conservation. And
the trust value of the routing node is evaluated by 1) the
direct trust value from the lower-lever nodes and 2) the
indirect trust value from the neighbor nodes.
Node trust value is computed through all the evalua-
tions from other nodes, which avoids computing the ex-
act local trust value and solves the dynamic trust value
problem. It enhances the veracity of the trust mechanism
and reflects the relativity of trust value and time. And
chances are that some reliable nodes with high trust val-
ue would take on more forwarding tasks from neighbor
nodes, which renders them more vulnerable to be trapped
in energy exhaust. Our construction, however, takes this
situation into consideration and achieves energy load
balancing to prolong the network existence and enhance
the effectiveness.
4. Beta Trust Model Based on Energy Load
Balancing
4.1. Basic Description
The goal of the trust model is to choose credible node for
routing information in order to ensure the data to reach
the base station safely without losing packets maliciously.
The evaluation of overall credibility of nodes in trust
model would be involved in direct credibility and rec-
ommended credibility comprehensively (namely indirect
credibility), where the previous is concluded from direct
interaction with evaluated node, while the latter is in-
ferred from others nodes to the evaluated node. While
selecting next hop node in the consideration of energy
load balancing of sensor network, we will take surplus
energy ratio as a standard. The trust model is based on
the following assumptions: 1) WSNs is safe after ini-
tialization and 2) after routing discovery, each node
stores multiple routing paths to base station.
In the process of research, it needs two important
terms derived from trust evaluation areas, namely credi-
bility and reputation, where credibility means subjective
expectations (usually a real number within 0-1) derived
from A to B, and reputation based on the observation of
an individual history behavior is an expectation for fu-
ture behavior. This paper will also introduce a new term,
the node energy surplus ratio, which means a ratio be-
tween the node surplus energy and initial total energy.
For the sake of discussion, we define the expectation
derived from future behaviors of B based on an observa-
tion of B history behaviors as direct credibility from A to
B. The expectation is based on an observation of history
behaviors and evaluation information derived from A to
B, so we define such expectation of future behaviors of B
deduced from the observation information from C to B
achieved by A from C as indirect credibility of A to B,
which is based on an observation of history behaviors
and evaluation information from C to B.
In the stage of data transmission, nodes need to select
routing paths, that is to say next hop node. Trust value
obtained from the evaluation of trust system will be a
basis of selecting routing, and the arbitrary node will try
to choose neighbor nodes with high trust value and high
energy surplus ratio as routing node. As for neighbor
nodes whose trust value is lower than threshold value,
the node will submit mistrust reports to base station. If
base station receives the same mistrust report from dif-
ferent nodes to some node many times, it will exclude
the node from routing table, so as to achieve the goal that
the network consists of trusted nodes.
In the process of routing in WSNs, each node will
generate a neighbors list, which stores other nodes iden-
tification (ID) within its communication region, mean-
while every ID corresponds to credibility. Therefore,
every node has a credibility list, which saves the credi-
bility of other nodes from this node. The credibility of
node can be divided into two parts- direct credibility and
indirect part, in which direct credibility is calculated
from downstream nodes of routing announcement nodes
according to the state of forwarding packets by routing
announcement nodes; indirect credibility is deduced
from neighbor nodes via monitoring the state of for-
warding packets by routing announcement nodes. Com-
bining direct credibility with indirect credibility, the total
credibility of routing announcement nodes can be calcu-
lated so as to judge whether the node is trusted.
4.2. Establ ishme nt an d Calc ulat io n of Node Tr ust
Trust value uses to judge whether monitored nodes are
malicious; it is also an expectation for future behaviors
of monitored nodes and has close relationship with past
behaviors of monitored nodes. As an evaluation of past
behaviors, reputation can use to calculate current trust

D. W. CHEN ET AL.375
value. This paper will adapt reputation model to detect
whether to exist malicious nodes.
Bayesian can calculate posterior probability via prior
probability as shown in Equation (1)

j
PB means the
prior probability, and means the posterior
probability. The priori probability is probability accord-
ing to random events calculated from past experience.
The posterior probability, after random experience,
means an amendment for priori probability

|
j
PB A



j
P
B un-
der conditions of
A
resulted from random experience.
Therefore, via combination the prior probability with the
posterior probability, it can calculate probability of fu-
ture possible completed tasks according to situation of
past data forwarding completed by node.
 
