research article esmart: energy-efficient slice-mix...

Hindawi Publishing CorporationInternational Journal of Distributed Sensor NetworksVolume 2013 Article ID 134509 9 pageshttpdxdoiorg1011552013134509

Research ArticleESMART Energy-Efficient Slice-Mix-Aggregate forWireless Sensor Network

Chaoran Li and Yun Liu

Key Laboratory of Communication amp Information Systems Beijing Municipal Commission of EducationBeijing Jiaotong University Beijing 100044 China

Correspondence should be addressed to Yun Liu liuyunbjtueducn

Received 30 July 2013 Accepted 19 November 2013

Academic Editor Ying-Hong Wang

Copyright copy 2013 C Li and Y Liu This 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

Wireless sensor network consists of a large number of resource-constrained sensor nodes and is usually deployed in unattendedarea to collect specific information Energy consumption is always a major concern in the research field of wireless sensor networkThus data aggregation schemes emerged and were deployed for prolonging network lifetime by reducing data transmitted withinthe network Meanwhile along with the wide application of the data aggregation schemes the security issues of data aggregationhave been increasingly drawing attention and the designing of secure data aggregation scheme is becoming a hot spot In thispaper we proposed an energy-efficient secure data aggregation scheme ESMART energy-efficient slice-mix-aggregate based onthe technique of data slicing and mixing The proposed scheme performs secure data aggregation in a more efficient way whilekeeping a good performance of privacy preservation And the simulation result shows that the security performance of ESMARTscheme is better than that of some existing and widely used schemes

1 Introduction

A wireless sensor network is composed of a large number ofresource-constrained sensor nodes The base station of thenetwork sends queries to the child nodes and the sensornodes that received the queries will collect the informationfollowing the request of the query The data acquired bysensor nodes will be transmitted to the base station withmultihop transmission scheme Obviously the energy con-sumption of transmissionwill affect the life cycle of the sensornodes dramatically Hence the technology of data aggrega-tion has been widely used in wireless sensor network as aneffective mechanism to reduce the amount of data transferin the network In the mechanism of data aggregation thenodes which aggregate raw data from their neighbor nodesand process it by specific algorithms are called aggregatorsand the other sensor nodes which center on the aggregatorsand only perform information collecting and transmittingare called leaf nodes The processed and aggregated datawill be returned to the base station through the aggregationtree constructed dynamically After the application of dataaggregation technology the amount of data transfer inWSNs

has been reduced greatly and then prolonged the operatinglife of the network

Meanwhile some security problems have arisen becauseof the deployment of data aggregation scheme Asmentionedabove sensor nodes are usually resource constrained thecomputing and communicating capacity of the nodes are verylimited which makes the designing of the secure scheme inWSN more difficult than in traditional wired network andmost of the existing mature secure protocols could not beapplied to WSN directly From the perspective of the adver-sary the sensor nodes are usually deployed in unattendedeven hostile environment which makes sensor nodes veryvulnerable to the attacks We take a simple and widely useddata aggregation scheme TAG Tiny Aggregation proposedby Madden et al in [1] as an example We can see thatthere is no protection of data privacy in this scheme whichmeans a captured node in the network can eavesdrop on anddecrypt the raw readings of neighbor nodes easily withoutbeing detected and then lead to a serious security problemwhich is not acceptable [2 3] On the basis of the aboveanalysis a good secure data aggregation scheme for WSNshould be energy efficient while ensuring data security In this

2 International Journal of Distributed Sensor Networks

paper we propose a novel secure data aggregation schemefor WSN ESMART (energy-efficient slice-mix-aggregate)which addresses both energy-efficient and privacy preserva-tion of the network To achieve the privacy preservation of thenetwork we introduced data slicing and mixing techniqueinto the scheme designing A secure data aggregation schemebased on data slicing and mixing technique was detailedin [4] in this scheme in order to achieve the preservationof data privacy in the network the raw readings of thesensor nodes will be sliced into pieces and part of the pieceswill be encrypted and sent to the neighbor nodes But thiscomes at the price of some additional communication costwhich is shown in the simulation result of SMART schemethe communication cost of SMART is about 119869 times thecommunication cost of TAG where 119869 denotes the number ofdata pieces Thus we introduced the dynamic 119869 value into thescheme which can effectively decrease the communicationoverhead while maintaining the privacy preservation ofthe network Meanwhile a lower communication overheadmeans a lower transmission collision probability which leadsto a higher aggregation accuracy in the networkMoreover inESMART the times of data slicing of a sensor node are onlyknown by itself so even if a malicious node eavesdroppedon several data pieces from the node it is not sure that allthe data pieces of the node are eavesdropped on Thus it ismore difficult for the malicious nodes to steal private dataGiven the above the use of data slicing andmixing techniquein ESMART is more efficient which consequently reducesthe communication overhead and increases the aggregationaccuracy effectively

Wireless sensor network is usually deployed in unat-tended environment and consisted of a large number oflow-powered resource-constrained sensor nodes Each nodein the network can sense the environment information asrequested by the base station and perform wireless commu-nication within a small range of its location To reduce theenergy consumption of the nodes during the transmissiondata aggregation scheme has been widely used in WSNMadden et al proposed TAG scheme as a simple dataaggregation scheme based on the assumption that every nodein the network is trusted and there is no protection of dataprivacy in this scheme But in practice the application ofWSN is facing various kinds of security threats A goodsecure data aggregation scheme should possess the capabilityto resist the security threats while staying within energyconstraint To meet this requirement some secure dataaggregation schemes have been provided He et al proposed anovel data aggregation scheme with disjoint aggregation treestructure to ensure data integrity in [5] Liu et al proposed asecure data aggregation schemeHEEPwith a novel formationmethod of aggregation tree in [6] in 2013 Li et al proposeda secure data aggregation scheme based on the improvedSMART scheme which is more efficient [7] Both Li et al[8] and Bista and Chang [9] proposed survey papers of theprivacy-preserving techniques for WSN Meanwhile someresearchers analyzed the existing network attack technologiesagainst WSN and proposed corresponding solutions [10ndash12] Zhu et al proposed a secure data aggregation schemein which the base station composes a secret configuration

matrix and each sensor node is preloaded with a limitedpart of the matrix known as a secret share containing certainlocal instructions [13] Dong and Li presented a secure dataaggregation approach based on monitoring in WSN [14] Inthis paper a grid-based network architecture for monitoringin WSN is designed and the algorithms for network divisioninitialization and grid tree construction are proposed Leiet al proposed a secure data aggregation solution based ondynamic multiple cluster key management model in [15] thiskey management model consumes little storage space andcan resist the collusion attack effectively Sun et al proposeda lightweight secure data aggregation protocol to find thecompromised nodes and help the base station to verify thefinal results [16] Poornima and Amberker proposed a securedata aggregation scheme in [17] This scheme provides end-to-end data privacy and can effectively reduce the energyconsumption

2 Network Model andthe Requirement of Design

In this paper we used aggregation tree to organize the sensornodes in the network and consider three types of nodes inthe network base station aggregator node and leaf nodeThere is only one base station in the network which can beseen as the network control center with unlimited resourcesand the final destination for the aggregation results As theroot of the aggregation tree it is responsible for broadcastingqueries to all the other nodes and receiving and processingthe aggregation results A sensor node in the network canchoose to be an aggregator nodewith the probability119875

