International Journal of Scientific & Engineering Research, Volume 4, Issue 4, April-2013 212

ISSN 2229-5518

Faulty Node Detection in Wireless Sensor

Networks Using Cluster

Srikanta Kumar Sahoo

Abstract- Since the accuracy of data is important to the whole system’s performance, detecting nodes with faulty readings is an essenti al issue in network management. Removing nodes with faulty readings from a system or replacing them with good ones improve the whole system’s performance and at the same time prolong the lifetime of the network. In general, wireless sensor nodes may experience two types of fault s that would lead to the

degradation of performance. One type is function fault, which typically results in the crash of individual nodes, packet loss , routing failure or network partition. The other type of error is data fault, in which a node behaves normally in all aspects except for its sensing results, leading to either significant biased or random errors.

Index Terms- W ireless Sensor Networks, Data Fault Detection, Functional Fault.

Introduction

Wireless Sensor Networks have emerged as an important new area in wireless technology. A wireless network consisting of tiny devices which monitor physical or environmental conditions such as temperature, pressure, motion or pollutants etc. at different areas. Such sensor networks are expected to be widely deployed in a vast variety of environments for commercial, civil, and military applications such as surveillance, vehicle tracking, climate and habitat monitoring, intelligence, medica l, and acoustic data gathering. The key limitations of wireless sensor networks are the storage, power and processing. These limitations and the specific architecture of sensor nodes call for energy efficient and secure communication protocols. The k ey challenge in sensor network is to maximize the lifetime of sensor nodes due to the fact that it is not feasible to replace the batteries of thousands of sensor nodes. Therefore, computational operations of nodes and communication protocols must be made as energy efficient as possible. Since the accuracy of data is important to the whole system’s performance, detecting nodes with faulty readings is an essential issue in network management. The accuracy of individual node’s readings is crucial; the readings of sensor nodes must be accurate to avoid false alarms and missed detections. Some applications are designed to be fault tolerant to some extent, removing nodes with faulty readings from a system with some redundancy or replacing them with good ones can still significantly improve the whole system’s performance and at the same time prol ong the lifetime of the network. To conduct such after deployment maintenance (e.g., remove and replace), it is essential to investigate methods for detecting faulty nodes

Wireless Sensor Network

Wireless sensor networks are potentially one of the most important technologies of this century. Recent advancement in wireless communications and electronics has enabled the development of low-cost, low-power, multifunctional miniature devices for use in remote sensing applications. The combination of these factors has improved the viability of utilizing a sensor network consisting of a large number of intelligent sensors, enabling the collection, processing analysis and dissemination of valuable information gathered in a variety of environments. A sensor network is composed of a large number of sensor nodes which consist of sensing, data processing and communication capabilities.
The fundamental objectives for sensor networks are reliability, accuracy, flexibility, cost effectiveness and ease of deployment.

Sensor Faults

In this chapter, we will review some of the commonly observed sensor faults . Faults in sensor data can occur for many reasons. The first source stems from unpredictable environmental conditions, which can often cause sensors to behave erratically. Factors such as extreme temperatures or precipitation can affect sensor performance. During a deployment, sensors can be displaced or change orientation. This can be caused, for example, by animals or humans or wind or if they are deployed in a body of water that freezes. Another type of fault occurs when the environmental conditions travel outside the range of values the sensor is able to detect. In this case, the sensor becomes saturated and is unable to report the true readings.

IJSER © 2013 http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 4, Issue 4, April-2013 213

ISSN 2229-5518

In the following taxonomy, we discuss both the manifestation of faults in the data as well as specific reasons that different types of faults occur. After the fault types are introduced, we discuss the problem formulation and assumptions that will be made
throughout this thesis.

