Research Article Energy Efficient and Safe Weighted...

Research ArticleEnergy Efficient and Safe Weighted Clustering Algorithm forMobile Wireless Sensor Networks

Amine Dahane1 Abdelhamid Loukil1 Bouabdellah Kechar2 and Nasr-Eddine Berrached1

1 Intelligent Systems Research Laboratory University of Sciences and Technology of Oran Algeria2Laboratory of Industrial Computing and Networking Ahmed Ben Bella Oran University Algeria

Correspondence should be addressed to Abdelhamid Loukil abdelhamidloukiluniv-ustodz

Received 10 November 2014 Accepted 15 January 2015

Academic Editor Zahoor Khan

Copyright copy 2015 Amine Dahane et alThis is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

The main concern of clustering approaches for mobile wireless sensor networks (WSNs) is to prolong the battery life of theindividual sensors and the network lifetime For a successful clustering approach the need of a powerful mechanism to safelyelect a cluster head remains a challenging task in many research works that take into account the mobility of the network Theapproach based on the computing of the weight of each node in the network is one of the proposed techniques to deal with thisproblem In this paper we propose an energy efficient and safe weighted clustering algorithm (ES-WCA) for mobile WSNs using acombination of five metrics Among these metrics lies the behavioral level metric which promotes a safe choice of a cluster head inthe sense where this last one will never be a malicious node Moreover the highlight of our work is summarized in a comprehensivestrategy for monitoring the network in order to detect and remove the malicious nodes We use simulation study to demonstratethe performance of the proposed algorithm

1 Introduction

After the success of theoretical research contributions inprevious decade wireless sensor networks (WSNs) havebecome now a reality [1ndash3] Their deployment in manysocietal environmental and industrial applications makesthem very useful in practice These networks consisted oflarge number of small size nodes which sense ubiquitouslysome physical phenomenon (temperature humidity accel-eration noise light intensity wind speed etc) and reportthe collected data to the sink station by using multihopwireless communicationsAlthough the nodes are able to self-organize and collaborate together in order to establish andmaintain the network they are battery powered limited interms of processing storage and communication capabilities[4] WSNs are considered in many cases as stationary buttopology changes can happen due to a weak mobility (newnodes join the network and existing nodes experience hard-ware failure or exhaust their batteries) [5] In other scenariosthe mobility can occur when nodes are carried by externalforces such as wind water or air [6] so that the networktopology can be affected accordingly and can be changedslowly This second kind of mobility known as one form of

strong mobility in the literature in the sense where nodes areforced to move physically in the deployment area has beenconsidered in this paper Clustering means grouping nodeswhich are closed to each other and it has been widely studiedin ad hoc networks [3 7ndash14] More recently it has been usedin WSNs [14ndash21] where the purpose in general is to reduceuseful energy consumption and routing overhead Figure 1illustrates how inside the cluster two kinds of nodes can befound one node called cluster head (CH) or coordinator(in Figure 1 CH1 CH2 and CH3) which is responsible forcoordinating the cluster activities and several ordinary nodescalled cluster members (CMs) (in Figure 1 CM1 and CM2)that have direct access only to one CH An ordinary nodewhich is able to hear two or more CHs is called a gateway(G) (in Figure 1 the gateway G2 can hear CH1 CH2 andCH3 while the gateway G1 can hear CH1 and CH2) insteadSo each communication initiated by a cluster member toa destination inside the cluster must pass by CH If thedestination is outside the cluster the communication mustbe forwarded by a gateway Recent research studies recognizethat organizing mobile WSNs in the sense defined aboveinto clusters by using a clusteringmechanism is a challengingtask [9 19] This is due to the fact that CHs carry out extra

Hindawi Publishing CorporationMobile Information SystemsVolume 2015 Article ID 475030 18 pageshttpdxdoiorg1011552015475030

2 Mobile Information Systems

BSCM1

CM2

CH2

G1

G2

CH1

CH3

Figure 1 Clustering formation of WSNs composed of 150 sensornodes deployed in a 570m times 555m space area with a radio range =100m

work and consequently consume more energy comparedto CMs during the network operations and this will leadto untimely death causing network partition and thereforefailure in communication link For this reason one of themost frequently encountered problems in this mechanism isto search for the best way to elect CH for each cluster Indeeda CH can be selected by computing the quality of nodes Thismay depend on severalmetrics connectivity degreemobilityresidual energy and the distance of a node from its neighborsSignificant improvement in performance of this quality canbe achieved by combining these metrics [3 9 10 12 19 21]

In this paper we propose an energy efficient and safeweighted clustering algorithm for mobile WSNs using acombination of the above metrics to which we added abehavioral level metric The latter metric is decisive andallows the proposed clustering algorithm to avoid any mali-cious node in the neighborhood to become a CH even ifthe remaining metrics are in its favor The election of CHsis carried out using weights of neighboring nodes whichare computed based on selected metrics So this strategyensures the election of legitimate CHs with high weightsThe preliminary results obtained through simulation studydemonstrate the effectiveness of our algorithm in terms ofthe number of equilibrate clusters and the number of reaffili-ations compared to WCA (Weighted Clustering Algorithm)[3] DWCA (Distributed Weighted Clustering Algorithm)[9] and SDCA (Secure Distributed Clustering Algorithm)[21] These results also reveal that our approach is suitable ifwe plan to use it in network layer reactive routing protocolsinstead of proactive ones once the clustering mechanism islaunched

We can enumerate the contributions of our paper asfollows

(i) maintaining stable clustering structure and offeringa better performance in terms of the number ofreaffiliations using the proposed algorithm ES-WCA(Energy Efficient and SafeWeighted Clustering Algo-rithm)

(ii) detecting common routing problems and attacks inclustered WSNs based on behavior level

(iii) showing clearly the interest of the routing protocols inenergy saving and therefore maximizing the lifetimeof the global network

The remaining part of this paper is organized as followsSection 2 briefly surveys the related works on clusteringalgorithms proposed for ad hoc networks and in particularthose developed for WSNs In Section 3 we emphasize onthe security problems in WSNs Section 4 introduces andexplains the selected metrics for the proposed approach ofclustering More details on the proposed algorithm are givenin Section 5 Section 6 presents the simulation tool developedfor evaluation Simulation results are provided to show theeffectiveness of the proposed algorithm Section 7 concludesthe paper and outline directions of future works

2 Related Works

In this section we outline some approaches of clustering usedin ad hoc networks andWSNs Research studies on clusteringin ad hoc networks involve surveyed works on clusteringalgorithms [11 22] and cluster head election algorithms[10 16] Abbasi and Younis [17] presented taxonomy andclassification of typical clustering schemes then summa-rized different clustering algorithms for WSNs based onclassification of variable convergence time protocols andconstant convergence time algorithms and highlighted theirobjectives features complexity and so forth A single metricbased on clustering as in paper [23] shows that the nodewith the least stability value is elected as CH among itsneighbors However the choice of CH which has a lowerenergy level could quickly become a bottleneck of its clusterEr and Seah [8] designed and implemented a dynamic energyefficient clustering algorithm (DEECA) for mobile ad hocnetworks (MANETs) that increases the network lifetimeTheproposed model elects first the nodes that have a higherenergy and less mobility as cluster heads then periodicallymonitors the cluster headrsquos energy and locally alters theclusters to reduce the energy consumption of the sufferingcluster heads The algorithm defines a yellow threshold toachieve some sort of local load balancing and a red thresholdto trigger local reclustering in the network However thecluster formation in this scheme is not based on connectivityso the formed clusters are not well connected consequentlythis increases the reaffiliation rate andmaximizes reclusteringsituations Jain andReddy [24] have proposed a novelmethodof modeling wireless sensor network using fuzzy graphand energy efficient fuzzy based k-Hop clustering algorithmwhich takes into account the dynamic nature of networkvolatile aspects of radio links and physical layer uncertaintyThey have defined a new centrality metric namely fuzzy

Mobile Information Systems 3

k-hop centrality The proposed centrality metric considersresidual energy of individual nodes link quality hop distancebetween the prospective cluster head and respectivemembernodes to ensure better cluster head selection and clusterquality which results in better scalability balancing of energyconsumption of nodes and longer network lifetime Otherproposals use a strategy based on computed weight in orderto elect CHs [3 9 10 12] The main strategy of thesealgorithms is based mainly on adding more metrics such asthe connectivity degree mobility residual energy and thedistance of a node from its neighbors corresponding to someperformance in the process of electing CHs Although thealgorithms which use this strategy allow us to ensure theelection of better CHs based only on their high computedweight from the considered metrics they unfortunately donot ensure that the elected CHs are legitimated nodes thatis whether the election process of CHs is safe or not Safaet al [13] propose a novel cluster based trust-aware routingprotocol (CBTRP) forMANETs to protect forwarded packetsfrom intermediary malicious nodes The proposed protocolensures the passage of packets through trusted routes onlyby making nodes monitor the behavior of each other andupdate their trust tables accordingly However in CBTRPall nodes monitor the network which lead to rapid drainageof node energy and therefore minimize the lifetime of thenetwork In Section 3 we show that WSNs are vulnerable tovarious types of attacks [24 25] In the last decade severalstudies proposed solutions to solve attacks in WSNs by usingcryptography such as SPINS [26] However cryptographyalone is not enough to prevent node compromise attacks andnovel misbehavior in WSNs [27] Little effort has been madeto include the security aspect in the clustering mechanismYu et al [4 28] try to secure the clustering mechanismagainst wormhole attack in ad hoc networks (communicationbetween CHs) However this is done after forming clustersnot during the election procedure of CHs Liu [4 29] sur-veyed the clustering algorithms available for WSNs but thatwas done from the perspective of data routing Hai et al [30]propose a lightweight intrusion detection framework inte-grated for clustered sensor networks by using an overhearingmechanism to reduce the sending alert packets Elhdhiliet al [31] propose a reputation based clustering algorithm(RECA) that aims to elect trustworthy stable and highenergy cluster heads but during the election procedure notafter forming clusters Benahmed et al [21] used clusteringmechanism based on weighted computing as an efficientsolution to detect misbehavior nodes during distributedmonitoring process inWSNs However they focused only onthe misbehavior of malicious nodes and not on the natureof attacks the formed clusters are not homogeneous theproposed algorithm SDCA is not coupled with a routingprotocols and it does not give much importance to energyconsumption

In this paper the proposed approach focuses aroundstrategy of distributed resolutionwhich enables us to generatea reduced number of balanced and homogeneous clustersin order to minimize the energy consumption of the entirenetwork and prolong sensors lifetime The introduction ofa new metric (the behavioral level metric) promotes a safe

choice of a cluster head in the sense where this last one willnever be a malicious node Thus the highlight of our workis summarized in a comprehensive strategy for monitoringthe network in order to detect and remove the maliciousnodes

The fact that WSNs include limited energy resources(batteries) duemainly to their small size our algorithm showsclearly the interest of the routing protocols in energy savingwhich therefore maximize the lifetime of the network bycoupling it with AODV and then DSDV protocols [5 32 33]

3 Security in WSNs

The typical attacks in WSNs include Sinkhole attack BlackHole attack Hello Flood attack and Node Outage which arethe most common network layer attacks on WSNs [30 34ndash38] These selected attacks have been summarized in thefollowing sections

31 Sinkhole Sinkhole attack is one of the most devastatingones it is very hard to protect against [36 39] In a Sinkholeattack the adversaryrsquos goal is to redirect nearly all the trafficfrom a particular area through a compromised node creatinga metaphorical sinkhole with the adversary at the centerso that all traffic in the surrounding will be absorbed bythe malicious node Because nodes on or near the pathfollowed by transmitted packets have many opportunitiesto tamper with application data Sinkhole attacks can enablemany other attacks such as selective forwarding for example[40]

32 Black Hole In this attack malicious nodes advertise veryshort paths (sometimes zero-cost paths) to every other nodeforming routing black holes within the network [41] As theiradvertisement propagates the network routes more trafficin their direction In addition to disrupting traffic deliverythis causes intense resource contention around the maliciousnode as neighbors compete for limited bandwidth Theseneighbors may themselves be exhausted prematurely causinga hole or partition in the network

33 Hello Flood Attack Many routing protocols use ldquoHellordquobroadcastmessages to announce themselves to their neighbornodes The nodes that receive this message assume thatsource nodes are within range and add source nodes to theirneighbor listTheHello Flood attacks can be caused by a nodewhich broadcasts aHello packet with very high power so thata large number of nodes even far away in the network chooseit as the parent node [14]These nodes are then convinced thatthe attacker node is their neighbor so that all the nodes willrespond to the Hello message and waste their energy

34 Node Outage If a node acts as an intermediary anaggregation point or a cluster head what happens if thenode stops working Protocols used by the WSNs must berobust enough to mitigate the effects of failures by providingalternate routes [34]


Malicious Suspect Abnormal Normal

0 03 05 08 1

Behavior level

Figure 2 Behavior level BL119894isin [0 1]

4 Metrics for CHs Election

This section introduces the different metrics used for clusterhead election by focusing on behavior level metric

41 The Behavior Level of Node 119899119894(BL119894) The behavioral level

of a node 119899119894is a key metric in our contribution Initially

each node is assigned an equal static behavior level ldquoBL119894= 1rdquo

However this level can be decreased by the anomaly detectionalgorithm if a node misbehaves For computing the behaviorlevel of each node nodes with a behavior level less thanthreshold behavior will not be accepted as CH candidateseven if they have the other interesting characteristics suchas high energy high degree of connectivity or low mobilityNevertheless abnormal nodes and suspect nodes may belongto a cluster as CMbut never as CH So we define the behaviorlevel of each sensor node 119899

119894 noted BL

119894 in any neighborhood

of the network as illustrated in Figure 2BL119894is classified by the following mapping function

Mp (BL119894) =

Normal node 08 le BL119894le 1

Abnormal node 05 le BL119894lt 08

Suspect node 03 le BL119894lt 05

Malicious node 0 le BL119894lt 03

(1)

The values in formula (1) are chosen on the basis of severalreputed models of WSNs adopted by numerous researcherslike Shaikh et al [42] and Lehsaini et al [43] The monitornode watches its neighbors to know what each one of themdoes with the messages it receives from another neighborIf the neighbor of the monitor changes delays replicatesor simply keeps a message that should be retransmitted themonitor counts a failure Number of failures have influenceon the behavior of neighbors for instance if the monitorcounts one failure from a neighbor its behavior will decreaseby 01 units This allows the monitor (cluster head) todifferentiate malicious nodes (that make much failure) of alegitimate node (that make fewer failure) in case there arecollisions

42 The Mobility of Node 119899119894(119872119894) Our objective is to have

stable clusters So we have to elect nodes with low relativemobility as CHs To characterize the instantaneous nodalmobility we use a simple heuristic mechanism as presentedin the formula below (2) [4 44]

119872119894=

1

119879

119879

sum

119905=1

radic(119909119905minus 119909119905minus1)2+ (119910119905minus 119910119905minus1)2 (2)

where (119909119905 119910119905) and (119909

119905minus1 119910119905minus1) are the coordinates of node 119899

119894

at time 119905 and 119905 minus 1 respectively 119879 is the period for which thisparameter is estimated

In our previous paper [4] the considered mobility has aparticular sense by the fact that a mobile node does not movefrom one location to another in the space area of its ownwill but in our case it moves through the forces acting fromthe outside These external forces can act from time to timesporadically In contrary the malicious node can use its ownability to move freely in the space area The behavior of themalicious node by moving frequently inside the same cluster(case illustrated by Figure 3) or from a cluster to another is anormal behavior to not attract attention of the neighborhoodand therefore be detected The idea of our algorithm toensure the choice of a legitimate CH is to never elect a nodethat moves frequently and even it has the best performancemetrics but this malicious node does nothing just mobilityso in this paper our algorithm (ES-WCA) detects the internalmisbehavior of nodes during distributed monitoring processinWSNs by the follow-up of themessages exchanged betweenthe nodes ES-WCA is based on the ideas proposed by da Silvaet al [45] used in his efficient and accurate IDS in detectingdifferent kinds of simulated attacks

43 The Distance between Node 119899119894and Its Neighbors (119863

119894)

This is likely to reduce node detachments and enhance clusterstability For each node 119894 we compute the sum of the distance119863119894with all its neighbors 119895This distance is given as in [3 4 9]

by

119863119894= sum

119895 isin 119873(119894)

dist (119894 119895) (3)

44The Residual Energy of Node 119899119894(Er119894) The residual energy

of a node 119899119894 after transmitting a message of 119896 bits at distance

119889 from the receiver is calculated according to [4 16]

Er119894= 119864 minus (119864

119879119909 (119896 119889) + 119864119877119909 elec (119896)) (4)

where

(i) 119864 the nodersquos current energy

(ii) 119864119879119909(119896 119889) = 119896 sdot 119864elec + 119896 sdot 119864amp sdot 119889

2 it refers to therequired energy to send a message where 119864amp is therequired amplifier energy

(iii) 119864119877119909 elec(119896) = 119896119864elec it refers to the energy consumed

while receiving a message

45 The Degree of Connectivity of Node 119899119894at Time 119905 (119862

119894)

It represents the number of 119899119894rsquos neighbors given by (5)

according to [4]

119862119894= |119873 (119894)| (5)


5

13

6

24

Cluster

Cluster head (CH)

Cluster member (CM)

Radio rangeof node 3(old CH)

Malicious node

Moving directionCommunication link

(a)

5

5

4

4

31

1

4

4

4

6

2

Cluster

Initiallocation

Finallocation

Cluster head (CH)

Cluster member (CM)Malicious node


Radio rangeof node 1(new CH)

(b)

Figure 3 (a) Clustering mechanism in mobile WSNs before moving nodes and (b) after moving nodes 1 5 and 4

where

(i) 119873(119894) = 119899119894dist(119894 119895) lt 119905119909range with 119894 = 119895