119888which

is a preset value The aggregator nodes are responsible foraggregating target information and the query results collectedby other nodes The other nodes in the network will becomeleaf nodes and will be in charge of collecting and transmittinginformation to their neighbor aggregator nodes Each node inthe network can only communicate with its neighbor nodeswithin ℎ hops for a dense WSN we take ℎ = 1 It isimportant to note that the proposed aggregation scheme isused to perform addictive aggregation functions which arenot restrictive like sum average variance and so forth Andthe concrete details of addictive aggregation algorithms arenot covered in this paper

21 Network Model We consider a wireless network con-sisting of 119873 nodes and one base station Each node in thenetwork has the functionalities of sensing computing andtransmitting The collected primitive data of node 119894 can beexpressed as 119877

119894 and the aggregation function is defined as

119860119894= 119891(119877

119894) where 119860

119894denotes the aggregated data of node

119894 Meanwhile we introduced time variable into the functionand obtained 119860

119894(119905) = 119891(119877

119894(119905)) where 119877

119894(119905) denotes the

raw data collected by node 119894 at time 119905 and 119860119894(119905) means the

aggregation results of 119877119894(119905) To prevent the private data from

being eavesdropped on in the link level we used randomkey distribution mechanism in the scheme whose workmechanismwas introduced by Eschenauer andGligor in [18]and there are also some other good keymanagement schemes

International Journal of Distributed Sensor Networks 3

that have been proposed which are not covered in this paper[19ndash21] In the phase of key distribution a large key pool isgenerated and part of the keys in the pool will be drawn toform a key ring for a node After the key distribution anynodes that find out the neighbor which shared common keyswith them can maintain a secure communication link withthe neighbor In this paper we consider 119870

119901is the size of key

pool and 119870119903keys are taken out from the key pool randomly

Then we can express the probability that any third party hasthe shared key of a secure link as below

119875eavesdrop =119870119903

119870119901

(1)

The formula above can also be interpreted as the proba-bility that a secure link between two nodes is eavesdroppedon by a third node and we can see from the formula that thelarger the value of 119870

119901is the smaller 119875eavesdrop is the more

secure the link is If the size of key pool is big enough this keydistributionmechanism can ensure that the value of 119875eavesdropin the network is around 01 under the condition that anypair of nodes in the network shares at least one common keywith each other

22 The Design Goals of the Scheme

Efficiency Due to the energy constrained character of sensornodes data aggregation scheme was applied to reduce thecommunication overhead but consequently the applicationof secure scheme for data aggregation would inevitablycause some security problems Therefore how to maintaina balance between security and energy consumption hasbecome one of the core issues in the designing of securedata aggregation scheme In this paper the proposed schemeimproved the network security while keeping the networksystem at a low energy cost levelPrivacy Preservation The preservation of private data isalways a critical security issue in the design of secure dataaggregation scheme It means any node in the network couldonly know the private data of itself Some of the widely useddata aggregation schemes like TAG ignored this issue whichlead to the fact that a malicious node can eavesdrops on thecommunications of its neighbors to steal the private dataeasily and then cause a major security problem Thus thepreservation of data privacy is a key point in the designingof the proposed scheme in this paperAccuracy The accuracy of the aggregation result is a crucialcriterion in assessing the performance of data aggregationscheme It is affected by the success rate of data transmissionin the network directly and the factors influencing the successrate of data transmission include data loss transmittingcollision and so forth In a certain time the more the datais transmitted in the network the higher the risk of dataloss or transmitting collision Therefore in the designing ofthe proposed scheme we need to reduce the data traffic innetwork while providing the preservation of data privacyIn other words the more efficient the scheme is the moreaccurate the aggregation result would be

3 The Proposed Secure DataAggregation Scheme

31 The Formation of Aggregation Tree In the proposedscheme an aggregation tree rooted at the base station needsto be formed for organizing the sensor nodes in the networkFirst of all the base station broadcasts ldquoHirdquo message to all theneighbor nodes within one hop Any nodes without parentthat received ldquoHirdquo message should reply with ldquoJoin Requestrdquomessage to the originator but if the node received mul-tiple ldquoHirdquo messages from different senders then it shouldrandomly select one of them to be its parent node Oncea message of ldquoJoin Requestrdquo is received the BS accepts thenode to be its child node by replying to the message ofldquoJoin Acceptrdquo The probability that a child node becomes anaggregator node is 119875

119886which is a preset value and the rest

of the nodes will become leaf nodes That means there are119873 sdot 119875119886aggregator nodes and 119873 sdot (1 minus 119875

119886) leaf nodes in the

network consisting of119873 sensor nodes The aggregator nodeswill continue broadcasting ldquoHirdquo messages to find their childnodes If a node did not receive ldquoHirdquomessage it will broadcastldquoNo Parentrdquo message to its neighbor nodes to find its parentnode and the aggregator node that received this messagewill accept it as its child node After the formation of theaggregation tree if any aggregator node in the aggregationtree were to fail the parent of the failed aggregator nodewould broadcast ldquoHirdquo messages and try to communicate withthe child nodes of the failed aggregator node After receivingldquoHirdquo message the child nodes of the failed aggregator nodewill reply ldquoJoin Requestrdquo messages to the sender The parentof the failed aggregator node will randomly select one ofthese nodes to be the new aggregator node and reject allthe other join requests Next the new aggregator node willthen begin to broadcast ldquoHirdquo messages and accept all theother nodes which do not have a parent as their child nodesaccording to the rules introduced above The causes of nodefailure are numerous such as runout of power and physicaldamage compromised by a cyber-attack In future work wewill propose a trust management system of data aggregationfor monitoring node behavior and then analyze the causesof node failure to detect the attacking strategy of the attackerearly Figure 1 describes the whole procedure of formatting anaggregation tree

32 The Overview of the Proposed Secure Data AggregationScheme In this section we present the work mechanism ofESMART that can be divided into three stages data slicingdata mixing and data aggregationThe detailed introductionis as follows

Figure 2 describes an aggregation tree consisting ofone base station two aggregator nodes and six leaf nodesWe set 119869 = 4 which is a preset value that defines themaximum of 119895 which means the range of 119895 is from 2 to 119869and obeys the probability distribution formula 119865(119869 119895) Thecalculating concrete process of 119865(119869 119895) will be introduced inSection 4 We describe the work mechanism of ESMARTin this network environment After the collecting of targetdata a leaf node in the network slices the collected data


BSHi

Hi

JoinRequest

Join

Reque

st3

12

45

6

7

8

(a) BS broadcasts ldquoHirdquomessages to the neigh-bor nodes and the nodes that received themessage reply with ldquoJoin Requestrdquo to BS

JoinAcce

pt

Join

Accept

BS

12

3

45

6

7

8

(b) BS replies with ldquoJoin Acceptrdquo to thesenders and the nodes that received the replywill become the child nodes of BS

3

1

2

45

6

7

8

Hi

HiHi

HiHi

HiHi

Join Request

Join

Requ

est

Join

Requ

est

Join

Requ

est

Join

Reque

st

Join

Requ

est

BS

(c) Nodes 1 and 2 are child nodes of BS andchoose to be aggregator nodes so they broadcastldquoHirdquo messages to their neighbor nodes Node 6received ldquoHirdquo messages from different aggregatornodes and it randomly chooses one of them as itsparent