Clustering in WSN

It is widely accepted that the energy consumed in one bit of data transfer can be used to perform a large number of arithmeti c operations in the sensor processor. Moreover in a densely deployed sensor network the physical environment would produce very similar data in near-by sensor nodes and transmitting such data is more or less redundant. Therefore, all these facts encourage using some kind of grouping of nodes such that data from sensor nodes of a group can be combined or compres sed together in an intelligent way and transmit only compact data. This can not only reduce the global data to be transmitted and localized most traffic to within each individual group, but reduces the traffic and hence contention in a wireless sensor network. This process of grouping of sensor nodes in a densely deployed large-scale sensor network is known as clustering. The intelligent way to combined and compress the data belonging to a single cluster is known as data aggregation.
There are some issues involved with the process of clustering in a wireless sensor network [12]. First issue is, how many clusters
should be formed that could optimize some performance parameter [14]. Second could be how many nodes should be taken in to a single cluster. Third important issue is the selection procedure of cluster-head in a cluster. Another issue that has been focused in many research papers is to introduce heterogeneity in the network [17]. It means that user can put some more powerful nodes, in terms of energy, in the network which can act as a cluster-head and other simple node work as cluster- member only. Considering the above issues, many protocols have been proposed which deals with each individual issue. Motivation of the work
Sensor nodes are widely used in surveillance, vehicle tracking, climate and habitat monitoring, intelligence, medical, and acoustic data gathering. The accuracy of data is important to the whole system’s performance, detecting nodes with faulty readings is an essential issue in network management. The work in this dissertation is motivated by the problem of detecting the faulty sensors nodes in the WSN (Wireless Sensor Network).

Objective of the Work

Motivated by the need of a fault detection algorithm for WSN (Wireless Sensor Network), the objective of this work is given as follows:
(a) To propose a frame work for fault detection in WSN.
(b) To propose a faulty node detection mechanism.
(c) To propose a clustering approach to devise an efficient fault detection algorithm. (d) To study and validate the performance of the proposed fault detection algorithm
through simulation using MATLAB.
(e) To compare the proposed approach with the existing approach.

LITERATURE SURVEY

Review of the literature on fault detection in sensor networks

The approach presented in [13] describes faulty sensor nodes are identified based on comparisons between neighboring nodes and dissemination of the decision made at each node. Nodes with malfunctioning sensors are allowed to act as a communication node for routing, but they are logically isolated from the network as far as fault detection is concerned. It employs local comparisons of sensed data between neighbors and dissemination of the test results to enhance the accuracy of diagnosis. Transient faults in communication and sensor reading are tolerated by using time redundancy. Faulty nodes are isolated by correctly identifying fault-free nodes. Both the network connectivity and accuracy of diagnosis are taken into account since fault-free nodes isolated might be of little or no use even if they are determined to be fault-free, unless they can participate in the network via intermediate communication nodes with faulty sensors.
Sensor faults have been studied extensively in [21] in the context of fault diagnosis for systems. The problem of tolerating and modeling sensor failures was also studied in [Mar90]. However, studying faults in wireless sensing systems differs from the issues studied in these references in several ways that make the problem more difficult. The first issue is that sensor networks may involve many more sensors over larger areas. Second, the phenomena being observed by sensor networks are not well controlled or measured in most cases. Instead, they can be very heterogeneous, resulting in higher uncertainty when modeling sensor behavior and sensor faults.
The distributed fault detection scheme have been studied in [12], it checks out the failed nodes by exchanging data and mutually
testing among neighbor nodes in the network, but the fault detection accuracy of a DFD (Distributed Fault Detection) scheme
would decrease rapidly when the number of neighbor nodes to be diagnosed is small and the node’s failure ratio is high.

IJSER © 2013 http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 4, Issue 4, April-2013 214

ISSN 2229-5518

The author of this[15] paper uses a heartbeat based testing mechanism to detect failure in each cluster and take the advantage of cluster based architecture to forward the failure report to other cluster and their respective members.
The work in [14] use a cluster-based communication architecture to permit the FDS (Failure Detection Service) to be implemented in a distributed manner via intra-cluster heartbeat diffusion and to allow a failure report to be forwarded across clusters through the upper layer of the communication hierarchy. In doing so, we extensively exploit the message redundancy that is inherent in ad hoc wireless settings to mitigate the effects of message loss on the accuracy and completeness properties of failure detection.
The work in [22] identifies a small number of sensor faults observed in real deployments. The faults are briefly defined, and different methods of offline detection of these faults are examined. The performance of each of these fault detection methods are analyzed using three real deployments.

