A Security Model for
Wireless Sensor Networks
Hosein Marzi
Department of Information Systems,
St. Francis Xavier University
Antigonish, NS Canada
hmarzi@stfx.ca
AbstractUntil recently use of sensors to collect sensitive
parameters had only few risk factors such as sensor malfunction,
uncertainty of data collection, or missing data coverage. Some
issues in these categories were addressed by multi-sensor
application or sensor fusion. However, advancement in
technology and advent of wireless sensors in a networked
environment, brought along a new risk factor related to the
security in wireless sensor network.
Therefore security in Wireless Sensor Network WSNs is
challenging and critical to the functionality of the networked
sensors. This is very important in cases of highly secure
environment, especially in industrial, military, and medical
domains. The standard WSN protocols focus on energy
efficiency; transmission efficiency, and routing. WSN is known
for limitations on hardware and software and for resources-
constrained in general. An adaptive model of security that meets
requirements and constraints in WSN is Intrusion Detections.
This article investigates security in WSN and presents a design
process for achieving optimum security based on requirements
and constraints in WSNs. Further, comparative results between a
proposed technique and other security current approaches are
discussed.
Keywords Security, wireless sensor network, trust and
reputation system, client sensors, server sensors, security level, path
length, energy efficiency.
I. INTRODUCTION
Sensors have become a prevalent part of many aspects of
daily life and are greatly relied upon in technological
environments. The range of sensor usage varies from standard
to advanced levels. It is rarer to see the use of single senses to
accomplish a task. Multisensors are more frequently
employed to enhance coverage and accuracy, and application
of sensor fusion has become increasingly ubiquitous [22], [23].
The new technology facilitates sensor application in a
networked and wireless environment. While sensor fusion
benefits significantly from wireless technology, it may suffer
for security reasons. Not all applications benefit from a
standard resourceful environment. In many cases, the most
fundamental requirement of Wireless Sensor Networks WSNs
is security, under resource constraint operation. Various
approaches have been made to enhance network security,
without heavily straining resources such as energy. In any
networked environment, the security services of authentication
and confidentiality are highly important factors in maintaining
communications. The Trust and Reputation Model (TRM) is an
innovative security mechanism for wireless sensor networks.
TRM ensures a minimum security level between two entities
978-1-4799-2614-5/14/$31.00 2014 IEEE
Arash Marzi
School of Electrical Engineering & Computer Science
University of Ottawa
Ottawa, ON Canada
amarz052@uottawa.ca
having communications, data exchanges or interactions within
a distributed environment. This article extends and presents the
outcome of study in security of WSN environment [21], and
proposes improvements to the concept of Trust and Reputation
Model. TRM in a general sense is combination of Bio-inspired
Trust and Reputation Model for Wireless Sensor Networks
BTRM-WSN as well as Peer Trust system PTS. A review of
wireless sensor networks and importance of their security will
be given here in this section. Description of Trust and
Reputation Model will be presented and two security
approaches, that is, BTRM and PTS that led to the proposed
model will be explored. Finally performance of the original
approach and our proposed will be compared, leading to
conclusion and planning for future research.
A. Review of Wireless Sensor Networks
Wireless sensor networks consist of distributed autonomous
sensors that are used for monitoring certain physical condition
such as temperature, pressure, sound, motion, etc. The network
system incorporates a gateway that provides wireless
connectivity to the distributed nodes and back to the wired
network [20]. WSNs are designed to perform a variety of
information processing functionalities including security and
military sensing [11], environmental monitoring [12], as well
as healthcare supporting [13]. Applications of WSNs can vary
significantly in requirements and deployment environments.
WSNs can play a critical role in military Commands, Controls,
Communications, Computing, Intelligence, Surveillances,
Reconnaissance, and Targeting (C4ISRT) systems [14]. They
can be used to, for instance, monitor friendly and enemy
forces, communicate commands, or sense different forms of
attacks. Environmental applications have been proposed for
WSNs. Common uses of these applications are: habitat
monitoring, animal tracking, forest-fire detecting, precision
farming, and disaster relief. Embedding WSNs in natural
environment enables researchers to collect long-range data on
previously unattainable scales. WSN applications are
frequently employed in extreme environments, such as the
Antarctic Pole. Several prototypes and commercial WSN
applications have been developed and applied for healthcare
monitoring for the elderly, children and chronically ill people.