(ii) dist(119894 119895) outdistance separating two nodes 119899119894and 119899119895

(iii) 119905119909range the transmission radius

For each node we must calculate its weight 119875119894 according to

the equation

119875119894= 1199081lowast BL119894+ 1199082lowast Er119894+ 1199083lowast119872119894+ 1199084lowast 119862119894+ 1199085

lowast 119863119894

(6)

where1199081119908211990831199084 and119908

5are the coefficients correspond-

ing to the system criteria so that

1199081+ 1199082+ 1199083+ 1199084+ 1199085= 1 (7)

We propose to generate homogeneous clusters whose size liesbetween two thresholds 119879ℎ119903119890119904ℎ

119880119901119901119890119903and 119879ℎ119903119890119904ℎ

119871119900119908119890119903

These thresholds are arbitrarily selected or they dependon the topology of the network Thus if their values dependon the topology of the network they are calculated as followsaccording to [43]

(i) 119906 the node that has the maximum number of neigh-bors with one jump

12057512 (119906) = min (120575

12(119906119894) 119906119894isin 119880) (8)

(ii) V the node that has theminimal number of neighborswith one jump

12057512 (

V) = min (12057512(V119894) V119894isin 119880) (9)

We denote AVG by the average cardinal of the groups withone jump of all the nodes of the network

AVG =sum119899

119894=112057512(119906119894)

119873

(10)

where 119873 represents the number of nodes in the networkThus the two thresholds are calculated as follows

119879ℎ119903119890119904ℎ119880119901119901119890119903

=

1

2

(12057512 (119906) + AVG)

119879ℎ119903119890119904ℎ119871119900119908119890119903

=

1

2

(12057512 (

V) + AVG) (11)

The calculated weight for each sensor is based on theabove parameters (BL

119894119872119894 119863119894Er119894 and 119862

119894) The values of

coefficients119908119894should be chosen depending on the basis of the

importance of each metric in considered WSNs applicationsFor instance it is possible to assign a greater value to themetric BL

119894compared to other metrics if we promote the

safety aspect in the clusteringmechanism It is also possible toassign the same value for each coefficient119908

119894in the case where

all metrics are considered as having the same importance Anapproach based on these weight types will enable us to builda self-organizing algorithm which forms a small number ofhomogenous clusters in size and radius by geographicallygrouping close nodes The resulting weighted clusteringalgorithm reduces energy consumption and guaranties thechoice of legitimate CHs

5 Weighted Clustering Algorithm (ES-WCA)

In this section we first present some assumptions of theproposed algorithm Energy Efficient and Safe Weighted


Clustering algorithm (ES-WCA)Thenwe present in detail anextended version of ES-WCA [4] followed by an illustrativeexample

51 Assumptions This paper is based on the followingassumptions

(i) The network formed by the nodes and the links can berepresented by an undirected graph119866 = (119880 119864) where119880 represents the set of nodes 119899119894 and 119864 represents theset of links 119890119894 [3 4]

(ii) All sensor nodes are deployed randomly in a 2-dimension (2D) plane

(iii) A node interacts with its one-hop neighbors directlyand with other nodes via intermediate nodes usingmultihop packet forwarding based on a routing pro-tocol such as ad hoc on demand distance vector [5 32]or DSDV [33]

(iv) The radio coverage of sensor nodes is a circular regioncentered on this node with radius 119877

(v) Two sensor nodes cannot be deployed in exactly thesame position 119909 119910 in a 2D space

(vi) All sensor nodes are identical or homogeneous Forexample they have the same radio coverage radius 119877

(vii) Each node can determine its position at any momentin a 2D space

(viii) Each cluster is monitored by only one CH(ix) Each CM communicates directly with its CH for the

transmission of security metrics(x) A CH communicates directly with the base station for

the transmission of security information and possiblealerts

52 Proposed Algorithm The ES-WCA algorithm that wepresent below is based on the ideas proposed by Chatterjeeet al [3] Lehsaini et al [43] and Zabian et al [10] withmodifications made for our application This algorithm runsin three phases the setup phase the reaffiliation phase andthe monitoring phase ES-WCA combines each of the abovesystem parameters with certain weighting factors chosenaccording to the system needs

521 The Setup Phase ES-WCA uses three types of messagesin the setup phase (Algorithm 1)Themessage CHmsg is sentin the network by the sensor node which has the greatestweighThe second one is the JOINmsg message which is sentby the neighbor of CH if it wants to join this cluster Finallya CH must send a response ACCEPTmsg message as shownin Figure 4

The node which has the greatest weight begins the pro-cedure by broadcasting CHmessage to their 1-hop neighborsto confirm its role as a leader of the cluster The neighborsconfirm their role as being member nodes by broadcastinga JOINmsg message In the case when nodes have thesame maximum weight the CH is chosen by using the bestparameters ordered by their importance If all parameters ofnodes are equal the choice is random

U CH

ACCEPT_CH message

REQ_JOIN message

ADV_CH message

Figure 4 Procedure of affiliation of node ldquoUrdquo to a cluster

U

CH

RE_AFF_CHREQ_RE_AFFACCEPT_RE_AFF

Figure 5 Procedure of reaffiliation of node ldquoUrdquo to a cluster

Table 1 Values of the various criteria of normal nodes

Ids BL119894

Er119894

119862119894

119863119894

119872119894

119875119894

1 086 384212 3 115 120 7696324 081 483254 5 230 030 9681335 088 405325 3 130 055 8118296 085 462043 0 000 020 9243618 081 481680 4 105 140 96475310 095 365025 2 055 010 73080511 091 481960 1 070 220 964753

522 The Reaffiliation Phase ES-WCA uses four types ofmessages in the reaffiliation phase (Algorithm 2) The mes-sage RE AFF CH is sent in the network by the CH whosecluster size is less than 119879ℎ119903119890119904ℎ

119880119901119901119890119903 The second one is the

REQ RE AFF message which is sent by the neighbors of CHif it wants to join this cluster Finally a CH must send aresponse ACCEPT RE AFFmessage or DROP AFFmessageas illustrated by Figure 5 Accordingly in this phase wepropose to reaffiliate the sensor nodes belonging to clustersthat have not attained the cluster size 119879ℎ119903119890119904ℎ

119871119900119908119890119903to those

that did not achieve 119879ℎ119903119890119904ℎ119880119901119901119890119903

in order to reduce thenumber of clusters formed and organize them so as to obtainhomogeneous and balanced clusters

With the help of 3 figures (Figures 6 7 and 8) ouralgorithm setup phase is demonstrated Table 1 shows thequantitative results of the different criteria applied on thenormal nodes (BL

119894ge 08) Table 2 shows the weights 119875

119894

of neighbors for each node which has behavior BL119894higher


Begin(1) Assign values to the coefficients 119908

1 1199082 1199083 1199084 1199085

(2) For any node 119899119894isin 119866 make

(3) 119899119894forms a list of its neighbors119873(119894) through the Message who are neighbors

(4) 119873(119894) = 0(5) Calculate its weight 119875

119894

(6) 119875119894= 1199081lowastBL119894+ 1199082lowastEr119894+ 1199083lowast119872119894+ 1199084lowast119862119894+ 1199085lowast119863119894

(7) Initialize Time Cluster and the state vector of allnodes 119899

119894isin 119866 Vector State (Id CH Weight List Neighbors Size Nature)

(8) CH = 0 Size = 0(9) Nature = ldquoNonerdquo(10) Repeat(11) Any node 119899

119894isin 119866 Broadcasts a message ldquoHellordquo

(12) If 119873(119894) ltgt 0 Then(13) Choose V isin 119873(119894)(14) 119882119890119894119892ℎ119905(V) = max119908119890119894119892ℎ119905(119908) 119908 isin 119873(119894)(15) the node that have the same maximum weight the CH is

the node that has the best criteria ordered by their

importance (BL119894Er119894119862119894 119863119894and 119872

119894) if all criteria of

nodes are equal the choice is random

(15) Else 119899119894is a CH of itself

EndIf(16) Update the state vector of the elected CH(17) CH = ID(18) Size = 1(19) Nature = CH(20) Send the message ldquoCHmsgrdquo by CH to its neighbors119873(CH)(21) 119869 = Count (119873(CH))(22) For 119868 = 1 to 119869 Do(23) If (119899

119894isin 119873(CH) receives the message ampamp119899

119894rarr CH = 0)

(24) Then 119899119894sends a message ldquoJOINmsgrdquo to CH

(25) If (CH rarr Size lt 119879ℎ119903119890119904ℎ119880119901119901119890119903

)(26) Then CH sends a message ldquoACCEPTmsgrdquo to Node 119899

119894

(27) CH executes the accession process(28) CH rarr Size = CH rarr Size + 1(29) 119899

119894executes the accession process

(30) 119899119894rarr CH = CH rarr Id

(31) Else go to (10)EndIf

EndIfEnd For

(32) Until expired (TimeCluster)End

Algorithm 1 Algorithm setup phase

Table 2 Weights of neighbors

Ids 1 4 5 6 8 10 111 769632 964753 9647534 968133 811829 9647535 968133 811829 7308056 9243618 769632 96475310 968133 811829 73080511 769632 964753


Inputs 119879ℎ119903119890119904ℎ119880119901119901119890119903

119879ℎ119903119890119904ℎ119871119900119908119890119903

Outputs set of clustersBegin(1) For num cl = 1 to Count (Cluster)Do(2) If (Size (Cluster [num cl]) lt 119879ℎ119903119890119904ℎ

119880119901119901119890119903)

Then(3) CH sends a message ldquoRE AFF CHrdquo to its neighbors

(119873(CH))(4) 119869 = Count (119873(CH))

EndIf(5) For 119868 = 1 to 119869 Do(6) If (119899

119894isin 119873(CH) receives the message)

ampamp (119899119894isin (Size (Cluster [num cl]) lt 119879ℎ119903119890119904ℎ

119871119900119908119890119903)

Then(7) 119899

119894sends a Select message ldquoREQ RE AFFrdquo to the CH

(8) If (Size (Cluster [num cl]) lt 119879ℎ119903119890119904ℎ119880119901119901119890119903

)Then

(9) CH sends a message ldquoACCEPT RE AFFrdquo to 119899119894

(10) CH updates its state vector(11) CH rarr CH rarr Size = Size + 1(12) 119899

119894updates its state vector

(13) 119899119894rarr CH rarr ID = ID

(14) Else CH sends a ldquoFIN AFFrdquo message to 119899119894

(15) Go to (2)EndIF

(16) Else 119899119894sends a ldquoDROP AFFrdquo message to CH

EndIfEnd For

End ForEnd

Algorithm 2 Algorithm reaffiliation phase

12055

7048

10095

2036

3045

5088

4081

8081

9050

1

086

11091

6

085

Figure 6 Topology of the network

than 08 The circles in Figure 6 represent the nodes theiridentity Ids are at the top and their levels of behavior are atthe bottom According to Table 2 node 1 could be attachedto either CH11 or CH8 (since they have the same weight)However the behavior level of node 11 is greater than that ofnode 8 (BL

11gt BL8) So node 1 will be attached to CH11

For the other nodes we have various conditions Node 4declares itself as a CH Node 5 will be attached to CH4 Node6 declares itself as a CH because it is an isolated node Node8 will be attached to CH4 Node 10 is connected to CH5 but

node 5 is attached to CH4 Thus node 10 declares itself asa CH Node 11 declares itself as a CH These results give usthe representation shown in Figure 7 Node 2 is connectedto CH4 and CH10 Node 2 will be attached to CH4 becauseCH4 has themaximumweight (968133) Node 3 is connectedto CH4 which implies that node 3 will be attached to CH4Node 7 is not connected to any CH so node 7 declares itselfas CH Node 9 is connected to CH4 and then node 9 will beattached to CH4 Node 12 is not connected to any CH whichimplies that node 12 declares itself as a CH These resultsgive us the representation shown in Figure 8 We propose togenerate homogeneous clusters whose size lies between twothresholds 119879ℎ119903119890119904ℎ

119880119901119901119890119903= 9 and 119879ℎ119903119890119904ℎ

119871119900119908119890119903= 6 For that

we suggest to reaffiliate the sensor nodes belonging to theclusters that have not attained the cluster size 119879ℎ119903119890119904ℎ

119871119900119908119890119903to

those that did not reach 119879ℎ119903119890119904ℎ119880119901119901119890119903

Node 4 has the highestweight and his size is less than 119879ℎ119903119890119904ℎ

119880119901119901119890119903 Nodes 1 7 and

10 are neighbors of node 4 with 2 hops and belong to theclusters that have not attained the cluster size 119879ℎ119903119890119904ℎ

119871119900119908119890119903 so

these nodes get merged to cluster 2 Clusters 1 3 and 4 willbe homogeneous with cluster 1 when the network becomesdensely

At the end of this example we obtain a network of fourclusters (as shown in Figure 9)


Cluster 2

10

095

7

048

2

036

5

088

3

045

4

081

8

081

9

050

1

086

11091

6

085

12055

Cluster 4 Cluster 3

Cluster 1

Figure 7 Identification of clusters node

Cluster 2

10

095

7

048

2

036

5

088

3

045

4

081

8

081

9

0501

086

11091

6

085

12055

Cluster 4

Cluster 5

Cluster 6

Cluster 3

Cluster 1

Figure 8 The final identification of clusters

Cluster 210

095

7

048

2

036

5

088

3

045

4

081

8

081

9

0501

086

11091

6

085

12055

Cluster 4

Cluster 3

Cluster 1

Figure 9 Final cluster structure (reaffiliation phase)

There are five situations that require the maintenance ofclusters

(i) battery depletion of a node(ii) behavior level of a node less than or equal 03(iii) adding moving or deleting a node

In all of these cases if a node 119899119894is CH then the setup phase

will be repeated

523 The Monitoring Phase Monitoring in WSNs can beboth local and global The local monitoring can be withrespect to a node and the global monitoring can be withrespect to the network but in sensor networks for detecting

some types of errors and security anomalies the local moni-toring would be insufficient [46] For this reason we adopt inthis paper a hybrid approach that is global monitoring basedon distributed local monitoring The general architectureof our approach is illustrated in Figure 10 Our simulatorbaptized ldquoMercuryrdquo detects the internal misbehavior nodesduring distributed monitoring process in WSNs by thefollow-up of the messages exchanged between the nodesWe assume that the network has already a mechanism ofprevention to avoid the external attacks By using a setof rules all the received messages are analyzed A similarapproach is used by da Silva et al [45] and Benahmed et al[21]


Cluster 2

Cluster 1

BS

Local monitoring

Global monitoring

Figure 10 Monitoring phase architecture

CHi broadcasts a ldquostartmonitoringrdquo message to CMs

Each node ni calculatesits security metrics

Each node ni sends allmetrics to the CHi

Called the punishingalgorithm

Node ni sends a message to its CHi

for monitoring purposesYes

State (ni ti)-state (ni timinus1) gt 120598

Yes

NoNo

ni is a normal node

Misbehavior detectionNo information is sent to the CH

Compute the deviation d(S) byusing equation (15)

d(S) gt Th

Figure 11 Monitoring phase

Algorithm 4 (monitoring phase algorithm) The monitor-ing process involves a series of steps as illustrated by theflowchart in (Figure 11)

Step 1 (this step runs in each 119862119867119894) Each CH

119894becomes the

monitor node of its cluster members and broadcasts a ldquoStartMonitoringrdquo message with its Idi to its entire cluster CMs

Step 2 (calculation of security metrics performed by eachmember 119899

119894of the cluster 119894) Each node 119899

119894(119894 ltgt 119895) receives the

message ldquoStartMonitoringrdquo and calculates its securitymetricsas follows

(i) Number of packets sent by 119899119894at time interval is Δ119905 =

[1199050 119905] 119873119887119901 119878119890119899119889(119899119894 Δ119905)

(ii) Number of packets received by node 119899119894at time

interval is Δ119905 = [1199050 1199050] 119873119887119901 119877119890119888119890119894V119890119889(119899

119894 Δ119905)

(iii) Delay between the arrivals of two consecutive packetsis

119863119890119897119886119910 119861119875 (119899119894 119905) = 119860119903119903119894V119886119897 119875119879

119894minus 119860119903119903119894V119886119897 119875119879

119894minus1 (12)

(iv) Energy consumption the energy consumed by thenode 119895 in receiving and sending packets is measuredusing the following equation

119864119888 (119899119894 Δ119905) = Er (119899

119894 1199050) minus Er (119899

119894 1199051) (13)

where Δ119905 is the time interval [1199050 1199051]Er(119899

119894 1199050) is the

residual energy of node 119899119894at time 119905

0 Er(119899

119894 1199051) is the


1and 119864119888(119899

119894 Δ119905) is

the energy consumption of node 119899119894at time intervalΔ119905

Step 3 (sending all metrics to the CH) After each consumptionof the security metrics the state of a node 119899

119894at time 119905 is

denoted by state (119899119894 119905119894) For storage volume economy each

node keeps only the latest calculation state

(i) In the initial deployment eachCM in cluster ldquo119894rdquo sendssome states (state(119899

119894 119905119894)) to the CHi for making a

normal behavior model of node 119899119894by using a learning

mechanism


(ii) Each state contains the following information

(119868119889119873119887119901119878119890119899119889(119899119894Δ119905)

119873119887119901119877119890119888119890119894V119890119889(119899119894 Δ119905) 119863119890119897119886119910119861119875(119899119894 119905)

119864119888 (119899119894 Δ119905))

(14)

(iii) If (state (119899119894 119905119894) minus state (119899

119894 119905119894minus1

) gt 120598)

then node 119899119894sends a message (120598 a given thresh-

old)Msg = (119868119889119873119887119901

119878119890119899119889(119899119894Δ119905) 119873119887119901

119877119890119888119890119894V119890119889(119899119894 Δ119905)119863119890119897119886119910

119861119875(119899119894 119905) 119864119888(119899

119894 Δ119905)) to its CHi for

monitoring purposesOtherwise no information is sent to the CH

(iv) The message received by CHi will be stored in a tableTmet for future analysis