PROPOSED SYSTEM

System and Fault Model

We assume that the wireless ad hoc network is a large connected network in which there are totally N sensor nodes denoted by

1, 2, 3… N. The nodes are distributed randomly in some physical domain and become stationary after deployment. The transmission range for each node is fixed and link between two hosts is bi-directional. If host u is in the transmission range of another host v, then there
must be a link between
the two. The system can be modeled as a communication graph G =
{V, E}, where V= {1,
2….N}, and E= {(v1, v2):
v1 is in transmission
range of v2 and vice versa}.
A cluster is a unit disk with a radius equal to the center node’s transmission range. As a result, any non-center nodes in a cluster are one- hop neighbors of the center node. The center node is called the cluster head (CH), while a node that is a one hop neighbor of the CHs of two different clusters can become the gateway (GW) node (see Figure 1) . After the autonomous cluster formation, only CH and GW node, which are elected in a fully distributed fashion,
For any unselected node v
{
If ((node v is an indispensable node) || (node v is the only node with highest quality Qv among
unselected neighbor) || (among unselected neighbor with same quality node v is with the smallest ID)
{
Update status to selected; Regard itself as a CH;
Send an invite packet, invite (v) to all neighbors ;
}
On receiving an invite packet from neighboring node v
If (node u is an indispensable node) Discard this packet;
Else
{
Regards itself as an ordinary node;
Updates status to selected;
Sends a join packet, join (u,v) to join the cluster constructed by v;
If (more than one such packets are received) Join the one with smallest ID;
Else
Joins sender with largest logical degree; Regards itself as a gateway node;
}
On receiving a join packet sent from neighboring node u decreases the logical degree by 1;
}
}
participate in the inter-cluster communication (see Figure 1(b)), while ordinary members (OMs) in each cluster talk only to their
CHs (and to other members when necessary).
The proposed system is not fully distributed. The total number of nodes is equally divided into a number of clusters. Each
cluster has a CH and there is a GW node between two clusters to forward the message from one cluster to another. The cluster is controlled by the CH. The fault is detected by the CH in each cluster and the message is forwarded to all nodes of the clust er and also forwarded to other CH . All the clusters are operating simultaneously.

IJSER © 2013 http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 4, Issue 4, April-2013 215

ISSN 2229-5518


1: Intra-Cluster and Inter-Cluster Communication
In the fault detection of wireless sensor networks, we assume that all the sensor nodes have the same transmission range. Sensor
nodes can be randomly deployed or placed in predetermined locations. Nodes with faulty sensors and permanent
communication faults are to be identified. Sensor nodes which generate incorrect sensing data or fail in communication intermittently are treated as usable nodes, and thus are diagnosed as fault-free. Sensor nodes with malfunctioning sensors could participate in the network operation since they are still capable of routing information. Only those sensor nodes with a permanent fault in communication (including lack of power) are detected and this information is disseminated throughout the network and removed from the network.

Algorithm for cluster formation

This section describes the algorithm for cluster formation in the proposed system model. The algorithm is given in a table. Table 1: Algorithm for Cluster formation
The system model uses an existing method FIND (Faulty Node Detection) to detect nodes with data faults [11]. After the nodes in a network detect a natural event, FIND ranks the nodes based on their sensing readings as well as their physical distances from the event. A node is considered faulty if there is a significant mismatch between the sensor data rank and the distance rank.
There are three stages in the proposed system
(1) Map Division
(2) Detection Sequence Mapping
(3) Fault Detection

Map Division: The first stage of our proposed system is map division, in which the map is divided into a number of subareas named faces based on the topology of the network. Each face is uniquely identified by a distance sequence, which is denoted as a sequence of sensor node IDs (e.g., 2-1-3-4-5). Within a distance sequence, the IDs are sorted in order of nodes’ distances from an arbitrary point within this face. Map division can be pre-computed before detecting faulty nodes such that a number of distance sequences can be obtained.