One of the most popular categories include monitoring daily
activities for cases of fall, movement detection, location
tracking, medical status monitoring, and meditation intake
[13], [20]. The present work significantly improves the current
state of research [21] to accelerate optimal search and pareto
optimality for real-time applications.
B. Issues Pertinent to Security in Wireless Sensor Networks
Wireless Sensor Networks need using effective protocols
and algorithms in order to be dispersed efficiently and securely
in variety of applications. New protocols design and test are
challenging, their security requirements are among the most
important aspect which need to be clearly analyzed and
defined. [15]-[19] reviews most occurring security attacks
against Wireless Sensor Networks. The attacks can be grouped
into five classes of selective forwarding, sinkhole attacks, Sybil
attacks, wormholes, and HELLO flood attacks.
C. Brief description of network layer attacks in WSN
Selective Forwarding: In a selective forwarding attack,
malicious nodes may assault WSN by refusing to forward
messages, or simply dropping packages received.
Sinkhole Attacks: In a sinkhole attack, a malicious node
attracts all the traffic from a particular area by making itself
look especially attractive to surrounding nodes with respect to
the routing algorithm.
The Sybil Attacks: In Sybil attack, a malicious node
disguises as other nodes by means of impersonating other
nodes, claiming false identities, or worst, generating a large
number of additional node identities using only one physical
device.
Wormholes: In wormhole attacks, a malicious node located
near the base station can tunnel messages over a low latency
link and thus, completely disrupting message traffic.
HELLO Flood Attack: In HELLO flood attacks, attacker
broadcasting routing or other information with stronger enough
transmission power could convince every node in the network
that the adversary is its neighbor.
This research focuses on trust and reputation systems
security technique. There are other techniques used to enhance
security in WSNs such as IDS or Intrusion Detection Systems,
and SRP or Secure Routing Protocols.
II. A NEW MODEL OF SECURITY
Trust and Reputation System management is an innovative
solution for maintaining a minimum security level between two
entities having transactions or interactions within a distributed
system. Trust is a particular level of the subjective probability
)  +

with which an agent will perform a particular action; while a
reputation [6] is an expectation about an agent's behavior based
on information about it or observations of its past behavior. In
) (1  )  (2)
most cases, these two terms are not distinguished explicitly and
could be used interchangeably. In WSN transactions, if we
define the sensors asking for services as client sensors, and
sensors providing services as server sensors, then the client
sensors will determine whether to have transactions with a
server sensor based on its trustworthiness or reputation. A trust
and reputation model is generally composed of five
components [1], [2]: gathering information, scoring and
ranking, selecting entities, having transaction, and reward or
punishment. Gathering information, the first component of a
trust and reputation system, is responsible for collecting
behavioral information about other entities, for instance peers,
agents, or paths. The information collected might come from
different sources [7]. It could be first-hand (direct observation
or own experience), or second-hand (information provided by
peers). Once information about an entity has been properly
aggregated and weighed, a reputation score is then computed
and given base on certain algorithm. The primary objective of
this procedure is to provide the clients a measurable approach
to decide which server node is most trustworthy. The next step
is selection of the most trustworthy or reputable server entity in
the community by the client, and commencement interaction
with it demanding certain service and followed by further
interactions. After receiving the service provided, the client
will access the result and give a score of satisfaction. Based on
the satisfaction obtained, the last step, punishing or rewarding,
is carried out. If a server node is unsuccessful in satisfying the
client with the service provider, its reputation score will suffer,
and the client is likely to abstain from engaging in any future
transaction. The following sections will introduce two
innovative trust and reputation models, namely BTRM-WSN
[3], and Peer Trust [4].
A. Bio-inspired Trust and Reputation Model (BTRM-WSN)
BTRM-WSN [3], [21] carries out the selection of the most
trustworthy node through the most reputable path offering a
certain service. It is based on a bio-inspired algorithm called
Ant Colony System (ACS) [8], where ants build paths in order
to fulfill certain conditions graphically. These ants leave some
pheromone traces that help next ants to find and follow those
routes. These pheromone values will help ants find the optimal
route solutions since the optimal path will have the largest
amount of pheromone value. When we apply this ACS
algorithm onto trust and reputation system, we use
"pheromone value" to represent the trustworthiness of sensors.