(v) If a sensor node 119899119894does not respond during this mon-

itoring period it will be considered as misbehaving(vi) The behavior level of sensor node 119899

119894is computed

using the following equation

BL119894= BL119894minus rate (15)

The ldquoraterdquo is fixed on the basis of the nature of theapplication For example if it is fault tolerant or notIn our case we took rate = 01

Step 4 (misbehavior detection which is performed by CHi)

(i) For each node 119899119894in the cluster ldquo119894rdquo the state in time

slot ldquo119905rdquo is expressed by the three-dimensional vector

119878 = (1198781199051 1198781199052 1198781199053) (16)

where

(a) 1198781199051

is the number of packets dropped by 119899119894

defined as follows

1198781199051= sum119875119904

119877119890119888119890119894V119890119889 119887119910 119899119894 minussum119875119904119878119890119899119905 119887119910 119899119894

minussum119875119904119889119890119904119905119894119899119890119889 119887119910 119899119894

(17)

with

sum119875119904119877119890119888119890119894V119890119889 119887119910 119899119894 = sum119875119904119878119890119899119905119887119910 119899119894 +sum119875119904119889119890119904119905119894119899119890119889119887119910 119899119894

+sum119875119904119897119900119904119905 119887119910 119899119894

(18)

For a normal node 1198781199051asymp 0

(b) 1198781199052

is the delay between the arrival of twoconsecutive packets

1198781199052= 119863119890119897119886119910 119861119875 (119899

119894 119905) (19)

(c) 1198781199053is the energy consumption

1198781199053= 119864119888 (119899

119894 Δ119905) (20)

Here 119905 isin [1199050 t] = Δ119905

(ii) In our case the first interval is used for the trainingdata set of 119899 time slots We calculate the mean vector119878 of 119878 by using

119878 =

sum119905119899minus1

119905=1199050119878119905

119899

(21)

(iii) After modeling a normal behavior model for eachsensor node the behaviors of all nodes are sent to thebase station for further analysisWe then compute thedeviation 119889(119878) by using

119889 (119878) =

10038161003816100381610038161003816119878 minus 119878

10038161003816100381610038161003816 (22)

(iv) When the deviation 119889(119878) is larger than threshold 119879ℎ

(which means that it is out of the range of normalbehavior) it will be judged as a misbehaving node Inthis case the level of behavior is BL

119894asymp 0This is called

the punishing algorithm

119889 (119878) gt 119879ℎ 119899119894is an abnormal node

119889 (119878) le 119879ℎ 119899119894Is a normal node

(23)

The punishing algorithm is presented in Algorithm 3

6 Simulation Results

This section presents the implementation of the proposedapproach using the Borland C++ language and the analysisof the obtained results

61The Simulator ldquoMercuryrdquo We try to complete the theoret-ical study by implementing our own wireless sensor networksimulator ldquoMercuryrdquo On the other hand a bit of simulatorsfor WSNs such as TOSSIM [47] and Power-TOSSIM [48]are irrelevant with our goal and purpose and in order toavoid many complications we established our own mercurysimulator It is established on an object-oriented design anda distributed approach such as self-organization mechanismwhich is distributed at the level of each sensor it provides aset of interfaces for configuring a simulation and for choosingthe type of event scheduler used to drive the simulation Asimulation script generally begins by creating an instanceof this class and calling various methods to create nodesand topologies and configure other aspects of the simulationMercury uses two routing protocols for delivering data fromsensor nodes to the Sink station a reactive protocol AODV(ad hoc on demand distance vector) [5] and a proactiveprotocol DSDV (destination sequenced distance vector) [6]To determine and evaluate the results of the execution ofalgorithms that are introduced previously the number ofsensors to deploy must be inferior or equal to 1000 Thereare two types of sensor nodes deployment on the sensorfield random and manual Mercury offers users the abilityto select a sensor type from 5 types of existing sensor eachof them has its proper characteristics (radius energy etc)


Begin(1) 119868 = 0(2) 119868 = 119868 + 1(3) If ((119868 = Rate) ampamp (BL

119894lt= 01))

Rate parameter of maximum number of faultsdefined by the administrator

BL119894= BL119894minus Rate

(4) Classification of the node according to its BL119894

(5) Mp(BLi) =

Normal node 08 le BLi le 1

Abnormal node 05 le BLi lt 08

Suspect node 03 le BLi lt 05

Malicious node 0 le BLi lt 03

(6) If (BL119894le 03)Then

(7) If (119899119894is CM)Then

(8) Suppression of the node of the list of the members(9) Addition of the node to the blacklist

EndIf(10) If (119899

119894is CH)Then CH Cluster Head

(11) Addition of the node to the blacklist(12) Set up Phase

EndIfEndIf

EndIfEnd

Algorithm 3 Punishing algorithm

Unity of the energy used is as Nanojoules (1 Joule = 109NJ)Mobility has influence on energy and the behavior of sensorsfor instance if the sensor moves one meter away from itsoriginal location its energy will diminish by 100000NJ andits behavior will also decrease by 0001 unitsThis allows usersto differentiate a malicious node (that moves frequently) of alegitimate node (that can changes position with reasonabledistances) Since sensors nodes move due to the forces actingfrom the outside no power consumption for mobility mustbe taken into consideration in all simulations that we havecarried for evaluation [4]

62 Discussion and Results To evaluate our ES-WCA algo-rithm we have performed extensive simulation experimentsThis section provides our experimental results and discus-sions In all the experiments 119873 varies between 10 and 1000sensor nodes The transmission range (119877) varies between10 and 175 meters (m) and the used energy (119864) is equal to50000NJ The sensor nodes are randomly distributed in aldquo570m times 555mrdquo space area by the following function

119891119900119903 (119894119899119905 119899 = 0 119899 lt 119899119900119889119890 119905119900119887119890 119889119890119901119897119900119910119890119889 119899 + +)

119883 = rand() 119894119898119886119892119890 119865119894119890119897119889 119874119891 119862119900119897119897119890119888119905119894119899119892

rarr 119908119894119889119905ℎ119884 = rand() 119894119898119886119892119890 119865119894119890119897119889 119874119891 119862119900119897119897119890119888119905119894119899119892

rarr 119867119890119894119892ℎ119905

The performance of the proposed ES-WCA algorithmis measured by calculating (i) the number of clusters (ii)number of reaffiliations (iii) choice of ES-WCA with AODVor DSDV and (iiii) detection of misbehavior nodes and thenature of attacks during the distributed monitoring process

In our experiments the values of weighting factors usedin the weight calculation are as follows 119908

1= 03 119908

2=

02 1199083= 02 119908

4= 02 and 119908

5= 01 It is noted that these

values are arbitrary at this time and for this reason theyshould be adjusted according to the system requirements Toevaluate the performance of the proposedES-WCAalgorithmby comparing it with alternative solutions we studied theeffect of the density of the networks (number of sensor nodesin a given area) and the transmission range on the averagenumber of formed clustersThenwe compared it with aWCAproposed in [3]DWCAproposed in [9] and SDCAproposedin [21]

Figure 12 illustrates the variation of the average numberof clusters with respect to the transmission range The resultsare shown for 119873 which varies between 200 and 1000 Wefound that there is opposite relationship between clusters andtransmission rangeThis is on the grounds that a cluster headwith a considerable transmission range will cover a large area

Figure 13 depicts the average number of clusters that areformed with respect to the total number of nodes in thenetwork The communication range used in this experimentis 200m From Figure 13 it is seen that ES-WCA consistentlyprovides about 6191 less clusters than DWCA and about3846 than SDCA when there were 100 nodes in thenetwork When the node number is equal to 20 nodes


0

5

10

15

20

25

30

35

40

75 100 125 150 175Transmission range

Aver

age n

umbe

r of c

luste

rs

N = 200

N = 600

N = 1000

Figure 12 Average number of clusters versus transmission range(119877)

0

5

10

15

20

25

10 20 30 40 50 60 70 80 90 100

DWCAES-WCASDCA

Aver

age n

umbe

r of c

luste

rs

Number of nodes

Figure 13 Average number of clusters versus number nodes (119873) forES-WCA DWCA and SDCA

the performance of ES-WCA is similar to DWCA in termsof number of clusters however if the node density hadincreased ES-WCA would have produced constantly lessclusters than SDCA and DWCA respectively regardless ofthe node number Because of the use of a random deploy-ment the result of ES-WCA is unstable between 60 and 90So the increase in the number of clusters depends on theincrease of the distance between the nodes As a result ouralgorithm gave better performance in terms of the number ofclusterswhen the node density in the network is high and thisis due to the fact that ES-WCA generates a reduced number ofbalanced and homogeneous clusters whose size lies betweentwo thresholds 119879ℎ119903119890119904ℎ

119880119901119901119890119903and 119879ℎ119903119890119904ℎ

119871119900119908119890119903(reaffiliation

0

5

10

15

20

25

30

10 20 30 40 50 60 70Transmission range

Aver

age n

umbe

r of c

luste

rs

WCA N = 20

ES-WCA N = 20

WCA N = 40

ES-WCA N = 40

WCA N = 60

ES-WCA N = 60

Figure 14 Average number of clusters versus transmission rangeES-WCA andWCA

phase) in order to minimize the energy consumption of theentire network and prolong sensors lifetime

Figure 14 shows the variation of the average number ofclusters with respect to the transmission rangeThe results areshown for varying119873 We notice an inverse relationship andthe average number of clusters decreases with the increase inthe transmission range As shown in Figure 14 the proposedalgorithm produced 16 to 35 fewer clusters than WCA[3] when the transmission range of nodes was 10m Whenthe node density increased ES-WCA constantly producedless clusters than WCA regardless of the node number With70 nodes in the network the proposed algorithm producedabout 47 to 73 less clusters than WCA The results showthat our algorithm gave a better performance in terms of thenumber of clusters when the node density and transmissionrange in the network are high

Figure 15 interprets the average number of reaffiliationsthat are established with esteem to the total number of nodesin the network The number of reaffiliations incrementedlinearly when there were 30 or more nodes in the network forboth WCA and DWCA But for our algorithm the numberof reaffiliations increased starting from 50 nodes We submitto engender homogeneous clusters whose size is betweentwo thresholds 119879ℎ119903119890119904ℎ

119880119901119901119890119903= 18 and 119879ℎ119903119890119904ℎ

119871119900119908119890119903= 9

According to the results our algorithm presented a betterperformance in terms of the number of reaffiliations Thebenefit of decreasing the number of reaffiliations mainlycomes from the localized reaffiliation phase in our algorithmThe result of the remaining amount of energy per node foreach protocol AODV and DSDV is presented in Figure 16such as 119877 equal to 35m As shown in the above-mentionedfigure the remaining energy for each node inAODVprotocolis greater than that in DSDV protocol such as AODV whichconsumes 22 74 less than DSDV According to the resultsthe network consumes 19 23of the total energywhenweusean AODV protocol (192322091 NJ) However it consumes


0

05

1

15

2

25

3

35

4

10 20 30 40 50 60 70Number of nodes

Aver

age n

umbe

r of r

eaffi

liatio

ns

DWCAWCA

ES-WCA

Figure 15 Average number of reaffiliations

05000

100001500020000250003000035000400004500050000

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

AODVDSDV

Nodes

Rem

aini

ng en

ergy

Figure 16 Remaining energy per node using ES-WCA

41 97 with a DSDV protocol (419740129 NJ) We alsoobserve that the network lost 6 nodes with DSDV but onlyone node with AODV because of the depletion of its batteryThis result clearly shows that AODV outperforms DSDVThis is due to the tremendous overhead incurred by DSDVwhen exchanging routing tables and the periodic exchangeof the routing control packets So our algorithm gave a betterperformance in terms of saving energy when it is coupledwith AODV

We consider that the network will be inoperative whenthe nodes of the neighborhood of the sink exhaust theirenergy as exemplified In Figure 17 we appraise the networklifetime by changing the number of nodes such as 119877 equalto 70m When there were 20 nodes in the network AODVincreases the network period about 88 47 compared toDSDV and about 579 for 119873 = 100 Also this is for thereason that in a DSDV protocol each nodemust have a global

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

20 40 60 80 100

Protocol AODVProtocol DSDV

Number of nodes

Net

wor

k lif

etim

e (s)

Figure 17 Network lifetime depending on number of nodes usingES-WCA

Table 3 Detection of the nature of attacks

IDs Packets Sent Packets Received Attack41 (19 13) (16 14) Node Outage71 (24 152) (20 34) Hello Flood162 (15 8) (22 112) Sinkhole181 (16 179) (26 42) Hello Flood190 (58 32) (50 51) Black Hole

view of the network This in turn raises the number of theexchanged control packets (overhead) in the full network andit decreases the residual energy of each node which has adirect effect on the network lifetime There are 9 nodes in anactive state but the network is inoperative We discover thatthe increase in the total of nodes does not have a powerfulfactor on the network lifetime except between 119873 = 60 and119873 = 80

To illustrate the effect of abnormal behavior in thenetwork in our experiments we propagated 200 nodes with5 malicious nodes The cases of the malicious nodes will passfrom a normal node with a yellow color to an abnormal nodewith a blue color to a suspicious node with a grey color andlastly to a malicious node with a black color All the casesof the CMs are discovered by their CH Malicious CHs aredisclosed by the base station

Figure 18(b) displays the total of clusters establishedaccording to the transmission range Figures 19(a) 19(b) and19(c) display themeasure results for a scenario withmaliciousnodes which are achieved by the generator of bad behaviorThe generated attacks are explained in Section 3 We canidentify that these nodes migrate from a normal case to anabnormal or suspicious state and finally to a malicious stateas expected Table 3 presents the Ids of malicious nodes and


BS

(a)

BS

(b)

Figure 18 (a) Graph connectivity of 200 nodes (b) Network after clustering formation

BS

(a)

BS

(b)

BS

(c)

Figure 19 (a) Sensors with a blue color are abnormal but not malicious (b) The grey sensors have a suspect behavior (c) The sensors with ablack color are compromised and are exhibiting malicious behavior

their categories of attacks in the course of the disseminationof a monitoring mechanism in the network by the follow-up of the messages exchanged between the nodes WhenPackets sent [11987311198732] Packets received [11987331198734] Thus1198731is the total of packets sent before attacks and1198732 is the totalof packets sent after attacks while 1198733 is the total of packetsreceived before attacks and1198734 is the total of packets receivedafter attacksWe regard that these malicious nodes increment1198731 as the sensors (71 181) reduce1198731 like the sensor (190)increment 1198733 as the sensor (162) and lastly break sendingdata like node (41) From Figure 20 it is observed that thesensor nodes (3 17) are malicious and have a behavior level

less than 03 its behavior decreased by 01 units and whenthe monitor (CH) counts one failure an alarm is raisedHowever packets from malicious nodes are not processedand no packet will be forwarded to them The sensor node(11) has the behavior level less then threshold behavior so itwill not be accepted as a CH candidate even if it has the otherinteresting characteristics (Er

119894 119862119894119863119894 and119872

119894) On the other

side the behavior level in Figure 21 decreased by 0001 units inour first work [4] when the malicious node moves frequentlyWe note that sensor (6) is suspicious so if it continues tomove frequently its behavior will gradually be decreased untilit reaches the malicious state in this case this node will be


0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Before attacksAfter attacks

Sensor ID

Beha

vior

leve

l

Figure 20 Behavior level of some sensors (moves frequently)

0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20


Sensor ID

Beha

vior

leve

l

Figure 21 Behavior level of some sensors before and after attacks

deleted from the neighborhood and finally it will be added tothe black list

7 Conclusion and Future Works

In this paper we have presented a new algorithm called ldquoES-WCArdquo for promoting the self-organization of mobile sensornetworks This algorithm is fully decentralized and aims atcreating a virtual topology with the purpose to minimizefrequent reelection of the cluster head (CH) and avoid overallrestructuring of the entire network Simulations result attestof the outperformance of our algorithm compared to WCAand DWCA in every sense It yields a low number of clustersand it preserves the network structure better than WCAand DWCA by reducing the number of reaffiliations Theproposed algorithm selects the most robust and safe CHs

with the responsibility of monitoring the nodes in theirclusters andmaintaining clusters locally Our third algorithmanalyses and detects specific misbehavior in the WSNs Theresults show that in scenarios in which mobile WSNs arewith a low density or with a small size the choice of ES-WCA with AODV is comparable to ES-WCA with DSDV toshow clearly the interest of the routing protocols in energysaving However the difference in favor between ES-WCAand AODV becomes very important in case of a high nodedensity This is due to the tremendous overheads incurredby ES-WCAwith DSDVwhen exchanging routing tables andexchanging routing control packets Future work includesconsidering further the concept of redundancy by using theldquosleeprdquo and ldquowakeuprdquo mechanism in case of node failureproviding in-network processing by aggregating correlateddata in order to reduce both the energy consumption and thecongestion issue

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgment

The authors are grateful to the anonymous referees for theirinsightful comments and valuable suggestions which greatlyimproved the quality of the paper

References

[1] I F Akyildiz W Su Y Sankarasubramaniam and E CayircildquoWireless sensor networks a surveyrdquo Computer Networks vol38 no 4 pp 393ndash422 2002

[2] J Yick B Mukherjee and D Ghosal ldquoWireless sensor networksurveyrdquoComputerNetworks vol 52 no 12 pp 2292ndash2330 2008

[3] M Chatterjee S Das and D Turgut ldquoWCA a weightedclustering algorithm for mobile ad hoc networksrdquo Journal ofCluster Computing vol 5 no 2 pp 193ndash204 2002