For two sensor nodes 1 and 2, the perpendicular bisector Div (1,2) divides the area into two subareas. For any position point below Div (1,2), node 1 is closer than node 2, so the distance sequence is 1-2. For any position point above Div (1,2), node 2 is closer than node 1, so the distance sequence is 2-1.

Figure 2: Map Division

Detection Sequence Mapping: The second stage is detection sequence mapping, a number of events appear in the monitored region and are detected by the sensor nodes. For a single event, the sensing result of each node (e.g., the received signal strength or time-of-arrival) varies depending on how far the node is away from the event. A detected sequence is then obtained by

IJSER © 2013 http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 4, Issue 4, April-2013 216

ISSN 2229-5518

ordering the sensing results of all the sensor nodes. Without knowing the location of the event, this detected sequence is mapped with one of the distance sequences corresponding to the face in which the event most likely takes place, yielding an

estimated sequence. The same mapping process is repeated for all the events such that an estimated sequence is obtained for every detected sequence after this stage.
Figure 3: Map Division with Events and Nodes

Detected Sequences Distance Sequences


1-3-2-4-5 1-2-3-4-5 2-1-3-4-5 1-3-2-5-4

3-4-5-1-2 3-4-5-2-1
5-3-4-1-2

4-3-2-5-1 4-5-3-2-1
2-1-3-4-5
3-1-2-5-4
…………
3-1-5-2-4
Figure 4: Detection Sequence Mapping
Let the N sensor nodes divide the map into M faces, identified by a set of distance sequences
V = {S1, S2,.· · · ,SM}.
Suppose S’ is a detected sequence from a single event and S is the distance sequence corresponding to the face where the event takes place.
{A1,A2, · · · ,AM} denotes the size of the M faces, the probability that an event takes places within the ith face can be computed as
the following:

Pr(S = Si) = Pr (Si) = Ai , 1 ≤ i ≤ M (1) S can be estimated by the method of Maximum A Posteriori (MAP) estimation as

S’’MAP(S’) = args max Pr (S| S’) Si =S
= arg max Pr( S’|Si) Pr(Si) /Pr( S’|Sk)Pr(Sk) where SiV
= arg max Pr( S’|Si)Pr(Si) where SiV (2)

Fault Detection: The third stage is fault detection using Ranking Differences, after a sufficient number of mappings becomes available; a blacklist is then obtained by analyzing the inconsistencies between the detected sequences and the estimated sequences.

Average Ranking Differences

We can identify faulty nodes by using the average ranking difference based on the following two cases.

Case 1. A node with a larger average ranking difference has a higher probability of being a faulty node.

IJSER © 2013 http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 4, Issue 4, April-2013 217

ISSN 2229-5518

The ranking difference di(K) for node K in sample i is given in the following equation:
di (K) = |R ( Si1, K)−R( Si2, K)| 1 ≤ K ≤ N
Where R (*, K) denotes the ranking of node K in sequence *,Si1 is estimated sequence & Si2 is the detected sequence. Then, the average ranking difference of node K in the n samples (denoted as D(K)) is computed by averaging di(K):
D (K) = (3) D (K) is also known as the sample mean of the ranking differences.

Case 2. The majority of faulty nodes can be obtained by selecting nodes whose ranking differences are above a lower bound. A

node q is faulty if its average ranking difference D (q) is greater than a bound B given by
B = e + Ne)
Where Ne is the number of faulty nodes in an N-node network, μe is the arithmetic mean of the average ranking difference of
faulty nodes, and D (q) is calculated under a sufficient large sample size.

Fault Detection Algorithm

This section describes the algorithm for faulty sensor node detection in the proposed system model. The different notations used in this algorithm are presented in a table.

