13
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. IEEE TRANSACTIONS ON COMPUTATIONAL SOCIAL SYSTEMS 1 A Blockchain-Powered Crowdsourcing Method With Privacy Preservation in Mobile Environment Xiaolong Xu , Qingxiang Liu, Xuyun Zhang , Jie Zhang, Lianyong Qi , and Wanchun Dou Abstract—Crowdsourcing is a booming technique that enables participants to exchange data directly, thus making it possible to answer latency-sensitive service requests and relieve the burden of core networks. With some incentives, providers compete to furnish service requests, thus pledging the quality of experi- ence (QoE) for requestors. However, the decentralized communi- cation in crowdsourcing increases the probability of information tapering. Furthermore, providers’ arbitrary selection of the requests poses great threat to the efficient and profitable service provision for the requestors. To deal with these challenges, we propose a blockchain-powered crowdsourcing method, named BPCM, while considering the privacy preservation in mobile environment. Specifically, a mobile crowdsourcing framework based on blockchain is designed first to preserve the privacy of the participants and keep the integrity of the service request and pro- vision. Then, density-based spatial clustering of applications with noise (DBSCAN) and improved dynamic programming (IDP) are adopted to cluster the requestors and generate service strategies, respectively. Furthermore, simple additive weighting (SAW) and multiple criteria decision making (MCDM) are utilized to select the optimal strategy that achieves the tradeoffs among maxi- mizing the service time, increasing the profits, and reducing the energy consumption for the providers. Finally, comprehensive experiments are conducted to verify the accuracy and effective- ness of BPCM. Index Terms— Blockchain, crowdsourcing, mobile environ- ment, privacy preservation. I. I NTRODUCTION A. Background C ROWDSOURCING, as an efficient method in machine computation, provides such a scenario where the tasks Manuscript received January 18, 2019; revised February 27, 2019 and March 22, 2019; accepted April 1, 2019. This work was supported in part by the National Natural Science Foundation of China under Grant 61702277, Grant 61872219, and Grant 61672276, in part by the Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD), and in part by the Jiangsu Collaborative Innovation Center on Atmospheric Environment and Equipment Technology (CICAEET). (Corresponding author: Lianyong Qi.) X. Xu and Q. Liu are with the School of Computer and Software, Jiangsu Engineering Center of Network Monitoring, Nanjing University of Information Science and Technology, Nanjing 210044, China (e-mail: [email protected]; [email protected]). X. Zhang is with the Department of Electrical and Computer Engineer- ing, The University of Auckland, Auckland 1010, New Zealand (e-mail: [email protected]). J. Zhang and W. Dou are with the State Key Laboratory for Novel Software Technology, Nanjing University, Nanjing 210008, China (e-mail: [email protected]; [email protected]). L. Qi is with the School of Information Science and Engineering, Qufu Normal University, Qufu 276826, China (e-mail: [email protected]). Digital Object Identifier 10.1109/TCSS.2019.2909137 are outsourced to the recruited human workers who regularly collect information, process data, and execute the transmitted tasks for requestors, thus increasing the service quality [1]–[4]. The crowdsourcing requestors usually divide the transmitted tasks into substantial atomic tasks which are offloaded to dif- ferent providers to make the large tasks be executed as quickly as possible [5]. Each provider has to sacrifice battery capacity for the implementation of the transmitted tasks, thus generating energy consumption, and therefore, the requestors are obliged to compensate for the providers in forms of monetary rewards, reputation approval, and so forth [2]. Technically, instead of depending on specific networks, crowdsourcing is fertilized by device-to-device (D2D) communication in a great deal. To maintain the installed applications running, mobile devices sense the surroundings and generate big data, which makes the volume of data traffic in the network witness great increase. The dense coverage of the cellular network and Wi-Fi support the ever-increasing data traffic to some extent. However, telecom operators charge a lot for the services of cellular networks, and the insufficient bandwidth fails to sustain the service requests with large data volume. On the other hand, the unstable radio intensity and infinite coverage of Wi-Fi mean that mobile users enjoy the Wi-Fi services only when they are within the scope [6]. Furthermore, with the increase in data traffic, wireless networks can hardly satisfy the mobile users’ quality of experience (QoE), thus asking for the improved mobile data traffic [7], [8]. As a novel and powerful technique for improving the network performance, D2D com- munication enables mobile users to exchange data directly via D2D links rather than the base station nearby [9]. Generally, the orthogonal and the nonorthogonal spectrum sharing are two kinds of D2D link access. The orthogonal spectrum enables the cellular and D2D pairs to employ different spectra. On the contrary, in the nonorthogonal spectrum, the D2D links share the cellular links. The D2D communication improves the throughput and shares the traffic volume of the core network [10]. Despite the efficiency of D2D communication, through direct communication, mobile users’ privacy might be leaked [11]. When a participant forwards a service request or offloads a task for execution to another participant, the former is ignorant of whether the one answering the request or executing the offloaded task is a normal participant or an attacker which tries to catch and tamper with the participants’ private information. Hence, in crowdsourcing, sound mobile authentication techniques are introduced for the D2D scheme, 2329-924X © 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