[4] A Dahane N E Berrached and B Kechar ldquoEnergy efficientand safe weighted clustering algorithm for mobile wireless sen-sor networksrdquo in Proceedings of the 9th International Conferenceon Future Networks and Communications (FNC rsquo14) vol 34pp 63ndash70 Procedia Computer Science (Elsevier) Niagara FallsCanada August 2014

[5] Q Dong and W Dargie ldquoA survey on mobility and mobility-aware MAC protocols in wireless sensor networksrdquo IEEECommunications Surveys amp Tutorials vol 15 no 1 pp 88ndash1002011

[6] M Ali T Suleman and Z A Uzmi ldquoMMAC a mobility-adaptive collision-free MAC protocol for wireless sensor net-worksrdquo in Proceedings of the 24th IEEE International Perfor-mance Computing and Communications Conference (IPCCCrsquo05) pp 401ndash407 IEEE April 2005

[7] Y Yu and L Zhang ldquoA secure clustering algorithm in mobilead-hoc networksrdquo in Proceedings of the IACSIT Hong KongConferences vol 29 pp 73ndash77 2012

[8] I I Er and W K G Seah ldquoMobility-based d-hop clusteringalgorithm for mobile ad hoc networksrdquo in Proceedings of the


IEEE Wireless Communications and Networking Conference(WCNC rsquo04) pp 2359ndash2364 March 2004

[9] W Choi and M Woo ldquoA distributed weighted clustering algo-rithm for mobile ad hoc networksrdquo in Proceedings of the IEEEAdvanced International Conference on Telecommunications andInternational Conference on Internet and Web Applications andServices (AICTICIW 06) p 73 February 2006

[10] A Zabian A Ibrahim and F Al-Kalani ldquoDynamic head clusterelection algorithm for clustered Ad-Hoc networksrdquo Journal ofComputer Science vol 4 no 1 pp 42ndash50 2008

[11] M Chawla J Singhai and J L Rana ldquoClustering in mobilead- hoc networks a reviewrdquo International Journal of ComputerScience and Information Security vol 8 no 2 pp 293ndash301 2010

[12] R Agarwal R Gupta and M Motwani ldquoReview of weightedclustering algorithms for mobile ad-hoc networksrdquo ComputerScience and Telecommunications vol 33 no 1 pp 71ndash78 2012

[13] H Safa H Artail and D Tabet ldquoA cluster-based trust-awarerouting protocol for mobile ad hoc networksrdquo Wireless Net-works vol 16 no 4 pp 969ndash984 2010

[14] Sikander M Zafar A Raza M Babar S Mahmud and GKhan ldquoA survey of cluster-based routing schemes for wirelesssensor networksrdquo Smart Computing ReviewNetworks vol 3 no4 pp 261ndash275 2013

[15] X Liu ldquoA survey on clustering routing protocols in wirelesssensor networksrdquo Sensors vol 12 no 8 pp 11113ndash11153 2012

[16] S Soro and W B Heinzelman ldquoCluster head election tech-niques for coverage preservation in wireless sensor networksrdquoAd Hoc Networks vol 7 no 5 pp 955ndash972 2009

[17] A A Abbasi andM Younis ldquoA survey on clustering algorithmsfor wireless sensor networksrdquo Computer Communications Jour-nal vol 30 no 14-15 pp 2826ndash2841 2007

[18] K A Darabkh S S Ismail M Al-Shurman I F Jafar EAlkhader and M F Al-Mistarihi ldquoPerformance evaluationof selective and adaptive heads clustering algorithms overwireless sensor networksrdquo Journal of Network and ComputerApplications vol 35 no 6 pp 2068ndash2080 2012

[19] V Geetha P Kallapur and S Tellajeera ldquoClustering in wire-less sensor networks performance comparison of LEACH ampLEACH-Cprotocols usingNS2rdquo Procedia Technology vol 4 pp163ndash170 2012

[20] YWang XWu JWangW Liu andW Zheng ldquoAnOVSF codebased routing protocol for clustered wireless sensor networksrdquoInternational Journal of Future Generation Communication andNetworking vol 5 no 3 pp 117ndash128 2012

[21] K Benahmed M Merabti and H Haffaf ldquoDistributed moni-toring for misbehaviour detection in wireless sensor networksrdquoSecurity and Communication Networks vol 6 no 4 pp 388ndash400 2013

[22] J Y Yu and P H J Chong ldquoA survey of clustering schemes formobile ad hoc networksrdquo IEEE Communications Surveys andTutorials vol 7 no 1 pp 32ndash47 2005

[23] J Y Yu and P H J Chong ldquoA survey of clustering schemes formobile ad hoc networksrdquo IEEE Communications Surveys andTutorials vol 7 no 1ndash4 pp 32ndash47 2005

[24] A Jain and B V R Reddy ldquoA novel method of modelingwireless sensor network using fuzzy graph and energy efficientfuzzy based k-hop clustering algorithmrdquo Wireless PersonalCommunications vol 82 no 1 pp 157ndash181 2015

[25] T Kavita and D Sridharan ldquoSecurity vulnerabilities in wirelesssensor networks a surveyrdquo Journal of Information Assuranceand Security vol 5 pp 31ndash44 2010

[26] A Perrig R Szewczyk J D Tygar V Wen and D E CullerldquoSPINS security protocols for sensor networksrdquo Wireless Net-works vol 8 no 5 pp 521ndash534 2002

[27] S Ganeriwal and M B Srivastava ldquoReputation-based frame-work for high integrity sensor networksrdquo in Proceedings of the2nd ACMWorkshop on Security of Ad Hoc and Sensor Networks(SASN rsquo04) pp 66ndash77 October 2004

[28] Y Yu and L Zhang ldquoA secure clustering algorithm in mobilead-hoc networksrdquo in Proceedings of the 2012 IACSIT Hong KongConferences vol 29 pp 73ndash77 Hong Kong 2012


[30] T H Hai E-N Huh and M Jo ldquoA lightweight intrusiondetection framework for wireless sensor networksrdquo WirelessCommunications and Mobile Computing vol 10 no 4 pp 559ndash572 2010

[31] M E Elhdhili L B Azzouz and F Kamoun ldquoReputationbased clustering algorithm for security management in ad hocnetworks with liarsrdquo International Journal of Information andComputer Security vol 3 no 3-4 pp 228ndash244 2009

[32] S Taneja and A Kush ldquoA survey of routing protocols inmobile ad-hoc networksrdquo International Journal of InnovationManagement and Technology vol 1 no 3 pp 279ndash285 2010

[33] C E Perkins and P Bhagwat ldquoHighly dynamic destination-sequenced distance-vector routing (DSDV) for mobile com-putersrdquo in Proceedings of the Conference on CommunicationsArchitectures Protocols and Applications (SIGCOMM rsquo94) pp234ndash244 ACM London UK September 1994

[34] D G Padmavathi and D Shanmugapriya ldquoA survey of attackssecurity mechanisms and challenges in wireless sensor net-worksrdquo International Journal of Computer Science and Informa-tion Security vol 4 no 1-2 pp 1ndash9 2009

[35] P Li L Sun X Fu and L Ning ldquoSecurity in wireless sensornetworksrdquo in Wireless Network Security pp 179ndash227 HigherEducation Press Springer Berlin Germany 2013

[36] W Stallings Cryptography and Network Security Principles andPractice Prentice Hall 5th edition 2010

[37] M Safiqul-Islam and S Ashiqur-Rahman ldquoAnomaly intrusiondetection system in wireless sensor networks security threatsand existing approachesrdquo International Journal of AdvancedScience and Technology vol 36 pp 1ndash8 2011

[38] P Berwal ldquoSecurity in wireless sensor networks issues andchallengesrdquo International Journal of Engineering and InnovativeTechnology vol 3 no 5 pp 192ndash198 2013

[39] C Karlof and D Wagner ldquoSecure routing in wireless sensornetworks attacks and countermeasuresrdquo Ad-Hoc NetworksJournal vol 1 no 2-3 pp 293ndash315 2003

[40] S Dai X Jing and L Li ldquoResearch and analysis on routingprotocols for wireless sensor networksrdquo in Proceedings of theInternational Conference on Communications Circuits and Sys-tems vol 1 pp 407ndash411 May 2005

[41] M Tripathi M S Gaur and V Laxmi ldquoComparing the impactof black hole and gray hole attack on LEACH in WSNrdquo inProceedings of the 4th International Conference on AmbientSystems Networks and Technologies (ANT rsquo13) and the 3rdInternational Conference on Sustainable Energy InformationTechnology (SEIT rsquo13) vol 19 pp 1101ndash1107 June 2013

[42] R A Shaikh H Jameel S Lee S Rajput and Y J Song ldquoTrustmanagement problem in distributed wireless sensor networksrdquoin Proceedings of the 12th IEEE International Conference onEmbedded and Real-Time Computing Systems and Applications(RTCSA rsquo06) pp 411ndash414 August 2006


[43] M Lehsaini H Guyennet and M Feham ldquoAn efficient cluster-based self-organisation algorithm for wireless sensor networksrdquoInternational Journal of Sensor Networks vol 7 no 1-2 pp 85ndash94 2010

[44] Y Li F Wang F Huang and D Yang ldquoA novel enhancedweighted clustering algorithm for mobile networksrdquo in Pro-ceedings of the IEEE 5th International Conference on WirelessCommunications Networking and Mobile Computing (WiCOMrsquo09) pp 2801ndash2804 IEEE Beijing China September 2009

[45] A da Silva M Martins B Rocha A Loureiro L Ruiz andHWong ldquoDecentralized intrusion detection in wireless sensornetworksrdquo inProceedings of the 1st ACM InternationalWorkshoponQuality of Serviceamp Security inWireless andMobileNetworkspp 16ndash23 2005

[46] K Benahmed H Haffaf and M Merabti ldquoMonitoring of wire-less sensor networksrdquo in Sustainable Wireless Sensor NetworksY K Tan Ed chapter 3 InTech 2010

[47] P Levis N Lee M Welsh and D Culler ldquoTOSSIM accurateand scalable simulation of entire TinyOS applicationsrdquo inProceedings of the 1st International Conference on EmbeddedNetworked Sensor Systems (SenSys rsquo03) pp 126ndash137 November2003

[48] V Shnayder M Hempstead B-R Chen G W Allen andM Welsh ldquoSimulating the power consumption of large-scalesensor network applicationsrdquo in Proceedings of the 2nd Inter-national Conference on Embedded Networked Sensor Systems(SenSys 04) pp 188ndash200 November 2004

Submit your manuscripts athttpwwwhindawicom

Computer Games Technology

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Distributed Sensor Networks


Advances in

FuzzySystems

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014


ReconfigurableComputing

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014


Applied Computational Intelligence and Soft Computing

thinspAdvancesthinspinthinsp

Artificial Intelligence

HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014

Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Electrical and Computer Engineering

Journal of

Journal of

Computer Networks and Communications


Hindawi Publishing Corporation

httpwwwhindawicom Volume 2014

Advances in

Multimedia


Biomedical Imaging


ArtificialNeural Systems

Advances in


RoboticsJournal of



Computational Intelligence and Neuroscience

Industrial EngineeringJournal of


Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014


Human-ComputerInteraction

Advances in

Computer EngineeringAdvances in



BSCM1

CM2

CH2

G1

G2

CH1

CH3

Figure 1 Clustering formation of WSNs composed of 150 sensornodes deployed in a 570m times 555m space area with a radio range =100m

work and consequently consume more energy comparedto CMs during the network operations and this will leadto untimely death causing network partition and thereforefailure in communication link For this reason one of themost frequently encountered problems in this mechanism isto search for the best way to elect CH for each cluster Indeeda CH can be selected by computing the quality of nodes Thismay depend on severalmetrics connectivity degreemobilityresidual energy and the distance of a node from its neighborsSignificant improvement in performance of this quality canbe achieved by combining these metrics [3 9 10 12 19 21]

In this paper we propose an energy efficient and safeweighted clustering algorithm for mobile WSNs using acombination of the above metrics to which we added abehavioral level metric The latter metric is decisive andallows the proposed clustering algorithm to avoid any mali-cious node in the neighborhood to become a CH even ifthe remaining metrics are in its favor The election of CHsis carried out using weights of neighboring nodes whichare computed based on selected metrics So this strategyensures the election of legitimate CHs with high weightsThe preliminary results obtained through simulation studydemonstrate the effectiveness of our algorithm in terms ofthe number of equilibrate clusters and the number of reaffili-ations compared to WCA (Weighted Clustering Algorithm)[3] DWCA (Distributed Weighted Clustering Algorithm)[9] and SDCA (Secure Distributed Clustering Algorithm)[21] These results also reveal that our approach is suitable ifwe plan to use it in network layer reactive routing protocolsinstead of proactive ones once the clustering mechanism islaunched

We can enumerate the contributions of our paper asfollows

(i) maintaining stable clustering structure and offeringa better performance in terms of the number ofreaffiliations using the proposed algorithm ES-WCA(Energy Efficient and SafeWeighted Clustering Algo-rithm)

(ii) detecting common routing problems and attacks inclustered WSNs based on behavior level

(iii) showing clearly the interest of the routing protocols inenergy saving and therefore maximizing the lifetimeof the global network

The remaining part of this paper is organized as followsSection 2 briefly surveys the related works on clusteringalgorithms proposed for ad hoc networks and in particularthose developed for WSNs In Section 3 we emphasize onthe security problems in WSNs Section 4 introduces andexplains the selected metrics for the proposed approach ofclustering More details on the proposed algorithm are givenin Section 5 Section 6 presents the simulation tool developedfor evaluation Simulation results are provided to show theeffectiveness of the proposed algorithm Section 7 concludesthe paper and outline directions of future works

2 Related Works

In this section we outline some approaches of clustering usedin ad hoc networks andWSNs Research studies on clusteringin ad hoc networks involve surveyed works on clusteringalgorithms [11 22] and cluster head election algorithms[10 16] Abbasi and Younis [17] presented taxonomy andclassification of typical clustering schemes then summa-rized different clustering algorithms for WSNs based onclassification of variable convergence time protocols andconstant convergence time algorithms and highlighted theirobjectives features complexity and so forth A single metricbased on clustering as in paper [23] shows that the nodewith the least stability value is elected as CH among itsneighbors However the choice of CH which has a lowerenergy level could quickly become a bottleneck of its clusterEr and Seah [8] designed and implemented a dynamic energyefficient clustering algorithm (DEECA) for mobile ad hocnetworks (MANETs) that increases the network lifetimeTheproposed model elects first the nodes that have a higherenergy and less mobility as cluster heads then periodicallymonitors the cluster headrsquos energy and locally alters theclusters to reduce the energy consumption of the sufferingcluster heads The algorithm defines a yellow threshold toachieve some sort of local load balancing and a red thresholdto trigger local reclustering in the network However thecluster formation in this scheme is not based on connectivityso the formed clusters are not well connected consequentlythis increases the reaffiliation rate andmaximizes reclusteringsituations Jain andReddy [24] have proposed a novelmethodof modeling wireless sensor network using fuzzy graphand energy efficient fuzzy based k-Hop clustering algorithmwhich takes into account the dynamic nature of networkvolatile aspects of radio links and physical layer uncertaintyThey have defined a new centrality metric namely fuzzy






3 Security in WSNs








0 03 05 08 1

Behavior level








119894 noted BL



Mp (BL119894) =





(1)




119872119894=

1

119879

119879

sum

119905=1


where (119909119905 119910119905) and (119909


119894




119894)


by

119863119894= sum

119895 isin 119873(119894)

dist (119894 119895) (3)




Er119894= 119864 minus (119864

119879119909 (119896 119889) + 119864119877119909 elec (119896)) (4)

where







119894)


according to [4]

119862119894= |119873 (119894)| (5)


5

13

6

24

Cluster

Cluster head (CH)

Cluster member (CM)


Malicious node


(a)

5

5

4

4

31

1

4

4

4

6

2

Cluster

Initiallocation

Finallocation

Cluster head (CH)




(b)


where





the equation


lowast 119863119894

(6)

where1199081119908211990831199084 and119908



1199081+ 1199082+ 1199083+ 1199084+ 1199085= 1 (7)


119880119901119901119890119903and 119879ℎ119903119890119904ℎ

119871119900119908119890119903



12057512 (119906) = min (120575

12(119906119894) 119906119894isin 119880) (8)


12057512 (

V) = min (12057512(V119894) V119894isin 119880) (9)


AVG =sum119899

119894=112057512(119906119894)

119873

(10)


119879ℎ119903119890119904ℎ119880119901119901119890119903

=

1

2

(12057512 (119906) + AVG)

119879ℎ119903119890119904ℎ119871119900119908119890119903

=

1

2

(12057512 (

V) + AVG) (11)


119894119872119894 119863119894Er119894 and 119862


























U CH

ACCEPT_CH message

REQ_JOIN message

ADV_CH message


U

CH




Ids BL119894

Er119894

119862119894

119863119894

119872119894

119875119894

1 086 384212 3 115 120 7696324 081 483254 5 230 030 9681335 088 405325 3 130 055 8118296 085 462043 0 000 020 9243618 081 481680 4 105 140 96475310 095 365025 2 055 010 73080511 091 481960 1 070 220 964753




119871119900119908119890119903to those





119894




1 1199082 1199083 1199084 1199085




119894














119894rarr CH = 0)




119894



(30) 119899119894rarr CH = CH rarr Id


EndIfEnd For




Ids 1 4 5 6 8 10 111 769632 964753 9647534 968133 811829 9647535 968133 811829 7308056 9243618 769632 96475310 968133 811829 73080511 769632 964753


Inputs 119879ℎ119903119890119904ℎ119880119901119901119890119903

119879ℎ119903119890119904ℎ119871119900119908119890119903


119880119901119901119890119903)


(119873(CH))(4) 119869 = Count (119873(CH))