µe : arithmetic mean of average ranking difference of faulty nodes

Ne : number of faulty nodes

B : bounded value

J : variable used in for……end loop

N : total number of nodes

D (nj) : mean of ranking difference

Table 2: Meaning of Different Terms
Initially the blacklist is empty (Line 1) and Ne, μe and B are set to 0 (Line 2). Starting from n1 (which is the node with the largest average ranking difference) and check the nodes one by one (Line 3) until the ranking difference of a node is no greater than B

(Line 4 and 5). Specifically, a node can be added into the blacklist if and only if its ranking difference is greater than B. After adding a new node into the blacklist, Ne, μe and B are updated (Line 7 to 9). Then, an estimation of faulty nodes {n1, · · · ,nK} can be obtained if this loop breaks at node nK+1.
Input: Sorted node sequence n1, · · ·, nN, average ranking differences D
Output: The Blacklist {n1, · · · , nK}
1: Blacklist ← // Initialization
2: µe ←0, Ne ←0, B←0
3: for j ←1 to N do
4: if D (nj) ≤ B then
5: Break
6: else
7:µe
8: B (µe + Ne)
9: Blacklist Blacklist + {nj}
10: end if
11: end for
12: Return Blacklist
Table 3: Algorithm for Fault Detection

IJSER © 2013 http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 4, Issue 4, April-2013 218

ISSN 2229-5518

Example:

In the following example we consider five nodes 1, 2,3,4,5 and four events. The detected sequences and estimated sequences for different events are calculated.


Figure 5: Fault Detection
Figure 6: Fault Detection Using Ranking Differences
In the above Figure 6, four detected sequences are mapped with their estimated sequences and the ranking differences can be computed by comparing the rankings of each node in the two sequences. Let us take the square event as an example. For the square event, the detected sequence is 3-4-5-1-2 and the estimated sequence is 5-3-4-1-2, where nodes 1 and 2 have the same rankings such that their ranking differences are 0. Node 3 and 4 both shift by one and their ranking differences are 1. Node 5 ranks the first in estimated sequence and the third in detected sequence, and thus its ranking difference is 2. The ranking differences of all the nodes in all events are shown in the above figure. In this example, node 5 is a faulty node, with all its readings lower than expected. First, as a faulty node, node 5 has a non-zero ranking difference in most events due to its faulty reading. Second, node 5 also changes some normal nodes’ rankings. However, a faulty node at most changes normal nodes’ rankings by 1. Third, for different events, the sets of normal nodes whose rankings are changed by node 5 are different. The total (or average) ranking difference of node 5 is the largest and is the faulty node.

SIMULATION & RESULT ANALYSIS Simulation Model and Parameters

We implemented the proposed system using MATLAB. We setup a network having 25 sensor nodes and different numbers of events. All the nodes are randomly distributed throughout the terrain and divided into clusters. Each cluster having equal
number of nodes, a CH and events. Broadcasting events, identified by their unique event ids. After receiving a broadcasting packet, nodes measure the RSS and record it together with event id. The number of events generated varies from 19 to 49. After recording the RSS of all the events, we randomly select 1 to 5 nodes and inject errors into thei r readings, corresponding to a 4% to 20% defective rate.

We evaluate the performance of proposed system in two scenarios. In the first scenario, an accurate (defective rate) is
assumed so that the first N nodes with the largest ranking differences are selected as faulty nodes. This scenario is known as

-based detection (or detection for short). is the defective rate i.e

= (no. of faulty nodes / total no. of nodes) x100.

IJSER © 2013 http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 4, Issue 4, April-2013 219