JoinAccept

Join

Acce

ptJoinAcce

pt

Join

Accep

t

Join

Acce

pt

Join

Acce

pt

3

1

2

45

6

7

8

BS

(d) Nodes 1 and 2 reply with ldquoJoin Acceptrdquo tothe child nodes

1

2

3

45

6

7

8

BS

(e) The aggregation tree is formed

Figure 1 The formation of aggregation tree

1

2

3

45

6

7

8

BS

Figure 2 Aggregation tree (ℎ = 1 119869 = 4)

into 119895 pieces and randomly chooses 119895 minus 1 neighbor nodeswhich shared common keys with it to be the receivers ofthe data pieces and then sends 119895 minus 1 encrypted data piecesto the receivers and keeps the remaining one by itself It isworth nothing that 119895 is a variable which is different fromSMARTWhen the node is an aggregator node it just receivesthe data pieces sent by the leaf nodes instead of slicing thecollected data of its own because the raw data collected bythe aggregator node will be assembled with the received datapieces in the step of data mixing which achieved the target ofhiding private data Figure 3 describes the step of data slicing119903119886is the raw data collected by node 119886 and 119903

119886119887denotes a piece

of data sent from node 119886 to node 119887 We take node 3 and node1 as an example as a leaf node node 3 sliced its primitive datainto 3 pieces 119903

33 11990331 and 119903

34and then sent the encrypted 119903

31

r31

r34r33

r44

r54

r77

r67r56

r55

r65

r61r82

r66

r72

r87

r78 r88

r41r51

1

2

3

4

5

6

7

8

BS

Figure 3 Data slicing

and 11990334

to node 1 and node 4 separately 11990333

is kept by itselfAs an aggregator node node 1 did not participate in the dataslicing and just received the data pieces sent by leaf nodes

After the stage of data slicing the nodes decrypt all thereceived data pieces and sum them up with their raw datapieceThe process can be described as119860

119887= sum119873

119886=1119903119886119887 119903119886119887= 0

if node 119887 is not one of the receivers of node 119886 where 119860119887

denotes the aggregation result of node 119887 And then we canexpress the final aggregation result as 119877 = sum119873

119887=1sum119873

119886=1119903119886119887

Figure 4 described the step of data mixing we take node 4as an example it sums up the received data pieces 119903

34and 11990354


1

2

3

4

5

6

7

8

BS

A1 = r1 + r31 + r41 + r51 + r61

A3 = r33

A4 = r44 + r34 + r54

A5 = r55 + r65A7 = r77 + r67 + r87

A8 = r88 + r78

A6 = r66 + r56

A2 = r2 + r72 + r82

Figure 4 Data mixing

with its raw data piece 11990344 and then the aggregation result of

node 4 is 1198604= 11990344+ 11990334+ 11990354

When all the sent data pieces are received leaf nodesbegin to encrypt their aggregation results and send themto the aggregator nodes The aggregators need to wait fora certain time to receive all the encrypted data pieces andthen aggregate all the received data pieces with their owndata After the aggregation the aggregators encrypt and sendthe aggregation results to their parent nodes The processcontinues until all the aggregation results arrived at the basestation This procedure is described in Figure 5

As we can see from the work process of ESMART aleaf node needs to perform 2 encryption operations and 1decryption operation during the process of data aggregationFor an aggregator node only 1 encryption operation and 1decryption operation are needed to be performed Obviouslythis degree of computational complexity has very little effecton the life cycle of sensor network and does not requirea highly computational capacity Meanwhile compared tocomputational complexity communication overhead hasalways been a major factor that affects the life cycle of sensornetwork Thus the analysis of the communication overheadof ESMART will be one of the focuses in Section 4

4 Simulation and Analysis

In this section simulations are carried out for analyzing andcomparing the performance of TAG SMART and ESMARTscheme First we set up the simulation environment inMATLAB which is an area of 400m lowast 400m and there are600 sensor nodes deployed in this area randomly with 119875

119886=

03 119869 denotes the maximum number of data pieces that theprimitive data can be sliced into in the network Thus thevalue of 119869will be set based on the demands of simulationTheother network parameters which are beyond the scope of thispaper will not be discussed The performance comparisonsof the schemes are focused on three aspects privacy preser-vation communication overhead and accuracy

41 Privacy-Preservation Analysis of ESMART In ESMARTscheme a sensor node which received a query from the basestation will start a collection of target data and slice the

1

2

3

45

6

7

8

BS

Figure 5 Data aggregation

collected data into 119895 pieces and then encrypt 119895 minus 1 pieces ofthe data and send them to the neighbor nodesTherefore theoutdegree of the node is 119895minus1Meanwhile the nodewill receive119896 pieces of encrypted data from its neighbors so the indegreeof the node is 119896 By analyzing the attack model we know thatonly in the case that all the 119869 minus 1 outgoing and 119896 incominglinks are broken by the eavesdropper the eavesdropper willbe able to steal the private data of the node In the SMARTscheme 119869 is a fixed value which is different from ESMARTschemeTherefore 119875

119904(119902 119869) the disclose probability of private

data in SMART can be expressed as below

119875119904(119902 119869) = 119902

119869minus1

idmax

sum

119896=0

119875 (indegree = 119896) 119902119896 (2)

The variable 119902 denotes the probability that a secure linkof the node is broken idmax is the maximum indegree ofnodes in the network and the value of idmax depends on119869 119875(indegree = 119896) is the probability that the indegree ofthe node equals 119896 sumidmax

119896=0119875(indegree = 119896)119902119896 denotes the

probability that all the outdegrees of the node are brokenand 119902119869minus1 is the probability that all the indegrees of the nodeare broken In the ESMART scheme 119869 is a preset valuewhich denotes the maximum number of data pieces that theprimitive data can be sliced into and 119895 is a variable whoserange of values allowed is from 2 to 119869 Then the formula ofthe disclose probability of private data in ESMART can bedescribed as below

119875119890(119902 119869) =

119869

sum

119895=2

119875 (outdegree = 119895) 119902119895minus1

times

idmax

sum

119896=0

119875 (indegree = 119896) 119902119896

(3)

where 119875(outdegree = 119895) is the probability that the outdegreeof the node is 119895 and then sum119869

119895=2119875(outdegree = 119895)119902119895minus1 denotes

the probability that all the outgoing links of the node arebroken As discussed in the previous chapter the probabilitydistribution of the node outdegree in the network dependson the performance-cost ratio of 119895 which means the 119895 withhigher performance-cost ratio will have higher probability to


Table 1 The probability distribution of 119895 with 119869 = 6

The value of 119895 2 3 4 5 6probability 03448 02299 01724 01379 01150

be chosen as the outdegree of the nodeTheperformance-costratio formula of 119895 is as below

119891 (119895) =1 minus 119875119904(119902 119895)

119895

=1 minus 119902119895minus1

sumidmax119896=0119875 (indegree = 119896) 119902119896

119895

(4)

where 1 minus 119875119904(119902 119895) is the secure probability of the communi-

cation link in the network and 119895 denotes the communicationcost of the node according to the formula of communicationoverhead in ESMARTTherefore the probability distributionformula of 119869 with different values of 119895 can be concluded asbelow

119865 (119869 119895) = 119875 (outdegree) =119891 (119895)

sum119869

119895=2119891 (119895) (5)