EndIf(5) For 119868 = 1 to 119869 Do(6) If (119899



119871119900119908119890119903)

Then(7) 119899



)Then




(13) 119899119894rarr CH rarr ID = ID


(15) Go to (2)EndIF


EndIfEnd For

End ForEnd


12055

7048

10095

2036

3045

5088

4081

8081

9050

1

086

11091

6

085






119880119901119901119890119903= 9 and 119879ℎ119903119890119904ℎ

119871119900119908119890119903= 6 For that


119871119900119908119890119903to



119880119901119901119890119903 Nodes 1 7 and


119871119900119908119890119903 so




Cluster 2

10

095

7

048

2

036

5

088

3

045

4

081

8

081

9

050

1

086

11091

6

085

12055

Cluster 4 Cluster 3

Cluster 1


Cluster 2

10

095

7

048

2

036

5

088

3

045

4

081

8

081

9

0501

086

11091

6

085

12055

Cluster 4

Cluster 5

Cluster 6

Cluster 3

Cluster 1


Cluster 210

095

7

048

2

036

5

088

3

045

4

081

8

081

9

0501

086

11091

6

085

12055

Cluster 4

Cluster 3

Cluster 1





will be repeated




Cluster 2

Cluster 1

BS

Local monitoring

Global monitoring









Yes

NoNo

ni is a normal node



d(S) gt Th




119894becomes the







[1199050 119905] 119873119887119901 119878119890119899119889(119899119894 Δ119905)


interval is Δ119905 = [1199050 1199050] 119873119887119901 119877119890119888119890119894V119890119889(119899

119894 Δ119905)


119863119890119897119886119910 119861119875 (119899119894 119905) = 119860119903119903119894V119886119897 119875119879

119894minus 119860119903119903119894V119886119897 119875119879

119894minus1 (12)


119864119888 (119899119894 Δ119905) = Er (119899

119894 1199050) minus Er (119899

119894 1199051) (13)


119894 1199050) is the


0 Er(119899

119894 1199051) is the


1and 119864119888(119899

119894 Δ119905) is



119894at time 119905 is






mechanism



(119868119889119873119887119901119878119890119899119889(119899119894Δ119905)

119873119887119901119877119890119888119890119894V119890119889(119899119894 Δ119905) 119863119890119897119886119910119861119875(119899119894 119905)

119864119888 (119899119894 Δ119905))

(14)


119894 119905119894minus1

) gt 120598)


old)Msg = (119868119889119873119887119901

119878119890119899119889(119899119894Δ119905) 119873119887119901

119877119890119888119890119894V119890119889(119899119894 Δ119905)119863119890119897119886119910

119861119875(119899119894 119905) 119864119888(119899

119894 Δ119905)) to its CHi for





119894is computed


BL119894= BL119894minus rate (15)





119878 = (1198781199051 1198781199052 1198781199053) (16)

where

(a) 1198781199051


defined as follows

1198781199051= sum119875119904

119877119890119888119890119894V119890119889 119887119910 119899119894 minussum119875119904119878119890119899119905 119887119910 119899119894

minussum119875119904119889119890119904119905119894119899119890119889 119887119910 119899119894

(17)

with

sum119875119904119877119890119888119890119894V119890119889 119887119910 119899119894 = sum119875119904119878119890119899119905119887119910 119899119894 +sum119875119904119889119890119904119905119894119899119890119889119887119910 119899119894

+sum119875119904119897119900119904119905 119887119910 119899119894

(18)


(b) 1198781199052


1198781199052= 119863119890119897119886119910 119861119875 (119899

119894 119905) (19)


1198781199053= 119864119888 (119899

119894 Δ119905) (20)

Here 119905 isin [1199050 t] = Δ119905


119878 =

sum119905119899minus1

119905=1199050119878119905

119899

(21)


119889 (119878) =

10038161003816100381610038161003816119878 minus 119878

10038161003816100381610038161003816 (22)







(23)







119894lt= 01))




(5) Mp(BLi) =





(6) If (BL119894le 03)Then

(7) If (119899119894is CM)Then


EndIf(10) If (119899



EndIfEndIf

EndIfEnd




119891119900119903 (119894119899119905 119899 = 0 119899 lt 119899119900119889119890 119905119900119887119890 119889119890119901119897119900119910119890119889 119899 + +)

119883 = rand() 119894119898119886119892119890 119865119894119890119897119889 119874119891 119862119900119897119897119890119888119905119894119899119892

rarr 119908119894119889119905ℎ119884 = rand() 119894119898119886119892119890 119865119894119890119897119889 119874119891 119862119900119897119897119890119888119905119894119899119892

rarr 119867119890119894119892ℎ119905



1= 03 119908

2=

02 1199083= 02 119908

4= 02 and 119908






0

5

10

15

20

25

30

35

40


Aver

age n

umbe

r of c

luste

rs

N = 200

N = 600

N = 1000


0

5

10

15

20

25

10 20 30 40 50 60 70 80 90 100

DWCAES-WCASDCA

Aver

age n

umbe

r of c

luste

rs

Number of nodes



119880119901119901119890119903and 119879ℎ119903119890119904ℎ

119871119900119908119890119903(reaffiliation

0

5

10

15

20

25

30


Aver

age n

umbe

r of c

luste

rs

WCA N = 20

ES-WCA N = 20

WCA N = 40

ES-WCA N = 40

WCA N = 60

ES-WCA N = 60





119880119901119901119890119903= 18 and 119879ℎ119903119890119904ℎ

119871119900119908119890119903= 9



0

05

1

15

2

25

3

35

4

10 20 30 40 50 60 70Number of nodes

Aver

age n

umbe

r of r

eaffi

liatio

ns

DWCAWCA

ES-WCA


05000

100001500020000250003000035000400004500050000

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

AODVDSDV

Nodes

Rem

aini

ng en

ergy




0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

20 40 60 80 100


Number of nodes

Net

wor

k lif

etim

e (s)








BS

(a)

BS

(b)


BS

(a)

BS

(b)

BS

(c)




119894 119862119894119863119894 and119872




0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20


Sensor ID

Beha

vior

leve

l


0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20


Sensor ID

Beha

vior

leve

l








Acknowledgment


References



























































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in








3 Security in WSNs








0 03 05 08 1

Behavior level








119894 noted BL



Mp (BL119894) =





(1)




119872119894=

1

119879

119879

sum

119905=1


where (119909119905 119910119905) and (119909


119894




119894)


by

119863119894= sum

119895 isin 119873(119894)

dist (119894 119895) (3)




Er119894= 119864 minus (119864

119879119909 (119896 119889) + 119864119877119909 elec (119896)) (4)

where







119894)


according to [4]

119862119894= |119873 (119894)| (5)


5

13

6

24

Cluster

Cluster head (CH)

Cluster member (CM)


Malicious node


(a)

5

5

4

4

31

1

4

4

4

6

2

Cluster

Initiallocation

Finallocation

Cluster head (CH)




(b)


where





the equation


lowast 119863119894

(6)

where1199081119908211990831199084 and119908



1199081+ 1199082+ 1199083+ 1199084+ 1199085= 1 (7)


119880119901119901119890119903and 119879ℎ119903119890119904ℎ

119871119900119908119890119903



12057512 (119906) = min (120575

12(119906119894) 119906119894isin 119880) (8)


12057512 (

V) = min (12057512(V119894) V119894isin 119880) (9)


AVG =sum119899

119894=112057512(119906119894)

119873

(10)


119879ℎ119903119890119904ℎ119880119901119901119890119903

=

1

2

(12057512 (119906) + AVG)

119879ℎ119903119890119904ℎ119871119900119908119890119903

=

1

2

(12057512 (

V) + AVG) (11)


119894119872119894 119863119894Er119894 and 119862


























U CH

ACCEPT_CH message

REQ_JOIN message

ADV_CH message


U

CH




Ids BL119894

Er119894

119862119894

119863119894

119872119894

119875119894

1 086 384212 3 115 120 7696324 081 483254 5 230 030 9681335 088 405325 3 130 055 8118296 085 462043 0 000 020 9243618 081 481680 4 105 140 96475310 095 365025 2 055 010 73080511 091 481960 1 070 220 964753




119871119900119908119890119903to those





119894




1 1199082 1199083 1199084 1199085




119894














119894rarr CH = 0)




119894



(30) 119899119894rarr CH = CH rarr Id


EndIfEnd For




Ids 1 4 5 6 8 10 111 769632 964753 9647534 968133 811829 9647535 968133 811829 7308056 9243618 769632 96475310 968133 811829 73080511 769632 964753


Inputs 119879ℎ119903119890119904ℎ119880119901119901119890119903

119879ℎ119903119890119904ℎ119871119900119908119890119903


119880119901119901119890119903)


(119873(CH))(4) 119869 = Count (119873(CH))

EndIf(5) For 119868 = 1 to 119869 Do(6) If (119899



119871119900119908119890119903)

Then(7) 119899



)Then




(13) 119899119894rarr CH rarr ID = ID


(15) Go to (2)EndIF


EndIfEnd For

End ForEnd


12055

7048

10095

2036

3045

5088

4081

8081

9050

1

086

11091

6

085






119880119901119901119890119903= 9 and 119879ℎ119903119890119904ℎ

119871119900119908119890119903= 6 For that


119871119900119908119890119903to



119880119901119901119890119903 Nodes 1 7 and


119871119900119908119890119903 so




Cluster 2

10

095

7

048

2

036

5

088

3

045

4

081

8

081

9

050

1

086

11091

6

085

12055

Cluster 4 Cluster 3

Cluster 1


Cluster 2

10

095

7

048

2

036

5

088

3

045

4

081

8

081

9

0501

086

11091

6

085

12055

Cluster 4

Cluster 5

Cluster 6

Cluster 3

Cluster 1


Cluster 210

095

7

048

2

036

5

088

3

045

4

081

8

081

9

0501

086

11091

6

085

12055

Cluster 4

Cluster 3

Cluster 1





will be repeated




Cluster 2

Cluster 1

BS

Local monitoring

Global monitoring









Yes

NoNo

ni is a normal node



d(S) gt Th




119894becomes the







[1199050 119905] 119873119887119901 119878119890119899119889(119899119894 Δ119905)


interval is Δ119905 = [1199050 1199050] 119873119887119901 119877119890119888119890119894V119890119889(119899

119894 Δ119905)


119863119890119897119886119910 119861119875 (119899119894 119905) = 119860119903119903119894V119886119897 119875119879

119894minus 119860119903119903119894V119886119897 119875119879

119894minus1 (12)


119864119888 (119899119894 Δ119905) = Er (119899

119894 1199050) minus Er (119899

119894 1199051) (13)


119894 1199050) is the


0 Er(119899

119894 1199051) is the


1and 119864119888(119899

119894 Δ119905) is



119894at time 119905 is






mechanism



(119868119889119873119887119901119878119890119899119889(119899119894Δ119905)

119873119887119901119877119890119888119890119894V119890119889(119899119894 Δ119905) 119863119890119897119886119910119861119875(119899119894 119905)

119864119888 (119899119894 Δ119905))

(14)


119894 119905119894minus1

) gt 120598)


old)Msg = (119868119889119873119887119901

119878119890119899119889(119899119894Δ119905) 119873119887119901

119877119890119888119890119894V119890119889(119899119894 Δ119905)119863119890119897119886119910

119861119875(119899119894 119905) 119864119888(119899

119894 Δ119905)) to its CHi for





119894is computed


BL119894= BL119894minus rate (15)





119878 = (1198781199051 1198781199052 1198781199053) (16)

where

(a) 1198781199051


defined as follows

1198781199051= sum119875119904

119877119890119888119890119894V119890119889 119887119910 119899119894 minussum119875119904119878119890119899119905 119887119910 119899119894

minussum119875119904119889119890119904119905119894119899119890119889 119887119910 119899119894

(17)

with

sum119875119904119877119890119888119890119894V119890119889 119887119910 119899119894 = sum119875119904119878119890119899119905119887119910 119899119894 +sum119875119904119889119890119904119905119894119899119890119889119887119910 119899119894

+sum119875119904119897119900119904119905 119887119910 119899119894

(18)


(b) 1198781199052


1198781199052= 119863119890119897119886119910 119861119875 (119899

119894 119905) (19)


1198781199053= 119864119888 (119899

119894 Δ119905) (20)

Here 119905 isin [1199050 t] = Δ119905


119878 =

sum119905119899minus1

119905=1199050119878119905

119899

(21)


119889 (119878) =

10038161003816100381610038161003816119878 minus 119878

10038161003816100381610038161003816 (22)







(23)







119894lt= 01))




(5) Mp(BLi) =





(6) If (BL119894le 03)Then

(7) If (119899119894is CM)Then


EndIf(10) If (119899



EndIfEndIf

EndIfEnd




119891119900119903 (119894119899119905 119899 = 0 119899 lt 119899119900119889119890 119905119900119887119890 119889119890119901119897119900119910119890119889 119899 + +)

119883 = rand() 119894119898119886119892119890 119865119894119890119897119889 119874119891 119862119900119897119897119890119888119905119894119899119892

rarr 119908119894119889119905ℎ119884 = rand() 119894119898119886119892119890 119865119894119890119897119889 119874119891 119862119900119897119897119890119888119905119894119899119892

rarr 119867119890119894119892ℎ119905



1= 03 119908

2=

02 1199083= 02 119908

4= 02 and 119908






0

5

10

15

20

25

30

35

40


Aver

age n

umbe

r of c

luste

rs

N = 200

N = 600

N = 1000


0

5

10

15

20

25

10 20 30 40 50 60 70 80 90 100

DWCAES-WCASDCA

Aver

age n

umbe

r of c

luste

rs

Number of nodes



119880119901119901119890119903and 119879ℎ119903119890119904ℎ

119871119900119908119890119903(reaffiliation

0

5

10

15

20

25

30


Aver

age n

umbe

r of c

luste

rs

WCA N = 20

ES-WCA N = 20

WCA N = 40

ES-WCA N = 40

WCA N = 60

ES-WCA N = 60





119880119901119901119890119903= 18 and 119879ℎ119903119890119904ℎ

119871119900119908119890119903= 9



0

05

1

15

2

25

3

35

4

10 20 30 40 50 60 70Number of nodes

Aver

age n

umbe

r of r

eaffi

liatio

ns

DWCAWCA

ES-WCA


05000

100001500020000250003000035000400004500050000

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

AODVDSDV

Nodes

Rem

aini

ng en

ergy




0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

20 40 60 80 100


Number of nodes

Net

wor

k lif

etim

e (s)








BS

(a)

BS

(b)


BS

(a)

BS

(b)

BS

(c)




119894 119862119894119863119894 and119872




0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20


Sensor ID

Beha

vior

leve

l


0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20


Sensor ID

Beha

vior

leve

l








Acknowledgment


References



























































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in





0 03 05 08 1

Behavior level








119894 noted BL



Mp (BL119894) =





(1)




119872119894=

1

119879

119879

sum

119905=1


where (119909119905 119910119905) and (119909


119894




119894)


by

119863119894= sum

119895 isin 119873(119894)

dist (119894 119895) (3)




Er119894= 119864 minus (119864

119879119909 (119896 119889) + 119864119877119909 elec (119896)) (4)

where







119894)


according to [4]

119862119894= |119873 (119894)| (5)


5

13

6

24

Cluster

Cluster head (CH)

Cluster member (CM)


Malicious node


(a)

5

5

4

4

31

1

4

4

4

6

2

Cluster

Initiallocation

Finallocation

Cluster head (CH)




(b)


where





the equation


lowast 119863119894

(6)

where1199081119908211990831199084 and119908



1199081+ 1199082+ 1199083+ 1199084+ 1199085= 1 (7)


119880119901119901119890119903and 119879ℎ119903119890119904ℎ

119871119900119908119890119903



12057512 (119906) = min (120575

12(119906119894) 119906119894isin 119880) (8)


12057512 (

V) = min (12057512(V119894) V119894isin 119880) (9)


AVG =sum119899

119894=112057512(119906119894)

119873

(10)


119879ℎ119903119890119904ℎ119880119901119901119890119903

=

1

2

(12057512 (119906) + AVG)

119879ℎ119903119890119904ℎ119871119900119908119890119903

=

1

2

(12057512 (

V) + AVG) (11)


119894119872119894 119863119894Er119894 and 119862


























U CH

ACCEPT_CH message

REQ_JOIN message

ADV_CH message


U

CH




Ids BL119894

Er119894

119862119894

119863119894

119872119894

119875119894

1 086 384212 3 115 120 7696324 081 483254 5 230 030 9681335 088 405325 3 130 055 8118296 085 462043 0 000 020 9243618 081 481680 4 105 140 96475310 095 365025 2 055 010 73080511 091 481960 1 070 220 964753




119871119900119908119890119903to those





119894




1 1199082 1199083 1199084 1199085




119894














119894rarr CH = 0)




119894



(30) 119899119894rarr CH = CH rarr Id


EndIfEnd For




Ids 1 4 5 6 8 10 111 769632 964753 9647534 968133 811829 9647535 968133 811829 7308056 9243618 769632 96475310 968133 811829 73080511 769632 964753


Inputs 119879ℎ119903119890119904ℎ119880119901119901119890119903

119879ℎ119903119890119904ℎ119871119900119908119890119903


119880119901119901119890119903)


(119873(CH))(4) 119869 = Count (119873(CH))

EndIf(5) For 119868 = 1 to 119869 Do(6) If (119899



119871119900119908119890119903)