 
1
|
|, 1,2,...,
|
jj
jn
ii
i
PB PAB
PB Ajn
PB PAB



(1)
where presents a combination of
events, which means that a particular detecting node gets
a special reputation value and .

1, 2,...,
j
Bj n


0
j
PB 
A
means a
specific observed event.
In order to reflect dynamic changes of node trust val-
ues as time goes by, we need to improve accuracy of
node trust evaluation. This paper will adopt a method of
sending detected packets regularly to update reputation
value. The event that packets forwarded by from i
is expressed as
j


ij t

at the time . The behavior that
observes the behaviors of from to is
expressed as . So it comes to the following for-
mula:
t
ij1tt

ij t
D
 

 
 
11
1
0
|
|
ijij ij
tt
ij t
t
ijij ij
lt
l
PPD
P
PPD










 


t
l






(2)

ij t
P 

 not only is the probability of an event that
completes tasks derived from , but also is the prior
probability of reputation derived from to , where
this reputation is denoted as
j
ii j


t
ij
R. We can see that,
is not only the posterior probability of node repu-
tation at the time , but also the priori probability at
the time .

ij t
R
t1t
On the assumption that has assigned tasks to
for times, has completed tasks for times
with the probability
i j
mnjm
x
, so the past reputation derived
from to is belonged to binomial distribution ij


,Bm xn. The priori probability of next task com-
pleted is belonged to homogeneous distribution


0,1U
on (0, 1) under no previous knowledge. As for individual
monitored node, the monitored node has only two types
of behavior: forwarding data or not forwarding data.
Therefore, the binomial distribution can be used to model
for the monitored nodes. The probability of completing
next task

by obeys that j
1
0
(0,1)(, )
()
[()][()| ()]
t
ij lij tij l
l
UBmnx
PPD 




P 
(3)
So the posterior probability of completing next event

by will be as follows j
 
1! 1
!!
n
m
mn xx
P

mn
 
  (4)
This paper draws on the experience of beta reputation
system of e-commerce, and introduces establishment for
node’s trust model, because beta distribution expresses
node's reputation. The probability density function of
beta distribution can be described as follows
 
1
1
1
,Bp
|,p1,0 1,
q
p
xxxp
q

0, 0qfx q

 
0

0,q
(5)
where

q
 
1
,1,01,
q
xx dxxp

1
0
1p
Bp


 (6)
The beta-family of probability density functions is a
continuous family of functions indexed by the two pa-
rameters and . The beta distribution p q


|,
f
xpq
can be expressed using the gamma function as




 
1
1
|,
p1,
01 , 0, 0
q
p
pqx x
pq
xp q





 
fx q (7)
with the restriction that the probability variable 0x

if
1p

, and 1x

if 1q

. The probability expectation
value of the beta distribution is given by

p
Ex pq  (8)
Considering the property of gamma function,




1!mm

, Equation (4) can be changed into:
 


21
11
n
m
mn xx
mn
 

P  (9)
So


P

is subject to distribution

1, 1mn  


D. W. CHEN ET AL. 376

compared Equation (7) with Equation (9) The reputation
derived from to is that i

j

1, 1
ij
Rmn   (10)
The expectation of completing next task derived from
to can be calculated by the previous formula, namely
the credibility of to . Let the direct credibility de-
note as
i
j
ij

ij
D
R, so the credibility of routing announce-
ment node is calculated by its downstream nodes. Equa-
tion (12) can be deduced by the expectation of beta dis-
tribution Equation (11):

m
Emnmn  

 (11)