Table 1 illustrates the probability distribution of 119895 when119869 = 6

As we can see from Table 1 the larger value of 119895 thelower probability of the value to be selectedThat is the largervalue of 119895 the lower performance-cost ratio of the schemeAfter themathematical derivation the simulations are carriedout to compare the privacy-preservation performance ofSMART and ESMART TAG scheme was excluded from thecomparison due to lacking protection mechanism of dataprivacy The two schemes are simulated in the same networkenvironment with 119869 = 3 119875

119886= 03 and the simulation results

are shown in Figure 6As we can see from Figure 6 the disclose probability of

the private data in ESMART scheme is lower than that inSMART That is because the amount of communication inESMART is smaller than that in SMART which means lesschance to eavesdrop on the data pieces of one node that is thedisclose probability of the private data is reducedMeanwhilethe introduction of variable 119895 brings another advantage thetimes of data slicing of a sensor node in the network are onlyknown by itself so even if a malicious node eavesdroppedon several data pieces from the node it is not sure that allthe data pieces of the node are eavesdropped on unless thenumber of eavesdropped data pieces equals 119869 which is thetheoretical maximum value of 119895 Thus it is more difficult forthemalicious nodes to steal private data Given the above theprivacy-preservation performance of ESMART is better thanthat of SMART

42 Communication Overhead Analysis How to make areasonable tradeoff between security and communicationoverhead is always a critical issue in the designing of dataaggregation scheme for wireless sensor network As we cansee from the work in [4] SMART scheme introduced the

001 002 003 004 005 006 007 008 009 01The probability that communication link is eavesdropped on

SMARTESMART

The p

roba

bilit

y th

at p

rivat

e dat

a is e

xpos

ed (

)

0

005

01

015

02

025

03

035

04

045

05

Figure 6 Privacy preservation of SMART and ESMART

technique of data slicing and mixing to achieve privacypreservation of wireless sensor network but also bringssome additional communication overhead In our proposedscheme ESMART we decrease the communication overheadduring the procedure of aggregation while keeping a goodperformance of privacy preservation In this section tovalidate the efficiency of ESMART scheme we conclude thecalculating formulas of the communication overhead of TAGSMART and ESMART and then deploy these aggregationschemes in the same network environment separately toavoid the influence of other factors on the simulation resultsFirst we assume a network consisting of 119873 sensor nodes InTAG scheme the nodes upload their raw data directly with-out any protection of private data thus the communicationoverhead can be expressed as below

TAG CO119905= 119873 (6)

In SMART scheme each node slices the primitive data into119869 pieces and sends 119869 minus 1 pieces to its neighbor nodes topreserve privacy After that the nodes upload the processeddata to their parent nodes according to the procedure of TAGscheme so the communication overhead formula of SMARTcan be concluded as below

SMART CO119904= 119873 sdot 119869 (7)

In ESMART scheme the aggregator nodes do not participatein data slicing Thus the data traffic of aggregator nodes inthe network is119873 sdot 119875

119886 where 119875

119886denotes the probability that a

sensor node becomes an aggregator nodeTherefore there are119873(1minus119875

119886) leaf nodes in the network Each leaf node sends 119895

119894minus1

pieces of raw data to the neighbor nodes and 119895119894denotes the

value of 119895 of node 119894 After that it transmits the mixed data to


the aggregator node So we can express the communicationoverhead formula of ESMART as below

ESMART CO119890= 119873 sdot 119875

119886+

119873(1minus119875119886)

sum

119894=1

119895119894 (119895

119894isin [2 119869]) (8)

As we introduced in Section 3 119895 is a variable which takesvalues based on a probability distribution function 119865(119869 119895)and the range of values allowed is from 2 to 119869 Therefore weintroduce function119865(119869 119895) into the calculating formula ofCO

119890

and derived the formula which is shown below

CO119890= 119873 sdot 119875

119886+

119869

sum

119895=2

119865 (119869 119895) sdot 119873 (1 minus 119875119886) sdot 119895 (9)

To compare the communication overheads of SMART andESMART we set 119876 = CO

119904CO119890 Thus if 119876 ge 1 then

CO119904ge CO

119890 If 119876 le 1 then CO

119904le CO

119890 Therefore

119876 =119873 sdot 119869

119873 sdot 119875119886+ sum119869

119895=2119865 (119869 119895) sdot 119873 (1 minus 119875

119886) sdot 119895

=119869

119875119886+ (1 minus 119875

119886)sum119869

119895=2119865 (119869 119895) sdot 119895

(10)

From the definition of 119865(119869 119895) we can deduce thatsum119869

119895=2119865(119869 119895) sdot 119895 lt 119869 then we set 119876 gt 1198761015840 and the expression

of 1198761015840 is shown as below

119876 gt 1198761015840

=119869

119875119886+ (1 minus 119875

119886) sdot 119869=

119869

119869 + 119875119886minus119875119886sdot 119869

=119869

119869 + 119875119886(1 minus 119869)

(11)

For 119869 ge 2 and 119875119886isin (0 1) we can determine that 119875

119886(1 minus 119869) lt

0 Thus 119869 + 119875119886(1 minus 119869) lt 119869 and then 119876 gt 1198761015840 gt 1 From the

derivation above we can conclude that CO119904gt CO

119890 After

themathematical derivation we deployed TAG SMART andESMART scheme in the simulation environment separatelyand the value of 119869 in the simulation environment is set to 4119875119886= 03 The simulation results are shown in Figure 7It is concluded from the simulation results above that

because of lacking protection of data privacy the number ofmessages transmitted during the procedure of data acquisi-tion and aggregation under the TAG scheme is very smallMeanwhile owing to the introduction of data slicing tech-nique the communication overhead of the network deployedwith SMART is much higher than that with TAG And thecommunication overhead of the proposed scheme ESMARTis about 37 to 46 lower than that of SMART which meansthat ESMART is more energy efficient than SMART whilekeeping a better performance of privacy preservation

43 Accuracy Analysis The accuracy of aggregation resultis an important indicator of the performance of data aggre-gation scheme Theoretically the accuracy of aggregationresult is 100 which means the final aggregation result

0 05 1 15 2 250

500

1000

1500

2000

2500

3000

Com

mun

icat

ion

over

head

SMARTTAGESMART

Time interval (s) J = 4 Pa = 03

Figure 7 Communication overhead of TAG SMART andESMART

equals the sum of the collected data of every node in thenetwork In reality due to channel sharing character ofwireless network data loss is inevitable during the procedureof data aggregation The more transmissions per unit timein the network the more collisions and then the more dataloss In other words the data accuracy in wireless network isinversely associatedwith the communication overheadThusfrom the simulation results of communication overhead ofthe schemes we can theoretically deduce that the accuracyperformance of TAG is the best in these schemes secondlyESMART thirdly SMART After the analysis we deployed theschemes in the simulation environment with 119869 = 3 119875

119886= 03

and the simulation results are shown in Figure 8As we can see from the simulation results due to

the low amount of communication the accuracy of TAGscheme is about 90 when the interval time is long enoughInstead the accuracy of SMART scheme is approximately67 which is caused by the large amount of data traffic Theproposed scheme ESMART can provide a better accuracyof 83 than SMART while keeping a good performance ofprivacy preservation The simulation results confirmed theprevious deduction that the aggregation scheme with lowcommunication overhead can achieve a good performance ofdata accuracy