Then(7) 119899



)Then




(13) 119899119894rarr CH rarr ID = ID


(15) Go to (2)EndIF


EndIfEnd For

End ForEnd


12055

7048

10095

2036

3045

5088

4081

8081

9050

1

086

11091

6

085






119880119901119901119890119903= 9 and 119879ℎ119903119890119904ℎ

119871119900119908119890119903= 6 For that


119871119900119908119890119903to



119880119901119901119890119903 Nodes 1 7 and


119871119900119908119890119903 so




Cluster 2

10

095

7

048

2

036

5

088

3

045

4

081

8

081

9

050

1

086

11091

6

085

12055

Cluster 4 Cluster 3

Cluster 1


Cluster 2

10

095

7

048

2

036

5

088

3

045

4

081

8

081

9

0501

086

11091

6

085

12055

Cluster 4

Cluster 5

Cluster 6

Cluster 3

Cluster 1


Cluster 210

095

7

048

2

036

5

088

3

045

4

081

8

081

9

0501

086

11091

6

085

12055

Cluster 4

Cluster 3

Cluster 1





will be repeated




Cluster 2

Cluster 1

BS

Local monitoring

Global monitoring









Yes

NoNo

ni is a normal node



d(S) gt Th




119894becomes the







[1199050 119905] 119873119887119901 119878119890119899119889(119899119894 Δ119905)


interval is Δ119905 = [1199050 1199050] 119873119887119901 119877119890119888119890119894V119890119889(119899

119894 Δ119905)


119863119890119897119886119910 119861119875 (119899119894 119905) = 119860119903119903119894V119886119897 119875119879

119894minus 119860119903119903119894V119886119897 119875119879

119894minus1 (12)


119864119888 (119899119894 Δ119905) = Er (119899

119894 1199050) minus Er (119899

119894 1199051) (13)


119894 1199050) is the


0 Er(119899

119894 1199051) is the


1and 119864119888(119899

119894 Δ119905) is



119894at time 119905 is






mechanism



(119868119889119873119887119901119878119890119899119889(119899119894Δ119905)

119873119887119901119877119890119888119890119894V119890119889(119899119894 Δ119905) 119863119890119897119886119910119861119875(119899119894 119905)

119864119888 (119899119894 Δ119905))

(14)


119894 119905119894minus1

) gt 120598)


old)Msg = (119868119889119873119887119901

119878119890119899119889(119899119894Δ119905) 119873119887119901

119877119890119888119890119894V119890119889(119899119894 Δ119905)119863119890119897119886119910

119861119875(119899119894 119905) 119864119888(119899

119894 Δ119905)) to its CHi for





119894is computed


BL119894= BL119894minus rate (15)





119878 = (1198781199051 1198781199052 1198781199053) (16)

where

(a) 1198781199051


defined as follows

1198781199051= sum119875119904

119877119890119888119890119894V119890119889 119887119910 119899119894 minussum119875119904119878119890119899119905 119887119910 119899119894

minussum119875119904119889119890119904119905119894119899119890119889 119887119910 119899119894

(17)

with

sum119875119904119877119890119888119890119894V119890119889 119887119910 119899119894 = sum119875119904119878119890119899119905119887119910 119899119894 +sum119875119904119889119890119904119905119894119899119890119889119887119910 119899119894

+sum119875119904119897119900119904119905 119887119910 119899119894

(18)


(b) 1198781199052


1198781199052= 119863119890119897119886119910 119861119875 (119899

119894 119905) (19)


1198781199053= 119864119888 (119899

119894 Δ119905) (20)

Here 119905 isin [1199050 t] = Δ119905


119878 =

sum119905119899minus1

119905=1199050119878119905

119899

(21)


119889 (119878) =

10038161003816100381610038161003816119878 minus 119878

10038161003816100381610038161003816 (22)







(23)







119894lt= 01))




(5) Mp(BLi) =





(6) If (BL119894le 03)Then

(7) If (119899119894is CM)Then


EndIf(10) If (119899



EndIfEndIf

EndIfEnd




119891119900119903 (119894119899119905 119899 = 0 119899 lt 119899119900119889119890 119905119900119887119890 119889119890119901119897119900119910119890119889 119899 + +)

119883 = rand() 119894119898119886119892119890 119865119894119890119897119889 119874119891 119862119900119897119897119890119888119905119894119899119892

rarr 119908119894119889119905ℎ119884 = rand() 119894119898119886119892119890 119865119894119890119897119889 119874119891 119862119900119897119897119890119888119905119894119899119892

rarr 119867119890119894119892ℎ119905



1= 03 119908

2=

02 1199083= 02 119908

4= 02 and 119908






0

5

10

15

20

25

30

35

40


Aver

age n

umbe

r of c

luste

rs

N = 200

N = 600

N = 1000


0

5

10

15

20

25

10 20 30 40 50 60 70 80 90 100

DWCAES-WCASDCA

Aver

age n

umbe

r of c

luste

rs

Number of nodes



119880119901119901119890119903and 119879ℎ119903119890119904ℎ

119871119900119908119890119903(reaffiliation

0

5

10

15

20

25

30


Aver

age n

umbe

r of c

luste

rs

WCA N = 20

ES-WCA N = 20

WCA N = 40

ES-WCA N = 40

WCA N = 60

ES-WCA N = 60





119880119901119901119890119903= 18 and 119879ℎ119903119890119904ℎ

119871119900119908119890119903= 9



0

05

1

15

2

25

3

35

4

10 20 30 40 50 60 70Number of nodes

Aver

age n

umbe

r of r

eaffi

liatio

ns

DWCAWCA

ES-WCA


05000

100001500020000250003000035000400004500050000

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

AODVDSDV

Nodes

Rem

aini

ng en

ergy




0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

20 40 60 80 100


Number of nodes

Net

wor

k lif

etim

e (s)








BS

(a)

BS

(b)


BS

(a)

BS

(b)

BS

(c)




119894 119862119894119863119894 and119872




0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20


Sensor ID

Beha

vior

leve

l


0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20


Sensor ID

Beha

vior

leve

l








Acknowledgment


References



























































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




5

13

6

24

Cluster

Cluster head (CH)

Cluster member (CM)


Malicious node


(a)

5

5

4

4

31

1

4

4

4

6

2

Cluster

Initiallocation

Finallocation

Cluster head (CH)




(b)


where





the equation


lowast 119863119894

(6)

where1199081119908211990831199084 and119908



1199081+ 1199082+ 1199083+ 1199084+ 1199085= 1 (7)


119880119901119901119890119903and 119879ℎ119903119890119904ℎ

119871119900119908119890119903



12057512 (119906) = min (120575

12(119906119894) 119906119894isin 119880) (8)


12057512 (

V) = min (12057512(V119894) V119894isin 119880) (9)


AVG =sum119899

119894=112057512(119906119894)

119873

(10)


119879ℎ119903119890119904ℎ119880119901119901119890119903

=

1

2

(12057512 (119906) + AVG)

119879ℎ119903119890119904ℎ119871119900119908119890119903

=

1

2

(12057512 (

V) + AVG) (11)


119894119872119894 119863119894Er119894 and 119862


























U CH

ACCEPT_CH message

REQ_JOIN message

ADV_CH message


U

CH




Ids BL119894

Er119894

119862119894

119863119894

119872119894

119875119894

1 086 384212 3 115 120 7696324 081 483254 5 230 030 9681335 088 405325 3 130 055 8118296 085 462043 0 000 020 9243618 081 481680 4 105 140 96475310 095 365025 2 055 010 73080511 091 481960 1 070 220 964753




119871119900119908119890119903to those





119894




1 1199082 1199083 1199084 1199085




119894














119894rarr CH = 0)




119894



(30) 119899119894rarr CH = CH rarr Id


EndIfEnd For




Ids 1 4 5 6 8 10 111 769632 964753 9647534 968133 811829 9647535 968133 811829 7308056 9243618 769632 96475310 968133 811829 73080511 769632 964753


Inputs 119879ℎ119903119890119904ℎ119880119901119901119890119903

119879ℎ119903119890119904ℎ119871119900119908119890119903


119880119901119901119890119903)


(119873(CH))(4) 119869 = Count (119873(CH))

EndIf(5) For 119868 = 1 to 119869 Do(6) If (119899



119871119900119908119890119903)

Then(7) 119899



)Then




(13) 119899119894rarr CH rarr ID = ID


(15) Go to (2)EndIF


EndIfEnd For

End ForEnd


12055

7048

10095

2036

3045

5088

4081

8081

9050

1

086

11091

6

085






119880119901119901119890119903= 9 and 119879ℎ119903119890119904ℎ

119871119900119908119890119903= 6 For that


119871119900119908119890119903to



119880119901119901119890119903 Nodes 1 7 and


119871119900119908119890119903 so




Cluster 2

10

095

7

048

2

036

5

088

3

045

4

081

8

081

9

050

1

086

11091

6

085

12055

Cluster 4 Cluster 3

Cluster 1


Cluster 2

10

095

7

048

2

036

5

088

3

045

4

081

8

081

9

0501

086

11091

6

085

12055

Cluster 4

Cluster 5

Cluster 6

Cluster 3

Cluster 1


Cluster 210

095

7

048

2

036

5

088

3

045

4

081

8

081

9

0501

086

11091

6

085

12055

Cluster 4

Cluster 3

Cluster 1





will be repeated




Cluster 2

Cluster 1

BS

Local monitoring

Global monitoring









Yes

NoNo

ni is a normal node



d(S) gt Th




119894becomes the







[1199050 119905] 119873119887119901 119878119890119899119889(119899119894 Δ119905)


interval is Δ119905 = [1199050 1199050] 119873119887119901 119877119890119888119890119894V119890119889(119899

119894 Δ119905)


119863119890119897119886119910 119861119875 (119899119894 119905) = 119860119903119903119894V119886119897 119875119879

119894minus 119860119903119903119894V119886119897 119875119879

119894minus1 (12)


119864119888 (119899119894 Δ119905) = Er (119899

119894 1199050) minus Er (119899

119894 1199051) (13)


119894 1199050) is the


0 Er(119899

119894 1199051) is the


1and 119864119888(119899

119894 Δ119905) is



119894at time 119905 is






mechanism



(119868119889119873119887119901119878119890119899119889(119899119894Δ119905)

119873119887119901119877119890119888119890119894V119890119889(119899119894 Δ119905) 119863119890119897119886119910119861119875(119899119894 119905)

119864119888 (119899119894 Δ119905))

(14)


119894 119905119894minus1

) gt 120598)


old)Msg = (119868119889119873119887119901

119878119890119899119889(119899119894Δ119905) 119873119887119901

119877119890119888119890119894V119890119889(119899119894 Δ119905)119863119890119897119886119910

119861119875(119899119894 119905) 119864119888(119899

119894 Δ119905)) to its CHi for





119894is computed


BL119894= BL119894minus rate (15)





119878 = (1198781199051 1198781199052 1198781199053) (16)

where

(a) 1198781199051


defined as follows

1198781199051= sum119875119904

119877119890119888119890119894V119890119889 119887119910 119899119894 minussum119875119904119878119890119899119905 119887119910 119899119894

minussum119875119904119889119890119904119905119894119899119890119889 119887119910 119899119894

(17)

with

sum119875119904119877119890119888119890119894V119890119889 119887119910 119899119894 = sum119875119904119878119890119899119905119887119910 119899119894 +sum119875119904119889119890119904119905119894119899119890119889119887119910 119899119894

+sum119875119904119897119900119904119905 119887119910 119899119894

(18)


(b) 1198781199052


1198781199052= 119863119890119897119886119910 119861119875 (119899

119894 119905) (19)


1198781199053= 119864119888 (119899

119894 Δ119905) (20)

Here 119905 isin [1199050 t] = Δ119905


119878 =

sum119905119899minus1

119905=1199050119878119905

119899

(21)


119889 (119878) =

10038161003816100381610038161003816119878 minus 119878

10038161003816100381610038161003816 (22)







(23)







119894lt= 01))




(5) Mp(BLi) =





(6) If (BL119894le 03)Then

(7) If (119899119894is CM)Then


EndIf(10) If (119899



EndIfEndIf

EndIfEnd




119891119900119903 (119894119899119905 119899 = 0 119899 lt 119899119900119889119890 119905119900119887119890 119889119890119901119897119900119910119890119889 119899 + +)

119883 = rand() 119894119898119886119892119890 119865119894119890119897119889 119874119891 119862119900119897119897119890119888119905119894119899119892

rarr 119908119894119889119905ℎ119884 = rand() 119894119898119886119892119890 119865119894119890119897119889 119874119891 119862119900119897119897119890119888119905119894119899119892

rarr 119867119890119894119892ℎ119905



1= 03 119908

2=

02 1199083= 02 119908

4= 02 and 119908






0

5

10

15

20

25

30

35

40


Aver

age n

umbe

r of c

luste

rs

N = 200

N = 600

N = 1000


0

5

10

15

20

25

10 20 30 40 50 60 70 80 90 100

DWCAES-WCASDCA

Aver

age n

umbe

r of c

luste

rs

Number of nodes



119880119901119901119890119903and 119879ℎ119903119890119904ℎ

119871119900119908119890119903(reaffiliation

0

5

10

15

20

25

30


Aver

age n

umbe

r of c

luste

rs

WCA N = 20

ES-WCA N = 20

WCA N = 40

ES-WCA N = 40

WCA N = 60

ES-WCA N = 60





119880119901119901119890119903= 18 and 119879ℎ119903119890119904ℎ

119871119900119908119890119903= 9



0

05

1

15

2

25

3

35

4

10 20 30 40 50 60 70Number of nodes

Aver

age n

umbe

r of r

eaffi

liatio

ns

DWCAWCA

ES-WCA


05000

100001500020000250003000035000400004500050000

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

AODVDSDV

Nodes

Rem

aini

ng en

ergy




0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

20 40 60 80 100


Number of nodes

Net

wor

k lif

etim

e (s)








BS

(a)

BS

(b)


BS

(a)

BS

(b)

BS

(c)




119894 119862119894119863119894 and119872




0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20


Sensor ID

Beha

vior

leve

l


0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20


Sensor ID

Beha

vior

leve

l








Acknowledgment


References



























































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in



















U CH

ACCEPT_CH message

REQ_JOIN message

ADV_CH message


U

CH




Ids BL119894

Er119894

119862119894

119863119894

119872119894

119875119894

1 086 384212 3 115 120 7696324 081 483254 5 230 030 9681335 088 405325 3 130 055 8118296 085 462043 0 000 020 9243618 081 481680 4 105 140 96475310 095 365025 2 055 010 73080511 091 481960 1 070 220 964753




119871119900119908119890119903to those





119894




1 1199082 1199083 1199084 1199085




119894














119894rarr CH = 0)




119894



(30) 119899119894rarr CH = CH rarr Id


EndIfEnd For




Ids 1 4 5 6 8 10 111 769632 964753 9647534 968133 811829 9647535 968133 811829 7308056 9243618 769632 96475310 968133 811829 73080511 769632 964753


Inputs 119879ℎ119903119890119904ℎ119880119901119901119890119903

119879ℎ119903119890119904ℎ119871119900119908119890119903


119880119901119901119890119903)


(119873(CH))(4) 119869 = Count (119873(CH))

EndIf(5) For 119868 = 1 to 119869 Do(6) If (119899



119871119900119908119890119903)

Then(7) 119899



)Then




(13) 119899119894rarr CH rarr ID = ID


(15) Go to (2)EndIF


EndIfEnd For

End ForEnd


12055

7048

10095

2036

3045

5088

4081

8081

9050

1

086

11091

6

085






119880119901119901119890119903= 9 and 119879ℎ119903119890119904ℎ

119871119900119908119890119903= 6 For that


119871119900119908119890119903to



119880119901119901119890119903 Nodes 1 7 and


119871119900119908119890119903 so




Cluster 2

10

095

7

048

2

036

5

088

3

045

4

081

8

081

9

050

1

086

11091

6

085

12055

Cluster 4 Cluster 3

Cluster 1


Cluster 2

10

095

7

048

2

036

5

088

3

045

4

081

8

081

9

0501

086

11091

6

085

12055

Cluster 4

Cluster 5

Cluster 6

Cluster 3

Cluster 1


Cluster 210

095

7

048

2

036

5

088

3

045

4

081

8

081

9

0501

086

11091

6

085

12055

Cluster 4

Cluster 3

Cluster 1





will be repeated




Cluster 2

Cluster 1

BS

Local monitoring

Global monitoring









Yes

NoNo

ni is a normal node



d(S) gt Th




119894becomes the







[1199050 119905] 119873119887119901 119878119890119899119889(119899119894 Δ119905)


interval is Δ119905 = [1199050 1199050] 119873119887119901 119877119890119888119890119894V119890119889(119899

119894 Δ119905)


119863119890119897119886119910 119861119875 (119899119894 119905) = 119860119903119903119894V119886119897 119875119879

119894minus 119860119903119903119894V119886119897 119875119879

119894minus1 (12)


119864119888 (119899119894 Δ119905) = Er (119899

119894 1199050) minus Er (119899

119894 1199051) (13)


119894 1199050) is the


0 Er(119899

119894 1199051) is the


1and 119864119888(119899

119894 Δ119905) is



119894at time 119905 is






mechanism



(119868119889119873119887119901119878119890119899119889(119899119894Δ119905)

119873119887119901119877119890119888119890119894V119890119889(119899119894 Δ119905) 119863119890119897119886119910119861119875(119899119894 119905)

119864119888 (119899119894 Δ119905))

(14)


119894 119905119894minus1

) gt 120598)


old)Msg = (119868119889119873119887119901

119878119890119899119889(119899119894Δ119905) 119873119887119901

119877119890119888119890119894V119890119889(119899119894 Δ119905)119863119890119897119886119910

119861119875(119899119894 119905) 119864119888(119899

119894 Δ119905)) to its CHi for





119894is computed


BL119894= BL119894minus rate (15)





119878 = (1198781199051 1198781199052 1198781199053) (16)

where

(a) 1198781199051


defined as follows

1198781199051= sum119875119904

119877119890119888119890119894V119890119889 119887119910 119899119894 minussum119875119904119878119890119899119905 119887119910 119899119894

minussum119875119904119889119890119904119905119894119899119890119889 119887119910 119899119894