ISSN 2229-5518

In the second scenario, detection algorithm is used for

1000

900

800

9

16 12 20

7

700 22

15 18

600 3

500 21

8

400

13

10

11 23

19

300 24

4

200

25 2

14 5 1

100

6 17

0

0 100 200 300 400 500 600 700 800 900 1000

X

se
lecting faulty
nodes whose ranking differences are greater than B (bounded value). This scenario is known as B-based detection (or B- detection for short).
Two metrics are used for evaluating the proposed system: false negative rate and false positive rate. The former is defined as
the proportion of faulty nodes that are reported as normal, which is also known as miss detection rate. The latter is defined as the proportion of normal nodes that are reported as faulty, which is also known as the false alarm rate.
from the so-called cell arrays to the definition of classes in object oriented programming.
MATLAB syntax remains very simple and MATLAB programs can be written far more easily than programs in other high level languages or computer algebra programs. A command interface created for interactive management without much ado, plus a simple integration of particular functions, programs, and libraries supports the operation of this software tool. This also makes it possible to learn MATLAB rapidly.
MATLAB is not just a numerical tool for evaluation of formulas, but is also an independent programming language capable of
treating complex problems and is equipped with all the essential constructs of a higher programming language. Since the MATLAB command interface involves a so-called interpreter and MATLAB is an interpreter language, all commands can be carried out directly. This makes the testing of particular programs much easier.
MATLAB 7 is equipped with a very well conceived editor with debugging functionality, which makes the writing and error analysis of large MATLAB programs even easier.
The last major advantage is the interaction with the special toolbox Simulink, This is a tool for constructing simulation programs based on a graphical interface in a way similar to block diagrams. The simulation runs under MATLAB and an easy interconnection between MATLAB and Simulink is ensured.
Figure 7 and 8 depicts simulation scenario for 25 nodes with 1000x1000 area and cluster formation.
Figure 7: Randomly Placed Sensor Nodes

IJSER © 2013 http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 4, Issue 4, April-2013 220

ISSN 2229-5518

cluster

1

0.

0.

0.

0.

event clusterhead cluster

CM to CH CM to CH CM to CH CM to CH CM to CH

CH to

CH to

CH to

CH to

0.

0.

0.

0.

0.

0 0. 0. 0. 0. 0. 0. 0. 0. 0. 1

Figure 7: Cluster Formation
Figure 8 shows the simulation result of false negative rate for 19 events based on A detection .

A-Detection

20

with clustering

without clustering

15

False Negative

10

5

0

0 0.05 0. 0.15 0.

Defective Rate(A)

Figure 8: False Negative Rate (19 Events, A Detection)
Figure 9 shows the simulation result of false negative rate for 29 events based on A detection.

IJSER © 2013 http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 4, Issue 4, April-2013 221

ISSN 2229-5518

A-Detection

20

with clustering without clustering

15

False Negative

10

5

0

0 0.05 0. 0.15 0.

Defective Rate(A)

Figure 8 False Negative Rate (29 Events, A Detection)
Figure 9 shows the simulation result of false negative rate for 39 events based on A detection .
Events, B Detection)
tection(19,29,39,49 events respectively).

CONCLUSION & FUTURE WORK

Future Work

It can be possible extension of present work. The work in this dissertation is motivated by the problem of fault detection for
WSN. We have presented a promising framework for fault detection using FIND to model faults and monitor the data to detect the occurrence of these faults. It is worth emphasizing that fault detection is not the end goal in sensor networks. The primary

False Negative Rate

goal is to obtain clean data which can be analyzed to achieve the scientific objectives. Fault detection methods are used to make
this goal possible. In this context, the future work is that persistent faults will be diagnosed as quickly as possible so that large
amounts of data are not wasted.

REFERENCES

[1] R. Szewczyk, A. Mainwaring, J. Anderson, and D. Culler. An Analysis of a Large Scale

Habit Monitoring Application. In SenSys’04, 2004.

[2] G. Tolle, J. Polastre, R. Szewczyk, N. Turner, K. Tu, S. Burgess, D. Gay, P. Buonadonna,

W. Hong, T. Dawson, and D. Culler. A Macroscope in the Redwoods. In SenSys’05,

November 2005.

[3] W. Su Y. Sankarasubramaniam E. Cayirci Akyildiz, I.F. A survey on sensor networks.

IEEE Communications Magazine, pages 102{114, 2002}.

[4] Kumar.S.P. Chee-Yee Chong. Sensor networks: Evolution, opportunities, and challenges. Proc IEEE, August 2003.