5 Conclusions

Wireless sensor network consists of a large number of low-powered resource-constrained sensor nodes and is usuallydeployed in unattended environment Sensor nodes in thenetwork can sense specific information and perform wirelesscommunication within a small range of their location To


0 5 10 15 20 250

10

20

30

40

50

60

70

80

90

100

Accu

racy

()

SMART

TAGESMART


Figure 8 Accuracy of TAG SMART and ESMART

reduce the energy consumption of the nodes during thetransmission data aggregation scheme has been widely usedin WSN Meanwhile how to provide the protection of dataprivacy during the aggregation is becoming a desiderativeproblem to be solved

In this paper we proposed an energy-efficient secure dataaggregation scheme ESMART The simulation results andanalysis proved that whether in energy saving ability or dataaccuracy ESMART scheme is better than SMART schemewhile keeping a good performance of privacy preservationwhich is not provided in TAG scheme It is an efficient securedata aggregation scheme with a value of practical applicationand it achieved the expected design requirements In futureresearch work we will focus on the designing of securedata aggregation scheme with the ability to guarantee dataintegrity

Acknowledgments

This research is supported by Beijing Science and TechnologyProgramunderGrant Z121100007612003 BeijingNatural Sci-ence Foundation under Grant 4132057 and National NaturalScience Foundation under Grant 61371071

References

[1] S Madden M J Franklin J M Hellerstein and H WeildquoTAG a Tiny AGgregation service for ad-hoc sensor networksrdquoin Usenix Association Proceedings of the 5th Symposium onOperating Systems Design and Implementation pp 131ndash1462002

[2] S Ozdemir and Y Xiao ldquoSecure data aggregation in wirelesssensor networks a comprehensive overviewrdquo Computer Net-works vol 53 no 12 pp 2022ndash2037 2009

[3] K J Mukesh and T Sharma ldquoSecure data aggregation inwireless sensor network a surveyrdquo International Journal ofEngineering Science and Technology vol 3 no 3 pp 2013ndash20192011

[4] W He X Liu H Viet K Nahrstedt and T AbdelzaherldquoPDA privacy-preserving data aggregation for informationcollectionrdquo ACM Transactions on Sensor Networks vol 8 no1 article no 6 2011

[5] W He H Nguyen X Liu K Nahrstedt and T AbdelzaherldquoiPDA an integrity-protecting private data aggregation schemeforwireless sensor networksrdquo inProceedings of the IEEEMilitaryCommunications Conference (MILCOM rsquo08) vol 1ndash7 pp 4071ndash4077 November 2008

[6] C X Liu Y Liu Z J Zhang and Z Y Cheng ldquoHigh energy-efficient and privacy-preserving secure data aggregation forwireless sensor networksrdquo International Journal of Communica-tion Systems vol 26 no 3 pp 380ndash394 2013

[7] H Li K Lin and K Li ldquoEnergy-efficient and high-accuracysecure data aggregation in wireless sensor networksrdquo ComputerCommunications vol 34 no 4 pp 591ndash597 2011

[8] N Li N Zhang S K Das and B Thuraisingham ldquoPrivacypreservation in wireless sensor networks a state-of-the-artsurveyrdquo Ad Hoc Networks vol 7 no 8 pp 1501ndash1514 2009

[9] R Bista and J-W Chang ldquoPrivacy-preserving data aggregationprotocols for wireless sensor networks a surveyrdquo Sensors vol10 no 5 pp 4577ndash4601 2010

[10] A Miyaji and K Omote ldquoEfficient and secure aggregation ofsensor data against multiple corrupted nodesrdquo IEICE Transac-tions on Information and Systems vol E94-D no 10 pp 1955ndash1965 2011

[11] HAlzaid E Foo J GNieto andEAhmed ldquoMitigatingOn-Offattacks in reputation-based secure data aggregation for wirelesssensor networksrdquo Security and Communication Networks vol 5no 2 pp 125ndash144 2012

[12] S ZareAfifi R Verma B King P Salama and D KimldquoSecure countermeasures to data aggregation attacks on sensornetworksrdquo inProceedings of the 55th IEEE InternationalMidwestSymposium on Circuits and Systems (Mwscas rsquo12) pp 856ndash8592012

[13] W T Zhu F Gao and Y Xiang ldquoA secure and efficient dataaggregation scheme for wireless sensor networksrdquo ConcurrencyComputation Practice amp Experience vol 23 no 12 pp 1414ndash1430 2011

[14] X M Dong and S S Li ldquoSecure data aggregation approachbased on monitoring in wireless sensor networksrdquo ChinaCommunications vol 9 no 6 pp 14ndash27 2012

[15] F-Y Lei C Fu X Li and J Chen ldquoSecure data aggregationsolution based on dynamic multiple cluster key managementmodelrdquo Journal of Internet Technology vol 12 no 3 pp 465ndash476 2011

[16] H M Sun C H Chen and P C Li ldquoA lightweight secure dataaggregation protocol for wireless sensor networksrdquo Interna-tional Journal of Innovative Computing Information and Controlvol 8 no 10A pp 6503ndash6514 2012

[17] A S Poornima and B B Amberker ldquoSecure end-to-end dataaggregation (SEEDA) protocols for wireless sensor networksrdquoAd Hoc amp Sensor Wireless Networks vol 17 no 3-4 pp 193ndash2192013


[18] L Eschenauer and V D Gligor ldquoA key-management schemefor distributed sensor networksrdquo in Proceedings of the 9th ACMConference on Computer and Communications Security pp 41ndash47 November 2002

[19] Q Xie ldquoA new authenticated key agreement for session initia-tion protocolrdquo International Journal of Communication Systemsvol 25 no 1 pp 47ndash54 2012

[20] D-C Lou and H-F Huang ldquoEfficient three-party password-based key exchange schemerdquo International Journal of Commu-nication Systems vol 24 no 4 pp 504ndash512 2011

[21] D He J Chen and J Hu ldquoA pairing-free certificatelessauthenticated key agreement protocolrdquo International Journal ofCommunication Systems vol 25 no 2 pp 221ndash230 2012

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Active and Passive Electronic Components

Control Scienceand Engineering

Journal of



RotatingMachinery


Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design



Shock and Vibration


Civil EngineeringAdvances in

Acoustics and VibrationAdvances in



Electrical and Computer Engineering

Journal of

Advances inOptoElectronics


Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of


Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014


Chemical EngineeringInternational Journal of Antennas and

Propagation




Navigation and Observation



DistributedSensor Networks



paper we propose a novel secure data aggregation schemefor WSN ESMART (energy-efficient slice-mix-aggregate)which addresses both energy-efficient and privacy preserva-tion of the network To achieve the privacy preservation of thenetwork we introduced data slicing and mixing techniqueinto the scheme designing A secure data aggregation schemebased on data slicing and mixing technique was detailedin [4] in this scheme in order to achieve the preservationof data privacy in the network the raw readings of thesensor nodes will be sliced into pieces and part of the pieceswill be encrypted and sent to the neighbor nodes But thiscomes at the price of some additional communication costwhich is shown in the simulation result of SMART schemethe communication cost of SMART is about 119869 times thecommunication cost of TAG where 119869 denotes the number ofdata pieces Thus we introduced the dynamic 119869 value into thescheme which can effectively decrease the communicationoverhead while maintaining the privacy preservation ofthe network Meanwhile a lower communication overheadmeans a lower transmission collision probability which leadsto a higher aggregation accuracy in the networkMoreover inESMART the times of data slicing of a sensor node are onlyknown by itself so even if a malicious node eavesdroppedon several data pieces from the node it is not sure that allthe data pieces of the node are eavesdropped on Thus it ismore difficult for the malicious nodes to steal private dataGiven the above the use of data slicing andmixing techniquein ESMART is more efficient which consequently reducesthe communication overhead and increases the aggregationaccuracy effectively