(17)

with

sum119875119904119877119890119888119890119894V119890119889 119887119910 119899119894 = sum119875119904119878119890119899119905119887119910 119899119894 +sum119875119904119889119890119904119905119894119899119890119889119887119910 119899119894

+sum119875119904119897119900119904119905 119887119910 119899119894

(18)


(b) 1198781199052


1198781199052= 119863119890119897119886119910 119861119875 (119899

119894 119905) (19)


1198781199053= 119864119888 (119899

119894 Δ119905) (20)

Here 119905 isin [1199050 t] = Δ119905


119878 =

sum119905119899minus1

119905=1199050119878119905

119899

(21)


119889 (119878) =

10038161003816100381610038161003816119878 minus 119878

10038161003816100381610038161003816 (22)







(23)







119894lt= 01))




(5) Mp(BLi) =





(6) If (BL119894le 03)Then

(7) If (119899119894is CM)Then


EndIf(10) If (119899



EndIfEndIf

EndIfEnd




119891119900119903 (119894119899119905 119899 = 0 119899 lt 119899119900119889119890 119905119900119887119890 119889119890119901119897119900119910119890119889 119899 + +)

119883 = rand() 119894119898119886119892119890 119865119894119890119897119889 119874119891 119862119900119897119897119890119888119905119894119899119892

rarr 119908119894119889119905ℎ119884 = rand() 119894119898119886119892119890 119865119894119890119897119889 119874119891 119862119900119897119897119890119888119905119894119899119892

rarr 119867119890119894119892ℎ119905



1= 03 119908

2=

02 1199083= 02 119908

4= 02 and 119908






0

5

10

15

20

25

30

35

40


Aver

age n

umbe

r of c

luste

rs

N = 200

N = 600

N = 1000


0

5

10

15

20

25

10 20 30 40 50 60 70 80 90 100

DWCAES-WCASDCA

Aver

age n

umbe

r of c

luste

rs

Number of nodes



119880119901119901119890119903and 119879ℎ119903119890119904ℎ

119871119900119908119890119903(reaffiliation

0

5

10

15

20

25

30


Aver

age n

umbe

r of c

luste

rs

WCA N = 20

ES-WCA N = 20

WCA N = 40

ES-WCA N = 40

WCA N = 60

ES-WCA N = 60





119880119901119901119890119903= 18 and 119879ℎ119903119890119904ℎ

119871119900119908119890119903= 9



0

05

1

15

2

25

3

35

4

10 20 30 40 50 60 70Number of nodes

Aver

age n

umbe

r of r

eaffi

liatio

ns

DWCAWCA

ES-WCA


05000

100001500020000250003000035000400004500050000

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

AODVDSDV

Nodes

Rem

aini

ng en

ergy




0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

20 40 60 80 100


Number of nodes

Net

wor

k lif

etim

e (s)








BS

(a)

BS

(b)


BS

(a)

BS

(b)

BS

(c)




119894 119862119894119863119894 and119872




0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20


Sensor ID

Beha

vior

leve

l


0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20


Sensor ID

Beha

vior

leve

l








Acknowledgment


References



























































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in





1 1199082 1199083 1199084 1199085




119894














119894rarr CH = 0)




119894



(30) 119899119894rarr CH = CH rarr Id


EndIfEnd For




Ids 1 4 5 6 8 10 111 769632 964753 9647534 968133 811829 9647535 968133 811829 7308056 9243618 769632 96475310 968133 811829 73080511 769632 964753


Inputs 119879ℎ119903119890119904ℎ119880119901119901119890119903

119879ℎ119903119890119904ℎ119871119900119908119890119903


119880119901119901119890119903)


(119873(CH))(4) 119869 = Count (119873(CH))

EndIf(5) For 119868 = 1 to 119869 Do(6) If (119899



119871119900119908119890119903)

Then(7) 119899



)Then




(13) 119899119894rarr CH rarr ID = ID


(15) Go to (2)EndIF


EndIfEnd For

End ForEnd


12055

7048

10095

2036

3045

5088

4081

8081

9050

1

086

11091

6

085






119880119901119901119890119903= 9 and 119879ℎ119903119890119904ℎ

119871119900119908119890119903= 6 For that


119871119900119908119890119903to



119880119901119901119890119903 Nodes 1 7 and


119871119900119908119890119903 so




Cluster 2

10

095

7

048

2

036

5

088

3

045

4

081

8

081

9

050

1

086

11091

6

085

12055

Cluster 4 Cluster 3

Cluster 1


Cluster 2

10

095

7

048

2

036

5

088

3

045

4

081

8

081

9

0501

086

11091

6

085

12055

Cluster 4

Cluster 5

Cluster 6

Cluster 3

Cluster 1


Cluster 210

095

7

048

2

036

5

088

3

045

4

081

8

081

9

0501

086

11091

6

085

12055

Cluster 4

Cluster 3

Cluster 1





will be repeated




Cluster 2

Cluster 1

BS

Local monitoring

Global monitoring









Yes

NoNo

ni is a normal node



d(S) gt Th




119894becomes the







[1199050 119905] 119873119887119901 119878119890119899119889(119899119894 Δ119905)


interval is Δ119905 = [1199050 1199050] 119873119887119901 119877119890119888119890119894V119890119889(119899

119894 Δ119905)


119863119890119897119886119910 119861119875 (119899119894 119905) = 119860119903119903119894V119886119897 119875119879

119894minus 119860119903119903119894V119886119897 119875119879

119894minus1 (12)


119864119888 (119899119894 Δ119905) = Er (119899

119894 1199050) minus Er (119899

119894 1199051) (13)


119894 1199050) is the


0 Er(119899

119894 1199051) is the


1and 119864119888(119899

119894 Δ119905) is



119894at time 119905 is






mechanism



(119868119889119873119887119901119878119890119899119889(119899119894Δ119905)

119873119887119901119877119890119888119890119894V119890119889(119899119894 Δ119905) 119863119890119897119886119910119861119875(119899119894 119905)

119864119888 (119899119894 Δ119905))

(14)


119894 119905119894minus1

) gt 120598)


old)Msg = (119868119889119873119887119901

119878119890119899119889(119899119894Δ119905) 119873119887119901

119877119890119888119890119894V119890119889(119899119894 Δ119905)119863119890119897119886119910

119861119875(119899119894 119905) 119864119888(119899

119894 Δ119905)) to its CHi for





119894is computed


BL119894= BL119894minus rate (15)





119878 = (1198781199051 1198781199052 1198781199053) (16)

where

(a) 1198781199051


defined as follows

1198781199051= sum119875119904

119877119890119888119890119894V119890119889 119887119910 119899119894 minussum119875119904119878119890119899119905 119887119910 119899119894

minussum119875119904119889119890119904119905119894119899119890119889 119887119910 119899119894

(17)

with

sum119875119904119877119890119888119890119894V119890119889 119887119910 119899119894 = sum119875119904119878119890119899119905119887119910 119899119894 +sum119875119904119889119890119904119905119894119899119890119889119887119910 119899119894

+sum119875119904119897119900119904119905 119887119910 119899119894

(18)


(b) 1198781199052


1198781199052= 119863119890119897119886119910 119861119875 (119899

119894 119905) (19)


1198781199053= 119864119888 (119899

119894 Δ119905) (20)

Here 119905 isin [1199050 t] = Δ119905


119878 =

sum119905119899minus1

119905=1199050119878119905

119899

(21)


119889 (119878) =

10038161003816100381610038161003816119878 minus 119878

10038161003816100381610038161003816 (22)







(23)







119894lt= 01))




(5) Mp(BLi) =





(6) If (BL119894le 03)Then

(7) If (119899119894is CM)Then


EndIf(10) If (119899



EndIfEndIf

EndIfEnd




119891119900119903 (119894119899119905 119899 = 0 119899 lt 119899119900119889119890 119905119900119887119890 119889119890119901119897119900119910119890119889 119899 + +)

119883 = rand() 119894119898119886119892119890 119865119894119890119897119889 119874119891 119862119900119897119897119890119888119905119894119899119892

rarr 119908119894119889119905ℎ119884 = rand() 119894119898119886119892119890 119865119894119890119897119889 119874119891 119862119900119897119897119890119888119905119894119899119892

rarr 119867119890119894119892ℎ119905



1= 03 119908

2=

02 1199083= 02 119908

4= 02 and 119908






0

5

10

15

20

25

30

35

40


Aver

age n

umbe

r of c

luste

rs

N = 200

N = 600

N = 1000


0

5

10

15

20

25

10 20 30 40 50 60 70 80 90 100

DWCAES-WCASDCA

Aver

age n

umbe

r of c

luste

rs

Number of nodes



119880119901119901119890119903and 119879ℎ119903119890119904ℎ

119871119900119908119890119903(reaffiliation

0

5

10

15

20

25

30


Aver

age n

umbe

r of c

luste

rs

WCA N = 20

ES-WCA N = 20

WCA N = 40

ES-WCA N = 40

WCA N = 60

ES-WCA N = 60





119880119901119901119890119903= 18 and 119879ℎ119903119890119904ℎ

119871119900119908119890119903= 9



0

05

1

15

2

25

3

35

4

10 20 30 40 50 60 70Number of nodes

Aver

age n

umbe

r of r

eaffi

liatio

ns

DWCAWCA

ES-WCA


05000

100001500020000250003000035000400004500050000

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

AODVDSDV

Nodes

Rem

aini

ng en

ergy




0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

20 40 60 80 100


Number of nodes

Net

wor

k lif

etim

e (s)








BS

(a)

BS

(b)


BS

(a)

BS

(b)

BS

(c)




119894 119862119894119863119894 and119872




0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20


Sensor ID

Beha

vior

leve

l


0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20


Sensor ID

Beha

vior

leve

l








Acknowledgment


References



























































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




Inputs 119879ℎ119903119890119904ℎ119880119901119901119890119903

119879ℎ119903119890119904ℎ119871119900119908119890119903


119880119901119901119890119903)


(119873(CH))(4) 119869 = Count (119873(CH))

EndIf(5) For 119868 = 1 to 119869 Do(6) If (119899



119871119900119908119890119903)

Then(7) 119899



)Then




(13) 119899119894rarr CH rarr ID = ID


(15) Go to (2)EndIF


EndIfEnd For

End ForEnd


12055

7048

10095

2036

3045

5088

4081

8081

9050

1

086

11091

6

085






119880119901119901119890119903= 9 and 119879ℎ119903119890119904ℎ

119871119900119908119890119903= 6 For that


119871119900119908119890119903to



119880119901119901119890119903 Nodes 1 7 and


119871119900119908119890119903 so




Cluster 2

10

095

7

048

2

036

5

088

3

045

4

081

8

081

9

050

1

086

11091

6

085

12055

Cluster 4 Cluster 3

Cluster 1


Cluster 2

10

095

7

048

2

036

5

088

3

045

4

081

8

081

9

0501

086

11091

6

085

12055

Cluster 4

Cluster 5

Cluster 6

Cluster 3

Cluster 1


Cluster 210

095

7

048

2

036

5

088

3

045

4

081

8

081

9

0501

086

11091

6

085

12055

Cluster 4

Cluster 3

Cluster 1





will be repeated




Cluster 2

Cluster 1

BS

Local monitoring

Global monitoring









Yes

NoNo

ni is a normal node



d(S) gt Th




119894becomes the







[1199050 119905] 119873119887119901 119878119890119899119889(119899119894 Δ119905)


interval is Δ119905 = [1199050 1199050] 119873119887119901 119877119890119888119890119894V119890119889(119899

119894 Δ119905)


119863119890119897119886119910 119861119875 (119899119894 119905) = 119860119903119903119894V119886119897 119875119879

119894minus 119860119903119903119894V119886119897 119875119879

119894minus1 (12)


119864119888 (119899119894 Δ119905) = Er (119899

119894 1199050) minus Er (119899

119894 1199051) (13)


119894 1199050) is the


0 Er(119899

119894 1199051) is the


1and 119864119888(119899

119894 Δ119905) is



119894at time 119905 is






mechanism



(119868119889119873119887119901119878119890119899119889(119899119894Δ119905)

119873119887119901119877119890119888119890119894V119890119889(119899119894 Δ119905) 119863119890119897119886119910119861119875(119899119894 119905)

119864119888 (119899119894 Δ119905))

(14)


119894 119905119894minus1

) gt 120598)


old)Msg = (119868119889119873119887119901

119878119890119899119889(119899119894Δ119905) 119873119887119901

119877119890119888119890119894V119890119889(119899119894 Δ119905)119863119890119897119886119910

119861119875(119899119894 119905) 119864119888(119899

119894 Δ119905)) to its CHi for





119894is computed


BL119894= BL119894minus rate (15)





119878 = (1198781199051 1198781199052 1198781199053) (16)

where

(a) 1198781199051


defined as follows

1198781199051= sum119875119904

119877119890119888119890119894V119890119889 119887119910 119899119894 minussum119875119904119878119890119899119905 119887119910 119899119894

minussum119875119904119889119890119904119905119894119899119890119889 119887119910 119899119894

(17)

with

sum119875119904119877119890119888119890119894V119890119889 119887119910 119899119894 = sum119875119904119878119890119899119905119887119910 119899119894 +sum119875119904119889119890119904119905119894119899119890119889119887119910 119899119894

+sum119875119904119897119900119904119905 119887119910 119899119894

(18)


(b) 1198781199052


1198781199052= 119863119890119897119886119910 119861119875 (119899

119894 119905) (19)


1198781199053= 119864119888 (119899

119894 Δ119905) (20)

Here 119905 isin [1199050 t] = Δ119905


119878 =

sum119905119899minus1

119905=1199050119878119905

119899

(21)


119889 (119878) =

10038161003816100381610038161003816119878 minus 119878

10038161003816100381610038161003816 (22)







(23)







119894lt= 01))




(5) Mp(BLi) =





(6) If (BL119894le 03)Then

(7) If (119899119894is CM)Then


EndIf(10) If (119899



EndIfEndIf

EndIfEnd




119891119900119903 (119894119899119905 119899 = 0 119899 lt 119899119900119889119890 119905119900119887119890 119889119890119901119897119900119910119890119889 119899 + +)

119883 = rand() 119894119898119886119892119890 119865119894119890119897119889 119874119891 119862119900119897119897119890119888119905119894119899119892

rarr 119908119894119889119905ℎ119884 = rand() 119894119898119886119892119890 119865119894119890119897119889 119874119891 119862119900119897119897119890119888119905119894119899119892

rarr 119867119890119894119892ℎ119905



1= 03 119908

2=

02 1199083= 02 119908

4= 02 and 119908






0

5

10

15

20

25

30

35

40


Aver

age n

umbe

r of c

luste

rs

N = 200

N = 600

N = 1000


0

5

10

15

20

25

10 20 30 40 50 60 70 80 90 100

DWCAES-WCASDCA

Aver

age n

umbe

r of c

luste

rs

Number of nodes



119880119901119901119890119903and 119879ℎ119903119890119904ℎ

119871119900119908119890119903(reaffiliation

0

5

10

15

20

25

30


Aver

age n

umbe

r of c

luste

rs

WCA N = 20

ES-WCA N = 20

WCA N = 40

ES-WCA N = 40

WCA N = 60

ES-WCA N = 60





119880119901119901119890119903= 18 and 119879ℎ119903119890119904ℎ

119871119900119908119890119903= 9



0

05

1

15

2

25

3

35

4

10 20 30 40 50 60 70Number of nodes

Aver

age n

umbe

r of r

eaffi

liatio

ns

DWCAWCA

ES-WCA


05000

100001500020000250003000035000400004500050000

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

AODVDSDV

Nodes

Rem

aini

ng en

ergy




0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

20 40 60 80 100


Number of nodes

Net

wor

k lif

etim

e (s)








BS

(a)

BS

(b)


BS

(a)

BS

(b)

BS

(c)




119894 119862119894119863119894 and119872




0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20


Sensor ID

Beha

vior

leve

l


0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20


Sensor ID

Beha

vior

leve

l








Acknowledgment


References



























































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




Cluster 2

10

095

7

048

2

036

5

088

3

045

4

081

8

081

9

050

1

086

11091

6

085

12055

Cluster 4 Cluster 3

Cluster 1


Cluster 2

10

095

7

048

2

036

5

088

3

045

4

081

8

081

9

0501

086

11091

6

085

12055

Cluster 4

Cluster 5

Cluster 6

Cluster 3

Cluster 1


Cluster 210

095

7

048

2

036

5

088

3

045

4

081

8

081

9

0501

086

11091

6

085

12055

Cluster 4

Cluster 3

Cluster 1





will be repeated




Cluster 2

Cluster 1

BS

Local monitoring

Global monitoring









Yes

NoNo

ni is a normal node



d(S) gt Th




119894becomes the







[1199050 119905] 119873119887119901 119878119890119899119889(119899119894 Δ119905)


interval is Δ119905 = [1199050 1199050] 119873119887119901 119877119890119888119890119894V119890119889(119899

119894 Δ119905)


119863119890119897119886119910 119861119875 (119899119894 119905) = 119860119903119903119894V119886119897 119875119879

119894minus 119860119903119903119894V119886119897 119875119879

119894minus1 (12)


119864119888 (119899119894 Δ119905) = Er (119899

119894 1199050) minus Er (119899

119894 1199051) (13)


119894 1199050) is the


0 Er(119899

119894 1199051) is the


1and 119864119888(119899

119894 Δ119905) is



119894at time 119905 is






mechanism



(119868119889119873119887119901119878119890119899119889(119899119894Δ119905)

119873119887119901119877119890119888119890119894V119890119889(119899119894 Δ119905) 119863119890119897119886119910119861119875(119899119894 119905)