In this BTRM-WSN, each sensor contains pheromone traces
for its neighbors ([0,1]), which determines probability for an
ant to select a path as well as the senor the path leading to as a
solution. In other word, can be considered as the trust that a
sensor gives another. The detailed steps and algorithms of
BTRM-WSN are as follows [3], [21]:
Gathering information: A set of artificial ants are created;
and then they leave the client sensor. When an ant moves from
a sensor i to sensor j, it gives a command for these two sensors
to modify the pheromone value of the path between them via
Eq. 1 and Eq. 2.
 = (1
(1)
 = (1 + (1
ij is the pheromone value of the path between sensor i and
sensor j; is the convergence value of ij; and is a parameter
controlling the amount of pheromone left by the ants.
When traveling along a WSN searching for the most
trustworthy routing to the server providing good service, each
ant needs to decide whether to stop and return the solution to
the client, or continue to find another one based on the
reputability of the server that is has found. When ant k arrives
at sensor s, several situations may occur. The first is that sensor
s provides services. If sensor s has more neighbors that have
pheromone value ( ) of the path followed by ant k from the
not been visited by ant k, then k computes the average
client until the sensor s is calculated,  [0,1]). If  is greater
that defined transition threshold (TraTh), then ant k stops and
returns the solution, vice versa. If sensor s has no more
neighbors or all of them have been visited, ant k stops and
returns solution. Another situation is that s does not provide any
services. If sensor s still has neighbours that have not been
visited by ant k yet, then k decides the next node to move. If
sensor s has no more neighbors or all of them have been
visited, ant k reaches a dead end. It has to go back to the route
that it has built until it arrives at: 1) a sensor offering the
requested service; or 2) a sensor not offering the requested
service but having more neighbors not visited yet. Scoring and
ranking: Client will examine and assess the quality of the
solution brought back by each launched ant. The quality of path
could be calculated by Eq. 3.
Q(S ) = %A


( ) !"
(3)
ij is the pheromone value of the path between sensor i and
sensor j; is the convergence value of ij; and is a parameter
controlling the amount of pheromone left by the ants.
S denotes the solution brought back by ant k, it is
Q(S ) denotes the quality of path S ;  denotes the average
represented the path leading to the sensor selected;
path pheromone of path S found by and k; PLF[0,1] denotes
a path length factor; and %A denote the percentage of ants
that have selected the same solution as ant k. After computing
the path quality of all solution brought back by ants, the client
selects the path with highest score and store it as Current_Best
solution. Then the client compares the path quality with the
best solution (Global_Best) found by earlier transactions. If
Current_Best solution is even better, then the client will
replace the previous Global_Best with the Current_Best
solution. Then an extra ant is sent to modify the pheromone
value of the current Global_Best.
Having Transaction: After the client selects the
Global_Best solution, it will have transaction with the selected
sensor. After receiving the service, the client will compare it
with the default service which the client expects to obtain.
There are two possible scenarios: first, the selected server
sensor might be completely benevolent and provide the exact
same service as it is supposed to; or it could be totally
malicious and provide extremely difference service. In the
former scenario, the client is satisfied and will give a
satisfaction value (Sat) as a random number between PunTh
and 1; while in the latter scenario, the satisfaction value (Sat)
is determined as a random number between 0 and PunTh as the
client is considered as unsatisfied. PunTh is a predefined
punishment threshold value.
Giving Reward or Punishment: A client will request the
desired service to what it things to be the most trustworthy
server via the most reputable route. Then punish or reward will
be given to all link in this route based on whether the client is
satisfied with the service provided by the server. This is done
by increasing or decreasing the pheromone value of the path.
B. Peer Trust System
The Peer Trust Model [3],[4], [21] a dynamic peer-to-peer
trust and reputation model, initially aims at estimating and
evaluating the trustworthiness, or goodness, of a peer in an e-
business environment. It identifies five factors related to trust
and reputation management for computing the trustworthiness
value of a given peer, namely: 1) the feedback a peer retrieves
from others; 2) the feedback scope, or field (number of
transactions); 3) the credibility factors of the source; 4) the
transaction context factor addressing the criticalness of
transactions; as well as 5) the community context factors
interpreting related characteristic. In a given WSN, the trust
value of a peer u could be computed via Eq. 4.