Wireless sensor network is usually deployed in unat-tended environment and consisted of a large number oflow-powered resource-constrained sensor nodes Each nodein the network can sense the environment information asrequested by the base station and perform wireless commu-nication within a small range of its location To reduce theenergy consumption of the nodes during the transmissiondata aggregation scheme has been widely used in WSNMadden et al proposed TAG scheme as a simple dataaggregation scheme based on the assumption that every nodein the network is trusted and there is no protection of dataprivacy in this scheme But in practice the application ofWSN is facing various kinds of security threats A goodsecure data aggregation scheme should possess the capabilityto resist the security threats while staying within energyconstraint To meet this requirement some secure dataaggregation schemes have been provided He et al proposed anovel data aggregation scheme with disjoint aggregation treestructure to ensure data integrity in [5] Liu et al proposed asecure data aggregation schemeHEEPwith a novel formationmethod of aggregation tree in [6] in 2013 Li et al proposeda secure data aggregation scheme based on the improvedSMART scheme which is more efficient [7] Both Li et al[8] and Bista and Chang [9] proposed survey papers of theprivacy-preserving techniques for WSN Meanwhile someresearchers analyzed the existing network attack technologiesagainst WSN and proposed corresponding solutions [10ndash12] Zhu et al proposed a secure data aggregation schemein which the base station composes a secret configuration

matrix and each sensor node is preloaded with a limitedpart of the matrix known as a secret share containing certainlocal instructions [13] Dong and Li presented a secure dataaggregation approach based on monitoring in WSN [14] Inthis paper a grid-based network architecture for monitoringin WSN is designed and the algorithms for network divisioninitialization and grid tree construction are proposed Leiet al proposed a secure data aggregation solution based ondynamic multiple cluster key management model in [15] thiskey management model consumes little storage space andcan resist the collusion attack effectively Sun et al proposeda lightweight secure data aggregation protocol to find thecompromised nodes and help the base station to verify thefinal results [16] Poornima and Amberker proposed a securedata aggregation scheme in [17] This scheme provides end-to-end data privacy and can effectively reduce the energyconsumption

2 Network Model andthe Requirement of Design

In this paper we used aggregation tree to organize the sensornodes in the network and consider three types of nodes inthe network base station aggregator node and leaf nodeThere is only one base station in the network which can beseen as the network control center with unlimited resourcesand the final destination for the aggregation results As theroot of the aggregation tree it is responsible for broadcastingqueries to all the other nodes and receiving and processingthe aggregation results A sensor node in the network canchoose to be an aggregator nodewith the probability119875

119888which

is a preset value The aggregator nodes are responsible foraggregating target information and the query results collectedby other nodes The other nodes in the network will becomeleaf nodes and will be in charge of collecting and transmittinginformation to their neighbor aggregator nodes Each node inthe network can only communicate with its neighbor nodeswithin ℎ hops for a dense WSN we take ℎ = 1 It isimportant to note that the proposed aggregation scheme isused to perform addictive aggregation functions which arenot restrictive like sum average variance and so forth Andthe concrete details of addictive aggregation algorithms arenot covered in this paper

21 Network Model We consider a wireless network con-sisting of 119873 nodes and one base station Each node in thenetwork has the functionalities of sensing computing andtransmitting The collected primitive data of node 119894 can beexpressed as 119877

119894 and the aggregation function is defined as

119860119894= 119891(119877

119894) where 119860

119894denotes the aggregated data of node

119894 Meanwhile we introduced time variable into the functionand obtained 119860

119894(119905) = 119891(119877

119894(119905)) where 119877

119894(119905) denotes the

raw data collected by node 119894 at time 119905 and 119860119894(119905) means the

aggregation results of 119877119894(119905) To prevent the private data from

being eavesdropped on in the link level we used randomkey distribution mechanism in the scheme whose workmechanismwas introduced by Eschenauer andGligor in [18]and there are also some other good keymanagement schemes






119875eavesdrop =119870119903

119870119901

(1)















BSHi

Hi

JoinRequest

Join

Reque

st3

12

45

6

7

8


JoinAcce

pt

Join

Accept

BS

12

3

45

6

7

8


3

1

2

45

6

7

8

Hi

HiHi

HiHi

HiHi

Join Request

Join

Requ

est

Join

Requ

est

Join

Requ

est

Join

Reque

st

Join

Requ

est

BS


JoinAccept

Join

Acce

ptJoinAcce

pt

Join

Accep

t

Join

Acce

pt

Join

Acce

pt

3

1

2

45

6

7

8

BS


1

2

3

45

6

7

8

BS



1

2

3

45

6

7

8

BS





33 11990331 and 119903


31

r31

r34r33

r44

r54

r77

r67r56

r55

r65

r61r82

r66

r72

r87

r78 r88

r41r51

1

2

3

4

5

6

7

8

BS


and 11990334




119887= sum119873

119886=1119903119886119887 119903119886119887= 0



119887=1sum119873

119886=1119903119886119887


34and 11990354


1

2

3

4

5

6

7

8

BS

A1 = r1 + r31 + r41 + r51 + r61

A3 = r33

A4 = r44 + r34 + r54

A5 = r55 + r65A7 = r77 + r67 + r87

A8 = r88 + r78

A6 = r66 + r56

A2 = r2 + r72 + r82



node 4 is 1198604= 11990344+ 11990334+ 11990354





119886=



1

2

3

45

6

7

8

BS





119875119904(119902 119869) = 119902

119869minus1

idmax

sum

119896=0

119875 (indegree = 119896) 119902119896 (2)




119875119890(119902 119869) =

119869

sum

119895=2

119875 (outdegree = 119895) 119902119895minus1

times

idmax

sum

119896=0

119875 (indegree = 119896) 119902119896

(3)








119891 (119895) =1 minus 119875119904(119902 119895)

119895

=1 minus 119902119895minus1


119895

(4)



119865 (119869 119895) = 119875 (outdegree) =119891 (119895)

sum119869

119895=2119891 (119895) (5)








SMARTESMART

The p

roba

bilit

y th

at p

rivat

e dat

a is e

xpos

ed (

)

0

005

01

015

02

025

03

035

04

045

05



TAG CO119905= 119873 (6)


SMART CO119904= 119873 sdot 119869 (7)


119886 where 119875




119894minus1





ESMART CO119890= 119873 sdot 119875

119886+

119873(1minus119875119886)

sum

119894=1

119895119894 (119895

119894isin [2 119869]) (8)


119890


CO119890= 119873 sdot 119875

119886+

119869

sum

119895=2

119865 (119869 119895) sdot 119873 (1 minus 119875119886) sdot 119895 (9)


119904CO119890 Thus if 119876 ge 1 then

CO119904ge CO

119890 If 119876 le 1 then CO

119904le CO

119890 Therefore

119876 =119873 sdot 119869

119873 sdot 119875119886+ sum119869

119895=2119865 (119869 119895) sdot 119873 (1 minus 119875

119886) sdot 119895

=119869

119875119886+ (1 minus 119875

119886)sum119869

119895=2119865 (119869 119895) sdot 119895

(10)