1
1, 12
ij D
m
REmn mn






m
(12)
We suppose that equals to kn , where is set
as 9, 5, 1, 1/5 and 1/9 respectively, so the relationship
between the credibility and can be described as
Figure 1.
k
mn
We can see from Figure 1 that node’s credibility will
increase as the number of tasks completed by node that
the number of forwarding packets successfully increases;
on the contrary, nodes credibility will be decrease as the
number of uncompleted tasks that the number of for-
warding packets unsuccessfully is larger. The behaviors
of forwarding packets as a judge basis can reflect true
situation of node.
4.3. The Initialization of Node Trust Value
For the sake of description, we introduce two concepts:
routing node and non-routing node. Routing node is a
type of next hop neighbor node selected to forward
packets to the base station. Non-routing node means one
of neighbor nodes except routing nodes. The credibility
system mainly uses to ensure route security, therefore in
order to save unnecessary expenses, the trust evaluation
is only for routing nodes, however half trust attitude is
adopted for non-routing nodes (that is to say that the cre-
dibility of non-routing nodes is set as 0.5). Note that
non-routing node is not fixed, it is possible to become a
routing node at some time, and when a non-routing node
has been changed into a routing node, the system will
re-evaluate the node's credibility.
We have detailed the ideas and methods of node’s
trust assessment in Section 4.2. This section will describe
initialization of node trust value in beta credibility sys-
tem based on energy load balancing. The node trust val-
ue of the system is between 0 and 1, which can evaluate
comprehensively direct trust value and indirect trust
Figure 1. Impact on the credibility derived from the ratio
between m and n.
value. The initialization of node trust values will start
after establishment of WSNs and route discovery. Trust
values of routing nodes will be initialized by trust detect
mechanism, while trust values of non-routing nodes will
be set as 0.5 initially. On the assumption that is
routing node of i, means the number of packets
forwarded by successfully, and represents the
number of packets forwarded by unsuccessfully, the
direct trust value of to is that
j
m
i
jn
j
j


1
1, 12
ij D
m
REmn mn 



 (13)
Suppose is a common neighbor node of and ,
ki j
k
j
mmeans the number of packets forwarded by suc-
cessfully in the process of trust detection by ,
j
kk
j
n
ik
R
means the number of packets forwarded by unsuc-
cessfully in the process of trust detection by k,
represents general trusted assessment of on , then
the indirect trust value of to is that
j
i k
i j

1
2
k
j
ijik kk
ID kjj
m
RR
mn


 (14)
In the credibility system, one node has absolutely be-
lieved in its direct assessment for other nodes, while has
reservedly trust in recommended assessment derived
from other nodes for evaluated nodes. In order to prevent
malicious slander or have a common conspiracy to en-
hance trust value of a malicious node, it should make a
consideration of its trust of recommended nodes while
combined with recommended credibility own. The over-
all credibility of to is that i j

D. W. CHEN ET AL. 377
value of to can be updated to Equation (19). ij

0.5, is a non-routing node,
1
2, is a routing node
1
k
j
ijik kk
D ij kjj
ik
k
j
m
RR
Rmn j
R









(15)
In the course of updating credibility system, we not
only take reputation accumulated by history behaviors of
nodes into account, but also need consider sensitivity of
trust changes dynamically, so that it needs to reflect be-
havior changes of nodes in order to minimize the impact
on network after some nodes being captured.
It can be seen from Equation (15) that directly
depends on the ratio between and n. The more this
ratio is, the higher the trust value of node will be gotten;
vice versa. It is in accord with the design of evaluating
trust value of node via forwarding packets, and has a
finer defensive effect for malicious loss packets of nodes.
ij
R
m4.5. Trust Decision
The main objective of the credibility system is to select
trust routing nodes and network topology excluding ma-
licious nodes to ensure security of data transmission. At
the same time, taken limitation of sensor node’s power
into account, neighbor nodes will regard the node as a
routing node for a long time, and then the node’s power
may be run out quickly so that it may lead to partial fail-
ure of sensor network. In order to prevent above situation,
it need consider energy load balancing. Therefore, trust
decision is very important.
4.4. The Update of Node Credibility Value
The behavior of node may change as time goes by, so the
credibility system must update the trust values of nodes
dynamically. The credibility system adopts a method of
opening detection mechanism regularly to monitor
changes of routing node’s behaviors and update routing
node trust value dynamically, so that reflect the changes
of the sensor network in time and ensure the security of
data transmission, as for non-routing node without updating.
We suppose that
j
E is residual energy rate of ,
and
j
r

, e

mean trust weights and energy weights
respectively. The trust decision value, , is defined as
ij
T
ijr ijej
TRE



 (20)
Suppose that is a routing node of with and
obtained in a new round of trust detection, we define
the aging parameters
ji m
n
age

as the impact on current trust
evaluation derived from past detected data, then
We define as the trust threshold of system. When
makes a decision to select next hop routing node ,
it obeys the following decision-making principles:
t
R
i j
1) Select routing node with ;
ij t
RR
newold age
mm m