T(u) = 234 S(u, i) CR(p(u, i)) TF(u, i) + CF(u)
0(1)
(4)
T(u) denotes the value of trustworthiness of peer u.
denotes the weight factor for the collective evaluation, a
weighted average of amount of satisfaction that peer u receives;
and denotes the weight factor for the community context
factor. I(u) denotes the total number of transactions that peer u
has had with all other peers. S(u,i) denotes the normalized the
amount of satisfaction which peer u receives in its ith
transaction. Some existing reputation-based systems simply
apply a binary reputation mechanism to evaluate the degree of
satisfaction, where 0 represents unsatisfied, while 1 represent
satisfied. However, this evaluating system may not function
well [9], since malicious node may hide its misbehavior via
increasing its transaction volume. Taking that into
consideration, a normalized satisfaction value is applied, where:
(1,2)6789:;
S(u, i) = , and S(u, i)<2=>? represent the binary
0(1)
satisfaction (0 or 1) peer u received in its ith transaction. Also,
CR(p(u,i)) denotes the credibility of the feedback that peer u
receives from the ith peer (p(u,i)) it has transaction with. The
reason why credibility of the feedback is important is that a
peer may make false statements for other peers. For instance, a
malicious sensor may give low satisfaction for a benevolent
sensor, or give high satisfaction for a malicious sensor. As a
result, a credibility of feedback should be introduced; and
feedback with higher credibility should be weighed more in
computing the trustworthiness of a give peer. Two mechanisms
evaluating the credibility of feedback is introduced. Eq. 5.
illustrates the first mechanism, and the metric using this one to
measure credibility of feedback is called TTVM (where TVM
stands for Trust Value Measurement). This mechanism uses the
trust value of peer p(u,i) to evaluate the trustworthiness of
feedback received from it to peer u.
CR@p(u, i)A = D(E)
B(C(1,2))
7FG B(C(1,2))
(5)
Eq. 6. introduces a personalized similarity measurement
this one to measure credibility of feedback is called THI .. If
(PSM) to rate the credibility of peer (u,i), and the metric using
we use w to represents the client sensor which wants to test
whether u is trustworthy, then Sim(p(u,i),w) is used to
measures the personalized similarity between peer w and

peer(u,i). This is computed via Eq. 7., where IJS(p(u,i).w) is
the common set of peers which peer w and peer p(u,i) have had
the set of peers; IJS(p(u, i), w) denotes the two vectors of
transaction with in the past, and x denotes a peer belonging to
feedback by peer w and peer p(u,i); and
Q(R,S(T,U)) Q(R,X)
UFG M(N,O(P,)) UFG M(N,W) Y
( )
V(N,O(P,)) V(N,W)
denote
deviation of the two feedback vectors.
CR(p(u, i), w) =
2Z(C(1,2),[)
D(E)
\FG 2Z(C(1,2),[)
D(`,a(E,7))
_(`,a(E,7))
Sim(p(u, i), u) = ^(j0f(C(1,2).1)
( 7FG
D(`,b)
hhhhhhhhhhhhhhhhhhhhhhhhhhhi|
|0f(C(1,g),1)
TF(u,i) denotes the transaction context factor of peer us ith
transaction. This is another importance factor since each
transaction may differ from one another. Transaction contexts
including size or category can have influence on the TF(u,i)
value. Finally, CF(u) denotes the community context factor
which could be applied to address some community-specific
issues, for instance, incentive problem.