119876 gt 1198761015840

=119869

119875119886+ (1 minus 119875

119886) sdot 119869=

119869

119869 + 119875119886minus119875119886sdot 119869

=119869

119869 + 119875119886(1 minus 119869)

(11)


119886(1 minus 119869) lt



119890 After




0 05 1 15 2 250

500

1000

1500

2000

2500

3000

Com

mun

icat

ion

over

head

SMARTTAGESMART




119886= 03



5 Conclusions



0 5 10 15 20 250

10

20

30

40

50

60

70

80

90

100

Accu

racy

()

SMART

TAGESMART





Acknowledgments


References

























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation














119875eavesdrop =119870119903

119870119901

(1)















BSHi

Hi

JoinRequest

Join

Reque

st3

12

45

6

7

8


JoinAcce

pt

Join

Accept

BS

12

3

45

6

7

8


3

1

2

45

6

7

8

Hi

HiHi

HiHi

HiHi

Join Request

Join

Requ

est

Join

Requ

est

Join

Requ

est

Join

Reque

st

Join

Requ

est

BS


JoinAccept

Join

Acce

ptJoinAcce

pt

Join

Accep

t

Join

Acce

pt

Join

Acce

pt

3

1

2

45

6

7

8

BS


1

2

3

45

6

7

8

BS



1

2

3

45

6

7

8

BS





33 11990331 and 119903


31

r31

r34r33

r44

r54

r77

r67r56

r55

r65

r61r82

r66

r72

r87

r78 r88

r41r51

1

2

3

4

5

6

7

8

BS


and 11990334




119887= sum119873

119886=1119903119886119887 119903119886119887= 0



119887=1sum119873

119886=1119903119886119887


34and 11990354


1

2

3

4

5

6

7

8

BS

A1 = r1 + r31 + r41 + r51 + r61

A3 = r33

A4 = r44 + r34 + r54

A5 = r55 + r65A7 = r77 + r67 + r87

A8 = r88 + r78

A6 = r66 + r56

A2 = r2 + r72 + r82



node 4 is 1198604= 11990344+ 11990334+ 11990354





119886=



1

2

3

45

6

7

8

BS





119875119904(119902 119869) = 119902

119869minus1

idmax

sum

119896=0

119875 (indegree = 119896) 119902119896 (2)




119875119890(119902 119869) =

119869

sum

119895=2

119875 (outdegree = 119895) 119902119895minus1

times

idmax

sum

119896=0

119875 (indegree = 119896) 119902119896

(3)








119891 (119895) =1 minus 119875119904(119902 119895)

119895

=1 minus 119902119895minus1


119895

(4)



119865 (119869 119895) = 119875 (outdegree) =119891 (119895)

sum119869

119895=2119891 (119895) (5)








SMARTESMART

The p

roba

bilit

y th

at p

rivat

e dat

a is e

xpos

ed (

)

0

005

01

015

02

025

03

035

04

045

05



TAG CO119905= 119873 (6)


SMART CO119904= 119873 sdot 119869 (7)


119886 where 119875




119894minus1





ESMART CO119890= 119873 sdot 119875

119886+

119873(1minus119875119886)

sum

119894=1

119895119894 (119895

119894isin [2 119869]) (8)


119890


CO119890= 119873 sdot 119875

119886+

119869

sum

119895=2

119865 (119869 119895) sdot 119873 (1 minus 119875119886) sdot 119895 (9)


119904CO119890 Thus if 119876 ge 1 then

CO119904ge CO

119890 If 119876 le 1 then CO

119904le CO

119890 Therefore

119876 =119873 sdot 119869

119873 sdot 119875119886+ sum119869

119895=2119865 (119869 119895) sdot 119873 (1 minus 119875

119886) sdot 119895

=119869

119875119886+ (1 minus 119875

119886)sum119869

119895=2119865 (119869 119895) sdot 119895

(10)




119876 gt 1198761015840

=119869

119875119886+ (1 minus 119875

119886) sdot 119869=

119869

119869 + 119875119886minus119875119886sdot 119869

=119869

119869 + 119875119886(1 minus 119869)

(11)


119886(1 minus 119869) lt



119890 After




0 05 1 15 2 250

500

1000

1500

2000

2500

3000

Com

mun

icat

ion

over

head

SMARTTAGESMART




119886= 03



5 Conclusions



0 5 10 15 20 250

10

20

30

40

50

60

70

80

90

100

Accu

racy

()

SMART

TAGESMART





Acknowledgments


References

























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










BSHi

Hi

JoinRequest

Join

Reque

st3

12

45

6

7

8


JoinAcce

pt

Join

Accept

BS

12

3

45

6

7

8


3

1

2

45

6

7

8

Hi

HiHi

HiHi

HiHi

Join Request

Join

Requ

est

Join

Requ

est

Join

Requ

est

Join

Reque

st

Join

Requ

est

BS


JoinAccept

Join

Acce

ptJoinAcce

pt

Join

Accep

t

Join

Acce

pt

Join

Acce

pt

3

1

2

45

6

7

8

BS


1

2

3

45

6

7

8

BS



1

2

3

45

6

7

8

BS





33 11990331 and 119903


31

r31

r34r33

r44

r54

r77

r67r56

r55

r65

r61r82

r66

r72

r87

r78 r88

r41r51

1

2

3

4

5

6

7

8

BS


and 11990334




119887= sum119873

119886=1119903119886119887 119903119886119887= 0



119887=1sum119873

119886=1119903119886119887


34and 11990354


1

2

3

4

5

6

7

8

BS

A1 = r1 + r31 + r41 + r51 + r61

A3 = r33

A4 = r44 + r34 + r54

A5 = r55 + r65A7 = r77 + r67 + r87

A8 = r88 + r78

A6 = r66 + r56

A2 = r2 + r72 + r82



node 4 is 1198604= 11990344+ 11990334+ 11990354





119886=



1

2

3

45

6

7

8

BS





119875119904(119902 119869) = 119902

119869minus1

idmax

sum

119896=0

119875 (indegree = 119896) 119902119896 (2)




119875119890(119902 119869) =

119869

sum

119895=2

119875 (outdegree = 119895) 119902119895minus1

times

idmax

sum

119896=0

119875 (indegree = 119896) 119902119896

(3)








119891 (119895) =1 minus 119875119904(119902 119895)

119895

=1 minus 119902119895minus1


119895

(4)



119865 (119869 119895) = 119875 (outdegree) =119891 (119895)

sum119869

119895=2119891 (119895) (5)








SMARTESMART

The p

roba

bilit

y th

at p

rivat

e dat

a is e

xpos

ed (

)

0

005

01

015

02

025

03

035

04

045

05



TAG CO119905= 119873 (6)


SMART CO119904= 119873 sdot 119869 (7)


119886 where 119875




119894minus1





ESMART CO119890= 119873 sdot 119875

119886+

119873(1minus119875119886)

sum

119894=1

119895119894 (119895

119894isin [2 119869]) (8)


119890


CO119890= 119873 sdot 119875

119886+

119869

sum

119895=2

119865 (119869 119895) sdot 119873 (1 minus 119875119886) sdot 119895 (9)