 (16)
2) If many trust values are all higher than , select
one of nodes with the greatest ;
t
R
ij
T
new old age
nn n

  (17)
Therefore the direct trust value of to can be
updated that
i j3) If many are same, choose one of routing nodes
with the highest ;
ij
T
ij
R

1
2
new
ij Dnew new new
m
Rmn

 (18) 4) If many are same, then choose one of routing
nodes with the shortest routing path.
ij
R
In the credibility system, when considers historical
behaviors of nodes, it should also take sensitivity of up-
dating trust values into account and needs reflect the be-
havior changes of nodes in time. Therefore, we consider
historical behaviors in direct trust while consider current
behaviors in indirect trust. On the assumption that

,
kk
After above decision-making, the routing node has
higher credibility with more remaining energy. The en-
ergy surplus ratio used in system also prevents the use of
high-power devices from attacking via using energy de-
fects of WSNs. As for one routing node with trust value
lower than , it will send a warning report to base sta-
tion. If base station recesives the same of warning report
from many nodes to a certain node, base station will
t
R

j
new jnew
mn is derived from monitoring of k on
in a new round of trust detection, then the overall trust
j
0.5, is a non-routing node,
1
() 2, is a routing node
1
k
jnew
ij Dnewiknewkk
ijnew kjnew jnew
iknew
k
j
m
RR
Rmn j
R









(19)

D. W. CHEN ET AL. 378
exclude this node from network topology, so that the
network consists of trust nodes.
5. Performance Analysis and Simulations of
Trust Routing Protocol
5.1. Performance Analysis of TRP and INSENS
INSENS (INtrusion-tolerant routing protocol for wireless
Sensor NetworkS) is a well-designed secure routing pro-
tocol, which achieve data efficient transmission by mak-
ing use of redundant routing [9]. In WSNs, it is essential
to save energy for the protocol designation; however,
INSENS cannot overcome more waste of energy from
sending packets multiply. We will introduce beta trust
model into INSENS to set up the trust detection mecha-
nism, evaluate routing node credibility, and make deci-
sion to choose some routing nodes to forward packets.
We will analyze security of our protocol based on the
beta trust model, called trust routing protocol (TRP) after
introduction of trust management model and resist
against current different typical attacks in WSNs.
In order to ensure packets to be forwarded to base sta-
tion safely, the way of sending packets of INSENS is
shown in Figure 2 (redundant routing mode). A data
packet is copied into a number of ectypes. The transmis-
sion path takes on a tree structure in network. Suppose
that a certain node with
H
hops to base station has
() routing paths, and each intermediate forwarding
node has all routing paths, and then including the
number of packets sent by source nodes and intermediate
forwarding nodes, the quantity of packets generated by
sending a data packet is that
N
1NN
1
21
H
iH iH H
i
NN
SNNNN





 (21)
Whereas the quantity generated under TRP is that
1SH (22)
Figure 2. INSENS: redundant routing protocol mode.
As the expansion of network scale, INSENS will make
network load increase exponentially, while the consump-
tion of TRP for network resource will almost increase
linearly. Thus, as long as proper control of communica-
tion consumption in the process of trust detection, the
communication consumption of TRP is much smaller
than INSENS. It can reduce node's energy consumption
and save network resources greatly, and be conducive to
network expansion. But the computing expense derived
from introduction of trust evaluation system relative to
communication expense is almost negligible.
5.2. Emulation of TRP and INSENS
The main goal of introducing beta trust model into IN-
SENS is to give up the way of sending packets multiply
via redundant routing path, and to adopt a way of trust
routing paths to send packets, which reduces energy con-
sumption of nodes and prolongs the survival time of
network, meanwhile alleviates network load and saves
communication resources.
In order to verify TRP described in this paper with an
introduction of energy load balancing beta trust model
whether satisfy the goal of this paper, this subsection will
make simulations for TRP and INSENS, and compare
the performance of two protocols according to simula-
tions. The weights used in the simulation are set as fol-
lows: weight_old = 0.6, weight_trust = 0.8, weigth_en-
ergy = 0.2, Rt = 0.6.
The paper will adopt two following evaluation indexes
to compare and analyze the performance of TRP and
INSENS.
1) The number of transmitted packets
Under the same conditions of sending the same pack-
ets, compare the total quantity of packets sent by all
nodes in the course of sending packets from source node
to destination node, including packets sent by source
node and forwarded by intermediate node. Because the
energy consumption of network is mainly embodied in
sending packets, this performance index can reflect not
only the difference of energy consumption in the process
of communication, but also the situation of network re-
sources usage.
2) Packet loss
It means a ratio of the number of packets not received
by destination node to the number of packets sent by
source node. This performance index can reflect the im-
pact on the protocol to network communication and
whether it is applicable to WSNs. The protocol with
higher packet loss is not obviously suitable to network
communications.
This paper also includes simulation of dynamic
changes of routing node trust value in order to verify two
additional problems: first, the ability of TRP resisting
malicious attacks; second, whether the node could dis-