119864119888 (119899119894 Δ119905))

(14)


119894 119905119894minus1

) gt 120598)


old)Msg = (119868119889119873119887119901

119878119890119899119889(119899119894Δ119905) 119873119887119901

119877119890119888119890119894V119890119889(119899119894 Δ119905)119863119890119897119886119910

119861119875(119899119894 119905) 119864119888(119899

119894 Δ119905)) to its CHi for





119894is computed


BL119894= BL119894minus rate (15)





119878 = (1198781199051 1198781199052 1198781199053) (16)

where

(a) 1198781199051


defined as follows

1198781199051= sum119875119904

119877119890119888119890119894V119890119889 119887119910 119899119894 minussum119875119904119878119890119899119905 119887119910 119899119894

minussum119875119904119889119890119904119905119894119899119890119889 119887119910 119899119894

(17)

with

sum119875119904119877119890119888119890119894V119890119889 119887119910 119899119894 = sum119875119904119878119890119899119905119887119910 119899119894 +sum119875119904119889119890119904119905119894119899119890119889119887119910 119899119894

+sum119875119904119897119900119904119905 119887119910 119899119894

(18)


(b) 1198781199052


1198781199052= 119863119890119897119886119910 119861119875 (119899

119894 119905) (19)


1198781199053= 119864119888 (119899

119894 Δ119905) (20)

Here 119905 isin [1199050 t] = Δ119905


119878 =

sum119905119899minus1

119905=1199050119878119905

119899

(21)


119889 (119878) =

10038161003816100381610038161003816119878 minus 119878

10038161003816100381610038161003816 (22)







(23)







119894lt= 01))




(5) Mp(BLi) =





(6) If (BL119894le 03)Then

(7) If (119899119894is CM)Then


EndIf(10) If (119899



EndIfEndIf

EndIfEnd




119891119900119903 (119894119899119905 119899 = 0 119899 lt 119899119900119889119890 119905119900119887119890 119889119890119901119897119900119910119890119889 119899 + +)

119883 = rand() 119894119898119886119892119890 119865119894119890119897119889 119874119891 119862119900119897119897119890119888119905119894119899119892

rarr 119908119894119889119905ℎ119884 = rand() 119894119898119886119892119890 119865119894119890119897119889 119874119891 119862119900119897119897119890119888119905119894119899119892

rarr 119867119890119894119892ℎ119905



1= 03 119908

2=

02 1199083= 02 119908

4= 02 and 119908






0

5

10

15

20

25

30

35

40


Aver

age n

umbe

r of c

luste

rs

N = 200

N = 600

N = 1000


0

5

10

15

20

25

10 20 30 40 50 60 70 80 90 100

DWCAES-WCASDCA

Aver

age n

umbe

r of c

luste

rs

Number of nodes



119880119901119901119890119903and 119879ℎ119903119890119904ℎ

119871119900119908119890119903(reaffiliation

0

5

10

15

20

25

30


Aver

age n

umbe

r of c

luste

rs

WCA N = 20

ES-WCA N = 20

WCA N = 40

ES-WCA N = 40

WCA N = 60

ES-WCA N = 60





119880119901119901119890119903= 18 and 119879ℎ119903119890119904ℎ

119871119900119908119890119903= 9



0

05

1

15

2

25

3

35

4

10 20 30 40 50 60 70Number of nodes

Aver

age n

umbe

r of r

eaffi

liatio

ns

DWCAWCA

ES-WCA


05000

100001500020000250003000035000400004500050000

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

AODVDSDV

Nodes

Rem

aini

ng en

ergy




0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

20 40 60 80 100


Number of nodes

Net

wor

k lif

etim

e (s)








BS

(a)

BS

(b)


BS

(a)

BS

(b)

BS

(c)




119894 119862119894119863119894 and119872




0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20


Sensor ID

Beha

vior

leve

l


0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20


Sensor ID

Beha

vior

leve

l








Acknowledgment


References



























































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




Cluster 2

Cluster 1

BS

Local monitoring

Global monitoring









Yes

NoNo

ni is a normal node



d(S) gt Th




119894becomes the







[1199050 119905] 119873119887119901 119878119890119899119889(119899119894 Δ119905)


interval is Δ119905 = [1199050 1199050] 119873119887119901 119877119890119888119890119894V119890119889(119899

119894 Δ119905)


119863119890119897119886119910 119861119875 (119899119894 119905) = 119860119903119903119894V119886119897 119875119879

119894minus 119860119903119903119894V119886119897 119875119879

119894minus1 (12)


119864119888 (119899119894 Δ119905) = Er (119899

119894 1199050) minus Er (119899

119894 1199051) (13)


119894 1199050) is the


0 Er(119899

119894 1199051) is the


1and 119864119888(119899

119894 Δ119905) is



119894at time 119905 is






mechanism



(119868119889119873119887119901119878119890119899119889(119899119894Δ119905)

119873119887119901119877119890119888119890119894V119890119889(119899119894 Δ119905) 119863119890119897119886119910119861119875(119899119894 119905)

119864119888 (119899119894 Δ119905))

(14)


119894 119905119894minus1

) gt 120598)


old)Msg = (119868119889119873119887119901

119878119890119899119889(119899119894Δ119905) 119873119887119901

119877119890119888119890119894V119890119889(119899119894 Δ119905)119863119890119897119886119910

119861119875(119899119894 119905) 119864119888(119899

119894 Δ119905)) to its CHi for





119894is computed


BL119894= BL119894minus rate (15)





119878 = (1198781199051 1198781199052 1198781199053) (16)

where

(a) 1198781199051


defined as follows

1198781199051= sum119875119904

119877119890119888119890119894V119890119889 119887119910 119899119894 minussum119875119904119878119890119899119905 119887119910 119899119894

minussum119875119904119889119890119904119905119894119899119890119889 119887119910 119899119894

(17)

with

sum119875119904119877119890119888119890119894V119890119889 119887119910 119899119894 = sum119875119904119878119890119899119905119887119910 119899119894 +sum119875119904119889119890119904119905119894119899119890119889119887119910 119899119894

+sum119875119904119897119900119904119905 119887119910 119899119894

(18)


(b) 1198781199052


1198781199052= 119863119890119897119886119910 119861119875 (119899

119894 119905) (19)


1198781199053= 119864119888 (119899

119894 Δ119905) (20)

Here 119905 isin [1199050 t] = Δ119905


119878 =

sum119905119899minus1

119905=1199050119878119905

119899

(21)


119889 (119878) =

10038161003816100381610038161003816119878 minus 119878

10038161003816100381610038161003816 (22)







(23)







119894lt= 01))




(5) Mp(BLi) =





(6) If (BL119894le 03)Then

(7) If (119899119894is CM)Then


EndIf(10) If (119899



EndIfEndIf

EndIfEnd




119891119900119903 (119894119899119905 119899 = 0 119899 lt 119899119900119889119890 119905119900119887119890 119889119890119901119897119900119910119890119889 119899 + +)

119883 = rand() 119894119898119886119892119890 119865119894119890119897119889 119874119891 119862119900119897119897119890119888119905119894119899119892

rarr 119908119894119889119905ℎ119884 = rand() 119894119898119886119892119890 119865119894119890119897119889 119874119891 119862119900119897119897119890119888119905119894119899119892

rarr 119867119890119894119892ℎ119905



1= 03 119908

2=

02 1199083= 02 119908

4= 02 and 119908






0

5

10

15

20

25

30

35

40


Aver

age n

umbe

r of c

luste

rs

N = 200

N = 600

N = 1000


0

5

10

15

20

25

10 20 30 40 50 60 70 80 90 100

DWCAES-WCASDCA

Aver

age n

umbe

r of c

luste

rs

Number of nodes



119880119901119901119890119903and 119879ℎ119903119890119904ℎ

119871119900119908119890119903(reaffiliation

0

5

10

15

20

25

30


Aver

age n

umbe

r of c

luste

rs

WCA N = 20

ES-WCA N = 20

WCA N = 40

ES-WCA N = 40

WCA N = 60

ES-WCA N = 60





119880119901119901119890119903= 18 and 119879ℎ119903119890119904ℎ

119871119900119908119890119903= 9



0

05

1

15

2

25

3

35

4

10 20 30 40 50 60 70Number of nodes

Aver

age n

umbe

r of r

eaffi

liatio

ns

DWCAWCA

ES-WCA


05000

100001500020000250003000035000400004500050000

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

AODVDSDV

Nodes

Rem

aini

ng en

ergy




0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

20 40 60 80 100


Number of nodes

Net

wor

k lif

etim

e (s)








BS

(a)

BS

(b)


BS

(a)

BS

(b)

BS

(c)




119894 119862119894119863119894 and119872




0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20


Sensor ID

Beha

vior

leve

l


0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20


Sensor ID

Beha

vior

leve

l








Acknowledgment


References



























































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in





(119868119889119873119887119901119878119890119899119889(119899119894Δ119905)

119873119887119901119877119890119888119890119894V119890119889(119899119894 Δ119905) 119863119890119897119886119910119861119875(119899119894 119905)

119864119888 (119899119894 Δ119905))

(14)


119894 119905119894minus1

) gt 120598)


old)Msg = (119868119889119873119887119901

119878119890119899119889(119899119894Δ119905) 119873119887119901

119877119890119888119890119894V119890119889(119899119894 Δ119905)119863119890119897119886119910

119861119875(119899119894 119905) 119864119888(119899

119894 Δ119905)) to its CHi for





119894is computed


BL119894= BL119894minus rate (15)





119878 = (1198781199051 1198781199052 1198781199053) (16)

where

(a) 1198781199051


defined as follows

1198781199051= sum119875119904

119877119890119888119890119894V119890119889 119887119910 119899119894 minussum119875119904119878119890119899119905 119887119910 119899119894

minussum119875119904119889119890119904119905119894119899119890119889 119887119910 119899119894

(17)

with

sum119875119904119877119890119888119890119894V119890119889 119887119910 119899119894 = sum119875119904119878119890119899119905119887119910 119899119894 +sum119875119904119889119890119904119905119894119899119890119889119887119910 119899119894

+sum119875119904119897119900119904119905 119887119910 119899119894

(18)


(b) 1198781199052


1198781199052= 119863119890119897119886119910 119861119875 (119899

119894 119905) (19)


1198781199053= 119864119888 (119899

119894 Δ119905) (20)

Here 119905 isin [1199050 t] = Δ119905


119878 =

sum119905119899minus1

119905=1199050119878119905

119899

(21)


119889 (119878) =

10038161003816100381610038161003816119878 minus 119878

10038161003816100381610038161003816 (22)







(23)







119894lt= 01))




(5) Mp(BLi) =





(6) If (BL119894le 03)Then

(7) If (119899119894is CM)Then


EndIf(10) If (119899



EndIfEndIf

EndIfEnd




119891119900119903 (119894119899119905 119899 = 0 119899 lt 119899119900119889119890 119905119900119887119890 119889119890119901119897119900119910119890119889 119899 + +)

119883 = rand() 119894119898119886119892119890 119865119894119890119897119889 119874119891 119862119900119897119897119890119888119905119894119899119892

rarr 119908119894119889119905ℎ119884 = rand() 119894119898119886119892119890 119865119894119890119897119889 119874119891 119862119900119897119897119890119888119905119894119899119892

rarr 119867119890119894119892ℎ119905



1= 03 119908

2=

02 1199083= 02 119908

4= 02 and 119908






0

5

10

15

20

25

30

35

40


Aver

age n

umbe

r of c

luste

rs

N = 200

N = 600

N = 1000


0

5

10

15

20

25

10 20 30 40 50 60 70 80 90 100

DWCAES-WCASDCA

Aver

age n

umbe

r of c

luste

rs

Number of nodes



119880119901119901119890119903and 119879ℎ119903119890119904ℎ

119871119900119908119890119903(reaffiliation

0

5

10

15

20

25

30


Aver

age n

umbe

r of c

luste

rs

WCA N = 20

ES-WCA N = 20

WCA N = 40

ES-WCA N = 40

WCA N = 60

ES-WCA N = 60





119880119901119901119890119903= 18 and 119879ℎ119903119890119904ℎ

119871119900119908119890119903= 9



0

05

1

15

2

25

3

35

4

10 20 30 40 50 60 70Number of nodes

Aver

age n

umbe

r of r

eaffi

liatio

ns

DWCAWCA

ES-WCA


05000

100001500020000250003000035000400004500050000

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

AODVDSDV

Nodes

Rem

aini

ng en

ergy




0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

20 40 60 80 100


Number of nodes

Net

wor

k lif

etim

e (s)








BS

(a)

BS

(b)


BS

(a)

BS

(b)

BS

(c)




119894 119862119894119863119894 and119872




0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20


Sensor ID

Beha

vior

leve

l


0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20


Sensor ID

Beha

vior

leve

l








Acknowledgment


References



























































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in





119894lt= 01))




(5) Mp(BLi) =





(6) If (BL119894le 03)Then

(7) If (119899119894is CM)Then


EndIf(10) If (119899



EndIfEndIf

EndIfEnd




119891119900119903 (119894119899119905 119899 = 0 119899 lt 119899119900119889119890 119905119900119887119890 119889119890119901119897119900119910119890119889 119899 + +)

119883 = rand() 119894119898119886119892119890 119865119894119890119897119889 119874119891 119862119900119897119897119890119888119905119894119899119892

rarr 119908119894119889119905ℎ119884 = rand() 119894119898119886119892119890 119865119894119890119897119889 119874119891 119862119900119897119897119890119888119905119894119899119892

rarr 119867119890119894119892ℎ119905



1= 03 119908

2=

02 1199083= 02 119908

4= 02 and 119908






0

5

10

15

20

25

30

35

40


Aver

age n

umbe

r of c

luste

rs

N = 200

N = 600

N = 1000


0

5

10

15

20

25

10 20 30 40 50 60 70 80 90 100

DWCAES-WCASDCA

Aver

age n

umbe

r of c

luste

rs

Number of nodes



119880119901119901119890119903and 119879ℎ119903119890119904ℎ

119871119900119908119890119903(reaffiliation

0

5

10

15

20

25

30


Aver

age n

umbe

r of c

luste

rs

WCA N = 20

ES-WCA N = 20

WCA N = 40

ES-WCA N = 40

WCA N = 60

ES-WCA N = 60





119880119901119901119890119903= 18 and 119879ℎ119903119890119904ℎ

119871119900119908119890119903= 9



0

05

1

15

2

25

3

35

4

10 20 30 40 50 60 70Number of nodes

Aver

age n

umbe

r of r

eaffi

liatio

ns

DWCAWCA

ES-WCA


05000

100001500020000250003000035000400004500050000

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

AODVDSDV

Nodes

Rem

aini

ng en

ergy




0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

20 40 60 80 100


Number of nodes

Net

wor

k lif

etim

e (s)








BS

(a)

BS

(b)


BS

(a)

BS

(b)

BS

(c)




119894 119862119894119863119894 and119872




0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20


Sensor ID

Beha

vior

leve

l


0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20


Sensor ID

Beha

vior

leve

l








Acknowledgment


References



























































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




0

5

10

15

20

25

30

35

40


Aver

age n

umbe

r of c

luste

rs

N = 200

N = 600

N = 1000


0

5

10

15

20

25

10 20 30 40 50 60 70 80 90 100

DWCAES-WCASDCA

Aver

age n

umbe

r of c

luste

rs

Number of nodes



119880119901119901119890119903and 119879ℎ119903119890119904ℎ

119871119900119908119890119903(reaffiliation

0

5

10

15

20

25

30


Aver

age n

umbe

r of c

luste

rs

WCA N = 20

ES-WCA N = 20

WCA N = 40

ES-WCA N = 40

WCA N = 60

ES-WCA N = 60





119880119901119901119890119903= 18 and 119879ℎ119903119890119904ℎ

119871119900119908119890119903= 9



0

05

1

15

2

25

3

35

4

10 20 30 40 50 60 70Number of nodes

Aver

age n

umbe

r of r

eaffi

liatio

ns

DWCAWCA

ES-WCA


05000

100001500020000250003000035000400004500050000

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

AODVDSDV

Nodes

Rem

aini

ng en

ergy




0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

20 40 60 80 100


Number of nodes

Net

wor

k lif

etim

e (s)








BS

(a)

BS

(b)


BS

(a)

BS

(b)

BS

(c)




119894 119862119894119863119894 and119872




0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20


Sensor ID

Beha

vior

leve

l


0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20


Sensor ID

Beha

vior

leve

l








Acknowledgment


References



























































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




0

05

1

15

2

25

3

35

4

10 20 30 40 50 60 70Number of nodes

Aver

age n

umbe

r of r

eaffi

liatio

ns

DWCAWCA

ES-WCA


05000

100001500020000250003000035000400004500050000

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

AODVDSDV

Nodes

Rem

aini

ng en

ergy




0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

20 40 60 80 100


Number of nodes

Net

wor

k lif

etim

e (s)








BS

(a)

BS

(b)


BS

(a)

BS

(b)

BS

(c)




119894 119862119894119863119894 and119872




0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20


Sensor ID

Beha

vior

leve

l


0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20


Sensor ID

Beha

vior

leve

l








Acknowledgment


References



























































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




BS

(a)

BS

(b)


BS

(a)

BS

(b)

BS

(c)




119894 119862119894119863119894 and119872




0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20


Sensor ID

Beha

vior

leve

l


0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20


Sensor ID

Beha

vior

leve

l








Acknowledgment


References



























































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20


Sensor ID

Beha

vior

leve

l


0

02

04

06

08

1

12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20


Sensor ID

Beha

vior

leve

l








Acknowledgment


References



























































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in





















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in

















Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in










Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in



Date post:	14-Jul-2020
Category:	Documents
Upload:	others
View:	0 times
Download:	0 times

Research Article Energy Efficient and Safe Weighted...

Documents