III. ENHANCED BIO-INSPIRED TRUST AND REPUTATION
SYSTEM (BTERM)
In this section we proposed an Enhanced bio-inspired trust
and reputation system inspired by BTRM-WSN and Peer Trust
system demonstrated in previous section. As mentioned in
earlier, the criterion that BTRM-WSN used to determine
whether a sensor is trustworthy is the value of the average
pheromone value of the solution path from the client sensor to
the sensor selected. Similarly, the quality of each solution is
evaluated based on the average pheromone value, the length of
those solution paths, as well as the number of ants which have
found the same solution. This approach has been proven to be
quite effective; however, the performance of this system may
get improved if the condition of the server sensors could be
taken into account. So, we are considering the possibility of
adding some principles in Peer Trust system into the original
BTRM-WSN, and to examine if this will modify some aspects
of the original system. The first modification is that when an
ant determines whether a sensor that it reaches is trust worthy
or not, it not only compares the average pheromone of the path
leading to the sensor to a predefined threshold value, but also
the trust value of that sensor which is computed by a simplified
Peer Trust algorithm. If the trust value of the sensor is larger
than a user-defined threshold value, then the sensor will be
considered as trustworthy, and vice versa. To compute the
trustworthiness of a sensor by applying Peer Trust principles,
only satisfaction that a sensor received from other sensors with
which it had transactions is considered. The original algorithm
(Eq. 8) is not applied here as it consumes too much resources.
For instances, more memory are required to hold the data, and
more energy is consumed to perform all these calculation.
Even though the performance in finding benevolent server
sensors might be better, we still want the proposed system to
be energy efficient.
T(u) = 234
0(1) (1,2)
0(1)
(8)
Similar to the original algorithm used in Peer Trust system,
the first step is to find a set of active server sensors. Then to
compute the trustworthiness of a specific sensor (sensor u in
our case), the client collect the satisfaction value that sensor u
the standard
received from its previous transactions. I(u) denotes the total
number of transaction that sensor u has had before. Here
(6) instead of using a random number between 0.5 and 1 to
represent satisfaction, and between 0 and 0.5 to represent
D(`,E)
_(`,E)
c 7FG )d
dissatisfaction, we simply use 1 and 0 to simplify the process
D(`,E)
(7) to make the result more stable. The second modification is that
when the client determines which sensor is more trustworthy, it
not only considers the path quality which is computed via Eq.
3., but also the trust value computed via Eq. 8. A sensor has to
outperform others in terms of both path quality and
trustworthiness to be selected as the final solution.
IV. PERFORMANCE ANALYSIS
We conduct three experiments simulating the proposed
EBTRM system in order to evaluate its performance and to
compare it with the other two original approaches, namely
BTRM-WSN, and Peer Trust system. The first experiment
aims at comparing the three systems in terms of the accuracy in
searching for trustworthy sensors, which evaluates the level of
security provided; the second test compares the average path
length leading to the trustworthy sensors selected, which
evaluates the efficiency, or the easiness in finding trustworthy
sensors, of the three systems; and finally, the overall energy
consumption of the WSNs applying these systems, which
evaluates the energy efficiency of the systems is also
compared.
A. TRMSim-WSN
TRMSim-WSN [10], [21] is a Java-based trust and
reputation model simulator aiming at providing an easy way to
test a trust and reputation model over WSNs and to compare it
against other models. Work has been done to develop similar
model in C++. Partial test carried out indicated that the
execution time is significantly reduced. On this basis, a WSN
template was designed using the same Network Parameter
settings in TRM-WSN as: Clients: 15%, Relay Servers: 5%,
Radio Range: 10, Min. Number of Nodes: 200, Max Number
of Nodes: 200, Number of Network 500, Number of
Execution: 100. Then the simulator will randomly create WSN
for experiments based this template. Note that 15% of all nodes
in a randomly created WSN are clients which will request
default services. The other 85% nodes will act as servers which
will be asked to provide services upon request. The following
subsections provide details of tests being conducted using new
software environment.
B. Accuracy
Here we use the concept accuracy to evaluate the
reliability and level of security provided by the trust and
reputation system. The accuracy of a trust and reputation
system is represented by the percentage that the number of
times when it successful selects trustworthy sensors (the
former situation) out of the total number of transactions. A
better trust and reputation system should have a good control
of the negative influence which the malicious nodes have on
the WSN. TABLE 1. compares Accuracy of the BTRM-WSN,
Peer Trust system, as well as the proposed EBTRM system.
From the previous discussion, we can conclude that when the
percentage of malicious sensors is not high (less than 40%), the
difference between these three trust and reputation system in
terms of accurately finding trustworthy sensor is not
significant, and they can all achieve accuracy of 98%.