[5] Ismail H. Kasimoglui Ian .F. Akyildiz. Wireless sensor and actor :research challenges.

(Elsevier) Journal, 2(38):351{367, 2004.

[6] Sarika Agarwal Leszek Lilien Maleq Khan, Bharat Bhargava and Pankaj. Self-configuring node clusters, data aggregation, and security in microsensor networks. Department of Management Information Systems Krannert Graduate School of Management Purdue University, West Lafayette, (IN 47907), 2007. pankaj@mgmt.purdue.edu.

[7] Sungha Pete Kim Bo-Cheng Charles Lai, David D. Hwang. Reducing radio energy consumption of key management protocols for wireless sensor networks. ACM 1-58113-

929-2/04/0008, pages 9{11, August 2004.

IJSER © 2013 http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 4, Issue 4, April-2013 222

ISSN 2229-5518

[8] Ismail H. Kasimoglui Ian .F. Akyildiz. Wireless sensor and actor :research challenges. (Elsevier) Journal, 2(38):351{367, 2004.

[9] K. Delin, S. Jackson, and R. Some, “Sensor Webs,” NASA Tech Briefs Journal, vol.

23, p. 80, Oct. 1999.

[10] J. Rabaey et al., “PicoRadios for wireless sensor networks — the next challenge in ultra- low power design,” in Proc. IEEE Int. Solid-State Circuits Conference (ISSCC), (San Francisco, CA), Feb. 2002.

[11] Shuo Guo, ZiguoZhong, Tian He “FIND: Faulty Node detection for Wireless Sensor

Networks”, in SenSys’09, November 4-6, 2009, Berkeley, CA, USA.

[12] Peng Jiang “A New Method for Node Fault Detection in Wireless Sensor Networks”

,Sensors 2009.

[13] Myeong-Hyeon Lee, Yoon-Hwa Choi “Fault detection of wireless sensor networks”

,(Elsevier) Computer Communications 31 (2008) 3469–3475.

[14] Ann T. Tai Kam S. Tso, William H. Sanders “Cluster-Based Failure Detection

Service for Large-Scale Ad HocWireless Network Applications” ,Proceedings of the

2004,IEEE.

[15] M.Yu, H.Mokhtar, and M.Merabti “A Survey on Fault Management in Wireless

Sensor Networks”.

[16] Sumit Gupta, Rong Zheng, Albert M. K. Cheng “ANDES: an Anomaly Detection

System for Wireless Sensor Networks”, 2007 IEEE.

[17] Abolfazl Akbari and Neda Beikmahdavi and Bahman Shahamat Naderi “Cluster- based and cellular approach to fault detection and recovery in wireless sensor network”, IJWMN February 2010.

[18] Sheela D. Nair “Finding Fault: Anomaly Detectio n for Embedded Networked

Sensing ”.

[19] Pabitra Mohan Khilar, Jitendra Kumar Singh, Sudipta Mahapatra “Design and Evaluation of a Failure Detection Algorithm for Large Scale Ad Hoc Networks Using Cluster Based Approach “,2008 IEEE.

[20] F.L.Lewis “ Wireless Sensor Networks “, To appear in Smart Environments: Technologies, Protocols, and Applications ed. D.J. Cook and S.K. Das, John Wiley, New York, 2004.

[21] R. Isermann. Fault-Diagnosis Sytems: An Introduction from Fault Detection to Fault

Tolerance. Springer, 2005.

[22] A. Sharma, L. Golubchik, and R. Govindan. \On the Prevalence of Sensor Faults in Real World Deployments." In IEEE Conference on Sensor, Mesh and Ad Hoc Communications and Networks (SECON).

AUTHORS

Srikanta Kumar Sahoo, is currently pursuing masters degree program in computer science and engineering department in Biju Patnaik University of Technolog,Odisha,India,Mobile No-+919776293140,Email- srikant.sahoo@live.com.

IJSER © 2013 http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 4, Issue 4, April-2013

ISSN 2229-5518

223

IJSER IS)2013

http://www ijserorg