D. W. CHEN ET AL.379
cover malicious routing nodes on upstream then exclude
them and select trust nodes. According to simulations,
we design two following scenarios.
Scenario 1: Suppose that there is a coordinate system
with base station at (0, 0). 100 nodes are distributed ran-
domly within the range of 1000*1000m2 in coordinate
system, and node’s communication distance is 250m.
INSENS and TRP will generate redundant routing paths.
For the sake of simplicity, the simulation will generate
two routing paths for each node as possible, however
some nodes may have a routing path because of topo-
logical structure. The system will select four nodes ran-
domly to generate 4 cbr data streams, where each cbr
data stream sends two packets per second, the length of
one data packet is 512 bytes, and the simulation time is
30 seconds. In trust routing, it can be seen from Figure 1,
when the number of packets detected by the system
reaches 30, it is more accurate for evaluation of node
trust value. Therefore, we select 30 packets sent by trust
detection at a time in simulation. In this scenario, it will
make a statistics about the number of transmitted packets
as shown in Figure 3 and packet loss as shown in Figure
4 in INSENS and TRP respectively.
Figure 3 shows that although the number of packets
sent by TRP is much more than INSENS in the initial
stage, after 12 seconds the later surpassed the former and
the gap between the two protocols becomes larger and
larger as time goes by. Because TRP starts trust routing
detection at the beginning and consume a certain amount
of network resources, once completes trust detection,
packets are forwarded in accordance with trust routing.
However INSENS always forwards packets according to
redundant routing. On the assumption that there are h
hops between the node and base station, and each
node has two routing paths, then the total quantity of
transmitted packets reaches about (3 * 2h-2) (2 + 22 + 23
+ ... + 2 * 2h). Obviously, according to INSENS, middle
nodes may discard duplicated packets, meanwhile be-
cause of signal conflict, network congestion and so on, it
also drops some packets, and in fact the quantity of a
packet transmitted in network may not reach (3 * 2h-2).
Figure 2 shows the network consumption of INSENS is
much more than that of TRP when sending the same
source packets, so the improvement derived from intro-
duction of trust evaluation system indeed saves a lot of
energy and network resources, extending survival time of
WSNs and improving effectiveness of completing tasks.
a
N
It can be seen from Figure 4, the average packet loss
of TRP is about 2.5%, which is higher than INSENS,
because of WSNs with higher packet loss. In TRP,
source packets are forwarded to base station only along a
routing path, while in INSENS, source packets are
spread over the network via redundant routing paths.
There are many copies of packets sent to base station
through multiple paths. Thus, INSENS has a slightly
05 10 1520 2530
0
1000
2000
3000
4000
5000
6000
7000
Time (s)
The number of transmmited packets
TRP
INSENS
Figure 3. The number of transmitted packets.
05 10 15 20 25 30
0
0.5
1
1.5
2
2.5
3
3.5
4
Time (s)
Packet loss (%)
TRP
INSENS
Figure 4. Packet loss.
lower packet loss. However, as expansion of network
scale and frequency of sending packets raises, the con-
sumption of INSENS for network resources will increase
exponentially, also it will result in more serious network
congestion and channel conflict, and its packet loss will
increase greatly. Whereas the consumption of TRP for
network resources under above mentioned situation will
almost increase linearly, it is much better than INSENS
in terms of network congestion and channel conflict.
Scenario 2: Suppose that there is a coordinate system
with base station at (0, 0). 100 nodes are distributed ran-
domly within the range of 1000*1000m2 in coordinate
system, and node’s communication distance is 250m.
The interval time of trust update is 30 seconds, and si-
mulation time is 60 seconds. In TRP, base station gener-
ates three routing paths for , of which the next hop
nodes are ,
V
N
A
N
B
N and . In the initial stage, ,
C
NA N
B
N and are all healthy nodes, while
C
N
B
N will be
captured within 0 ~ 30 seconds, which will discard all
packets without forwarding. In this simulation scenario,
trust values of to ,
V
NA N
B
N and can be
shown in Figure 5.
C
N