119904CO119890 Thus if 119876 ge 1 then

CO119904ge CO

119890 If 119876 le 1 then CO

119904le CO

119890 Therefore

119876 =119873 sdot 119869

119873 sdot 119875119886+ sum119869

119895=2119865 (119869 119895) sdot 119873 (1 minus 119875

119886) sdot 119895

=119869

119875119886+ (1 minus 119875

119886)sum119869

119895=2119865 (119869 119895) sdot 119895

(10)




119876 gt 1198761015840

=119869

119875119886+ (1 minus 119875

119886) sdot 119869=

119869

119869 + 119875119886minus119875119886sdot 119869

=119869

119869 + 119875119886(1 minus 119869)

(11)


119886(1 minus 119869) lt



119890 After




0 05 1 15 2 250

500

1000

1500

2000

2500

3000

Com

mun

icat

ion

over

head

SMARTTAGESMART




119886= 03



5 Conclusions



0 5 10 15 20 250

10

20

30

40

50

60

70

80

90

100

Accu

racy

()

SMART

TAGESMART





Acknowledgments


References

























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










1

2

3

4

5

6

7

8

BS

A1 = r1 + r31 + r41 + r51 + r61

A3 = r33

A4 = r44 + r34 + r54

A5 = r55 + r65A7 = r77 + r67 + r87

A8 = r88 + r78

A6 = r66 + r56

A2 = r2 + r72 + r82



node 4 is 1198604= 11990344+ 11990334+ 11990354





119886=



1

2

3

45

6

7

8

BS





119875119904(119902 119869) = 119902

119869minus1

idmax

sum

119896=0

119875 (indegree = 119896) 119902119896 (2)




119875119890(119902 119869) =

119869

sum

119895=2

119875 (outdegree = 119895) 119902119895minus1

times

idmax

sum

119896=0

119875 (indegree = 119896) 119902119896

(3)








119891 (119895) =1 minus 119875119904(119902 119895)

119895

=1 minus 119902119895minus1


119895

(4)



119865 (119869 119895) = 119875 (outdegree) =119891 (119895)

sum119869

119895=2119891 (119895) (5)








SMARTESMART

The p

roba

bilit

y th

at p

rivat

e dat

a is e

xpos

ed (

)

0

005

01

015

02

025

03

035

04

045

05



TAG CO119905= 119873 (6)


SMART CO119904= 119873 sdot 119869 (7)


119886 where 119875




119894minus1





ESMART CO119890= 119873 sdot 119875

119886+

119873(1minus119875119886)

sum

119894=1

119895119894 (119895

119894isin [2 119869]) (8)


119890


CO119890= 119873 sdot 119875

119886+

119869

sum

119895=2

119865 (119869 119895) sdot 119873 (1 minus 119875119886) sdot 119895 (9)


119904CO119890 Thus if 119876 ge 1 then

CO119904ge CO

119890 If 119876 le 1 then CO

119904le CO

119890 Therefore

119876 =119873 sdot 119869

119873 sdot 119875119886+ sum119869

119895=2119865 (119869 119895) sdot 119873 (1 minus 119875

119886) sdot 119895

=119869

119875119886+ (1 minus 119875

119886)sum119869

119895=2119865 (119869 119895) sdot 119895

(10)




119876 gt 1198761015840

=119869

119875119886+ (1 minus 119875

119886) sdot 119869=

119869

119869 + 119875119886minus119875119886sdot 119869

=119869

119869 + 119875119886(1 minus 119869)

(11)


119886(1 minus 119869) lt



119890 After




0 05 1 15 2 250

500

1000

1500

2000

2500

3000

Com

mun

icat

ion

over

head

SMARTTAGESMART




119886= 03



5 Conclusions



0 5 10 15 20 250

10

20

30

40

50

60

70

80

90

100

Accu

racy

()

SMART

TAGESMART





Acknowledgments


References

























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation













119891 (119895) =1 minus 119875119904(119902 119895)

119895

=1 minus 119902119895minus1


119895

(4)



119865 (119869 119895) = 119875 (outdegree) =119891 (119895)

sum119869

119895=2119891 (119895) (5)








SMARTESMART

The p

roba

bilit

y th

at p

rivat

e dat

a is e

xpos

ed (

)

0

005

01

015

02

025

03

035

04

045

05



TAG CO119905= 119873 (6)


SMART CO119904= 119873 sdot 119869 (7)


119886 where 119875




119894minus1





ESMART CO119890= 119873 sdot 119875

119886+

119873(1minus119875119886)

sum

119894=1

119895119894 (119895

119894isin [2 119869]) (8)


119890


CO119890= 119873 sdot 119875

119886+

119869

sum

119895=2

119865 (119869 119895) sdot 119873 (1 minus 119875119886) sdot 119895 (9)


119904CO119890 Thus if 119876 ge 1 then

CO119904ge CO

119890 If 119876 le 1 then CO

119904le CO

119890 Therefore

119876 =119873 sdot 119869

119873 sdot 119875119886+ sum119869

119895=2119865 (119869 119895) sdot 119873 (1 minus 119875

119886) sdot 119895

=119869

119875119886+ (1 minus 119875

119886)sum119869

119895=2119865 (119869 119895) sdot 119895

(10)




119876 gt 1198761015840

=119869

119875119886+ (1 minus 119875

119886) sdot 119869=

119869

119869 + 119875119886minus119875119886sdot 119869

=119869

119869 + 119875119886(1 minus 119869)

(11)


119886(1 minus 119869) lt



119890 After




0 05 1 15 2 250

500

1000

1500

2000

2500

3000

Com

mun

icat

ion

over

head

SMARTTAGESMART




119886= 03



5 Conclusions



0 5 10 15 20 250

10

20

30

40

50

60

70

80

90

100

Accu

racy

()

SMART

TAGESMART





Acknowledgments


References

























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











ESMART CO119890= 119873 sdot 119875

119886+

119873(1minus119875119886)

sum

119894=1

119895119894 (119895

119894isin [2 119869]) (8)


119890


CO119890= 119873 sdot 119875

119886+

119869

sum

119895=2

119865 (119869 119895) sdot 119873 (1 minus 119875119886) sdot 119895 (9)


119904CO119890 Thus if 119876 ge 1 then

CO119904ge CO

119890 If 119876 le 1 then CO

119904le CO

119890 Therefore

119876 =119873 sdot 119869

119873 sdot 119875119886+ sum119869

119895=2119865 (119869 119895) sdot 119873 (1 minus 119875

119886) sdot 119895

=119869

119875119886+ (1 minus 119875

119886)sum119869

119895=2119865 (119869 119895) sdot 119895

(10)




119876 gt 1198761015840

=119869

119875119886+ (1 minus 119875

119886) sdot 119869=

119869

119869 + 119875119886minus119875119886sdot 119869

=119869

119869 + 119875119886(1 minus 119869)

(11)


119886(1 minus 119869) lt



119890 After




0 05 1 15 2 250

500

1000

1500

2000

2500

3000

Com

mun

icat

ion

over

head

SMARTTAGESMART




119886= 03



5 Conclusions



0 5 10 15 20 250

10

20

30

40

50

60

70

80

90

100

Accu

racy

()

SMART

TAGESMART





Acknowledgments


References

























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










0 5 10 15 20 250

10

20

30

40

50

60

70

80

90

100

Accu

racy

()

SMART

TAGESMART





Acknowledgments


References

























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation
















RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation









research article esmart: energy-efficient slice-mix...

Documents