However, when the situation is getting more unsecured (the
percentage of malicious sensor is getting higher), we can tell
that the proposed EBTRM can provide the highest accuracy
and thus the highest level of reliability and security.
TABLE 1. ACCURACY IN SEARCHING FOR TRUSTWORTHY SENSORS,
COMPARISON OF BTRM-WSN, PEER TRUST, AND EBTRM
Accuracy in Percentage of Malicious Sensors
Searching for
Trustworthy
10% 20% 30% 40% 50% 60% 70% 80% 90%
Sensors (%)
BTRM-
99.88 99.63 99.35 99.03 98.72 98.01 97.33 96.01 93.01
WSN
TRM Peer
Trust
99.87 99.48 98.88 98.55 98.07 97.45 96.55 92.01 67.79
EBTRM 99.95 99.87 99.73 99.35 99.04 98.64 97.89 96.83 96.17
The comparative results of TABLE 1 with the three security
models of Bio-inspired Trust and Reputation Model, Peer Trust
Model, and Enhanced Bio-inspired Reputation Model confirms
that EBTRM has higher accuracy in searching for trustworthy
sensors. A better trust and reputation system should have a
good control of the negative influence which the malicious
nodes have on the WSN. TABLE 1 compares the BTRM-WSN,
Peer Trust system, as well as the proposed EBTRM system in
terms of accuracy
C. Path Length
Path length is the average hops leading to the most
trustworthy sensors which are selected by the client in a WSN
applying a certain type of trust and reputation system. It is
assumed that less average path length indicates a better
performance in efficiency and easiness in searching for
trustworthy sensors of a trust and reputation system. This is
because: 1) less number of intermediaries means higher
security level and less energy consumption; and 2) shorter path
length implies that it is easier to find trustworthy nodes and
thus, server nodes will response quicker to client nodes. Fig. 2.
will compare the BTRM-WSN, Peer Trust system, as well as
the proposed EBTRM system in terms of average path length
leading to trustworthy sensors. BTRM-WSN has the best
performance, and Peer Trust has the worst performance in
terms of shorting average path length. However, the difference
between BTRM-WSN and our proposed EBTRM is not
significant. Thus in smaller size WSN, we can consider that
they can both provide an easy and efficient approach in
searching trustworthy sensors.
TABLE.2. presents a comparison between average path
lengths leading to trustworthy sensor. The BTRM holds the
minimum of average path length. However, Enhance Bio-
inspired Trust and Reputation Model is close to the minimum
values of BTRM.
Table 2: average path length leading to trustworthy sensors,
COMPARISON OF BTRM-WSN, PEER TRUST, AND EBTRM
Accuracy in Percentage of Malicious Sensors
Searching for
Trustworthy 10% 20% 30% 40% 50% 60% 70% 80% 90%
Sensors (%)
BTRM-
2.22 2.25 2.23 2.26 2.29 2.39 2.56 2.74 3.62
WSN
TRM Peer
6.45 6.52 6.55 6.57 6.6 6.65 6.72 6.78 7.05
Trust
EBTRM 2.97 3.00 3.06 3.07 3.18 3.16 3.24 3.36 4.23
D. Conservation of Energy and Its Consumption
Energy consumption of the network is the overall energy
consumed in: 1) client nodes sending request messages; 2)
server nodes sending response services; 3) energy consumed
by malicious nodes which provide bad services; 4) relay nodes
which do not provide services; and 5) the energy to execute the
trustworthy sensor searching process of a certain trust and
reputation system. How to effectively reduce energy
consumption is a major issue in WSN researches.
TABLE 3. compares the BTRM-WSN, Peer Trust system, as
well as the proposed EBTRM system in terms of overall
network energy consumption. Accordingly, Peer Trust system
outperforms BTRM-WSN and the proposed EBTRM in terms
of energy efficiency significantly. EBTRM is slightly less
energy efficient than BTRM model, when the percentage of
malicious sensors is getting higher.