D. W. CHEN ET AL. 380
7. Acknowledgements
010 20 30 40 50 60
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Time (s)
Trust value
A
B
C
This work was sponsored by the National Natural Sci-
ence Foundation of P.R. China (No. 60973139, 60773041,
60905040), the Natural Science Foundation of Jiangsu
Province (BK2008451), Postdoctoral Foundation of Ji-
angsu Province (0801019C), Science & Technology In-
novation Fund for Higher Education Institutions of Ji-
angsu Province (CX08B-085Z, CX08B-086Z), and the
Six Kinds of Top Talent of Jiangsu Province (2008118).
8. References
[1] C. Karlof and D. Wanger, “Secure Routing in Wireless
Sensor Networks: Attacks and Counter-Measures,” First
IEEE International Workshop on Sensor Network Proto-
cols and Applications, IEEE Computer Society, Anchor-
age, May 2003, pp. 113-127.
Figure 5. Trust values of V
N
to
A
N
,
B
N
and C
N
.
It can be seen from Figure 5, the healthy node has rel-
ative higher trust value at 0s, namely at the time of ini-
tialization. When updating trust at 30s and 60s, its trust
value descends quickly due to that
B
N
A
N
is captured and
discards packets maliciously, while and have
still higher trust values. So we can see that trust value of
this paper can evaluate node behaviors accurately, and
detect malicious behaviors in time. The network may
select trust nodes to cooperate and minimize harm re-
sulted from malicious nodes as possible by trust assess-
ment.
C
N
[2] D. Liu and P. Ning, “Efficient Distribution of Key Chain
Commitments for Broadcast Authentication in Distrib-
uted Sensor Networks,” Technical Report: North Caro-
lina State University at Raleigh, September 2002.
[3] M. Blaze, J. Feigenbaum and J. Lacy, “Decentralized
Trust Management,” Proceedings of the 17th Symposium
on Security and Privacy, 1996.
[4] M. Blaze, J. Feigenbaum and J. Ioannidis, “The KeyNote
Trust Management System,” Version 2, Internet Engi-
neering Task Force, September 1999.
[5] N. H. Li, J. C. Mitchell, W. H. Winsborough, “Design of
a Role-Based Trust Management Framework,” In Header,
H. Ed., Proceedings of the IEEE Symposium on Security
and Privacy, IEEE Computer Society Press, Washington,
2002, pp. 114-130.
6. Conclusions
Nowadays it proves to be difficult to authenticate the
network entity in distributed network environment, espe-
cially in resource-constrained WSNs. Dynamic trust
management provides a novel solution of the security in
distributed environment. Due to own features of WSNs,
however, the traditional trust model fails to be applied in
the WSNs.
[6] P. Resnick and R. Zeckhauser, “Trust among Strangers in
Internet Transactions: Empirical Analysis of eBya’s
Reputation System,” National Bureau of Economic Re-
search Workshop on Empirical Studies of Electronic
Commerce, 2000.
[7] S. Buchegger and J. L. Boudec, “Performance Analysis of
the CONFIDANT Protocol,” Proceedings of ACM Mobi-
hoc, 2002.
This paper proposes the beta trust model based on en-
ergy load balancing, which is remarkable in computation,
communication consumption and energy load balancing,
combining with the trust system model and beta trust
model in the e-commerce. The simulation results indicate
that the evaluated node trust value reflects the node be-
haviors effectively. And our model achieves the expected
security goal with low energy and network consumption
in WSNs.
[8] L. Xiong and L. Liu, “A reputation-Based Trust Model
for Peer to Peer Ecommerce Communities,” IEEE Con-
ference on E-Commerce, 2003.
[9] J. Deng, R. Han and S. Mishra, “INSENS: Intrusion-
Tolerant Routing in Wireless Sensor Networks,” De-
partment of Computer Science Technical Report CU-CS-
939-02, University of Colorado, Boulder, 2006.