TABLE 3: OVERALL NETWORK ENERGY CONSUMPTION, COMPARISON OF
BTRM-WSN, PEER TRUST, AND EBTRM
Accuracy in Percentage of Malicious Sensors
Searching for
Trustworthy 10% 20% 30% 40% 50% 60% 70% 80% 90%
Sensors (%)
BTRM-
1.42 1.60 2.22 2.63 3.00 3.75 4.31 5.75 11.42
WSN
TRM Peer
0.28 0.33 0.38 0.42 0.47 0.51 0.59 0.71 1.05
Trust
EBTRM
2.3 2.6 2.96 3.41 3.92 4.4 4.76 5.87 12.21
E. Overall Performance Analysis
From these simulations we conducted, conclusion could be
drawn that our proposed EBTRM system successfully
increases the accuracy in finding trustworthy sensors and thus
the level of security of the original BTRM-WSN without
sacrifice its advantage in finding trustworthy sensors
efficiently, and the extra amount of energy required for those
add-ons is acceptable. EBTRM is proven to be able to
accurately distinguish benevolent sensors from malicious
sensors and thus protect WSNs from attackers. And more
importantly, the level of security it provides is not influenced
by the number of attackers as much as its two competitors do.
The trade-off is that, especially when the network is in a
relatively secured status (where there are only a small number
of malicious sensors), it becomes more complicated and less
energy efficient to search for trustworthy sensors because of
the extra conditioning and computation. However, the
difference is not significant. Overall, the modification is
successful, and our proposed EBTRM provide a better solution
to WSNs where a high level of security is required.
 V. CONCLUSION
In this study, two current solutions related to trust and
reputation systems namely Bio-inspired Trust and Reputation
Systems, and Peer Trust Systems for Wireless Sensor
Networks are reviewed. In order to produce stronger security
environment for WSNs an enhanced model of bio-inspired
trust and reputation model is designed and implemented. The
objectives of the enhanced model are to guarantee an
increased level of security within wireless sensor networks
without compromising the behavior of the original model as
well as keeping the energy consumption to its lowest possible.
In this investigation, based on the results presented earlier,
it is evident that the proposed enhanced system is offering an
increased level of accuracy in detecting trustworthy sensors
which in return increases the security level of wireless sensor
networks without sacrificing efficiency in finding trustworthy
sensors, in comparison with the original model and only with
an insignificant and acceptable level of energy consumption
increase. Comparing to the other two earlier models the
proposed enhanced model is able to find trustworthy sensors
with greater accuracy. On the other hand BTRM operates with
slightly less average distance between client and selected
trustworthy sensors as well as overall network energy
consumption is slightly less as compared with EBTRM.
In conclusion the enhanced mechanism of EBTRM is able
to improve security level by increasing the accuracy in finding
the trustworthy sensors. The research will continue to improve
the ease of search and energy consumption while maintain
enhanced security. This work significantly shortened
execution time through changing software environment.
Considering multi-objectives decision making, the optimum
solution depends on the particular requirements, which may
differ for different environment or applications. Therefore
developing pareto optimality facilitates selection of optimum
solutions. Ideally the new extension will be conducted using
HPC resources to develop near real-time solution.
REFERENCES
[1] Marti, S. and H. Garcia-Molina. "Taxonomy of Trust:
Categorizing P2P Reputation Systems." Computer Networks 50.4 (2006):
472-484.
[2] Mrmol, F. G. and G. M. Prez. "Towards Pre-Standardization of
Trust and Reputation Models for Distributed and Heterogeneous Systems."
Computer Standards' Interfaces 32.4 (2010): 185-196.J. Clerk Maxwell, A
Treatise on Electricity and Magnetism, 3rd ed., vol. 2. Oxford: Clarendon,
1892, pp.6873.
[3] Mrmol, F. G. and G. M. Prez. "Providing Trust in Wireless
Sensor Networks Using a Bio-Inspired Technique." Telecommunication
Systems 46.2 (2011): 163-180.
[4] Xiong, L. and L. Liu. "PeerTrust: Supporting Reputation-Based
Trust for Peer-to-Peer Electronic Communities." Knowledge and Data
Engineering, IEEE Transactions on 16.7 (2004): 843-857.
[5] Gambetta, D. "Can We Trust Trust?" Trust: Making and Breaking
Cooperative Relations. Ed. D. Gambetta. Department of Sociology University
of Oxford, 2000. 213-237.
[6] Abdul-Rahman, A. and S. Hailes. "Supporting Trust in Virtual
Communities." System Sciences, 2000. Proceedings of the 33rd Annual
Hawaii International Conference on . London, UK: IEEE, (2000)September.
[7] Srinivasan, A., et al. "Reputation and Trust-based Systems for Ad
Hoc and Sensor Networks." On Trust Establishment in Mobile Ad-Hoc
Networks. Ed. A. Boukerche. Wiley & Sons, 2007.
[8] Dorigo, M., Birattari, M and Stutzle, T.,. "Ant Colony
Optimization." Computational Intelligence Magazine, IEEE 1.4 (2006), 28-39.
[9] Dellarocas, C. "Analyzing the Economic Efficiency of EBay-Like
Online Reputation Reporting Mechanisms." Proceedings of the 3rd ACM
conference on Electronic Commerce. NY, USA: ACM, (2001), 171-179.
[10] Mrmol, F. G. and G. M. Prez. "TRMSim-WSN, Trust and
Reputation Models Simulator for Wireless Sensor Networks."
Communications, 2009. ICC '09. IEEE International Conference on. Dresden:
IEEE, (2009), 1-5.
[11] Callaway, E. H. Wireless Sensor Networks: Architectures and
Protocols. Boca Raton, FL: Auerbach Publications, (2004).
[12] Zhao, F. and L. J. Guibas. Wireless Sensor Networks: An
Information Processing Approach. Amsterdam: Morgan Kaufmann, (2004).
[13] Alemdar, H. and C. Ersoy. "Wireless Sensor Networks for
Healthcare: A Survey." Computer Networks 54.15 (2010): 2688-2710.
[14] Martincic, F. and L. Schwiebert. "Introduction to Wireless Sensor
Networking." Handbook of Sensor Networks: Algorithms and Architectures.
Hoboken, NJ, USA: John Wiley & Sons, (2005).
[15] Dimitrievski, V. P. and D. Davcex. Security Issues and
Approaches in WSN. Skopje, Republic of Macedonia: Department of
computer science, Faculty of Electrical Engineering and Information
Technology, (2011).
[16] Gerrigagoitia, K.; Uribeetxeberria, R.; Zurutuza, U.; Arenaza, I.,
Reputation-based Intrusion Detection System for wireless sensor networks,
Complexity in Engineering (COMPENG 2012), June 11-13, (2012), Aachen.
[17] Hortos, W.S., Bio-inspired, cross-layer protocol design for
intrusion detection and identification in wireless sensor networks, 2012 IEEE
37th Conference on Local Computer Networks Workshops (LCN Workshops),
Oct 22-25, (2012), 1030-1037.
[18] Xiaoyong Li, Feng Zhou, Junping Du, LDTS: A Lightweight and
Dependable Trust System for Clustered Wireless Sensor Networks, IEEE
Transactions on Information Forensics and Security, June (2013), Vol. 8 , Iss.
6, pp. 924-935.
[19] Naveed, A.; and Kanhere, S.; Authntication and Confidentiality in
Wireless Ad Hoc Networks; Chapter 1 of: Beyah, R,; McNair, J; and Corbett,
C.; Security in Ad Hoc and Sensor Network, ISBN: 13 978 981 4271 08 0,
(2010), World Science.
[20] Dargie, W.; and Poellabauer, C.; Fundamentals of wireless Sensor
Networks: Theory and Practice; John Wiley (2010), ISBN: 978 0 470 99765 9
[21] Marzi, H.; and Li, M.; An Enhanced Bio-Inspired Trust and
Reputation Model for Wireless Sensor Network, Elsevier Procedia Computer
Science, June (2013), Pages 11591166, Vol. 19.
[22] Marzi, H. Sensor Fusion and Use of Reconfigurable Neural
Networks in Condition Monitoring, Proceedings of IEEE International
Conference on Computational Intelligence for Measurement Systems and
Applications, CIMSA 2005, July 20-22, (2005), pp. 145 150, Giardini
Naxos, Sicily, Italy.
[23] Marzi, H., Modular Neural Network Architecture for Precise
Condition Monitoring, IEEE Transactions on Instrumentation and
Measurement, April (2008), Vol. 57, No. 4.