A Blockchain-Powered Crowdsourcing Method With Privacy ...static.tongtianta.site/paper_pdf/88073aae-e052-11e9-9b27-00163e08… · Xiaolong Xu , Qingxiang Liu, Xuyun Zhang , Jie Zhang,

  • Upload
    others

  • View
    1

  • Download
    0

Embed Size (px)

Citation preview

Page 1: A Blockchain-Powered Crowdsourcing Method With Privacy ...static.tongtianta.site/paper_pdf/88073aae-e052-11e9-9b27-00163e08… · Xiaolong Xu , Qingxiang Liu, Xuyun Zhang , Jie Zhang,

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

IEEE TRANSACTIONS ON COMPUTATIONAL SOCIAL SYSTEMS 1

A Blockchain-Powered Crowdsourcing MethodWith Privacy Preservation in Mobile Environment

Xiaolong Xu , Qingxiang Liu, Xuyun Zhang , Jie Zhang, Lianyong Qi , and Wanchun Dou

Abstract— Crowdsourcing is a booming technique that enablesparticipants to exchange data directly, thus making it possible toanswer latency-sensitive service requests and relieve the burdenof core networks. With some incentives, providers compete tofurnish service requests, thus pledging the quality of experi-ence (QoE) for requestors. However, the decentralized communi-cation in crowdsourcing increases the probability of informationtapering. Furthermore, providers’ arbitrary selection of therequests poses great threat to the efficient and profitable serviceprovision for the requestors. To deal with these challenges,we propose a blockchain-powered crowdsourcing method, namedBPCM, while considering the privacy preservation in mobileenvironment. Specifically, a mobile crowdsourcing frameworkbased on blockchain is designed first to preserve the privacy of theparticipants and keep the integrity of the service request and pro-vision. Then, density-based spatial clustering of applications withnoise (DBSCAN) and improved dynamic programming (IDP) areadopted to cluster the requestors and generate service strategies,respectively. Furthermore, simple additive weighting (SAW) andmultiple criteria decision making (MCDM) are utilized to selectthe optimal strategy that achieves the tradeoffs among maxi-mizing the service time, increasing the profits, and reducing theenergy consumption for the providers. Finally, comprehensiveexperiments are conducted to verify the accuracy and effective-ness of BPCM.

Index Terms— Blockchain, crowdsourcing, mobile environ-ment, privacy preservation.

I. INTRODUCTION

A. Background

CROWDSOURCING, as an efficient method in machinecomputation, provides such a scenario where the tasks

Manuscript received January 18, 2019; revised February 27, 2019 andMarch 22, 2019; accepted April 1, 2019. This work was supportedin part by the National Natural Science Foundation of China underGrant 61702277, Grant 61872219, and Grant 61672276, in part by thePriority Academic Program Development of Jiangsu Higher EducationInstitutions (PAPD), and in part by the Jiangsu Collaborative InnovationCenter on Atmospheric Environment and Equipment Technology (CICAEET).(Corresponding author: Lianyong Qi.)

X. Xu and Q. Liu are with the School of Computer and Software,Jiangsu Engineering Center of Network Monitoring, Nanjing Universityof Information Science and Technology, Nanjing 210044, China (e-mail:[email protected]; [email protected]).

X. Zhang is with the Department of Electrical and Computer Engineer-ing, The University of Auckland, Auckland 1010, New Zealand (e-mail:[email protected]).

J. Zhang and W. Dou are with the State Key Laboratory for NovelSoftware Technology, Nanjing University, Nanjing 210008, China (e-mail:[email protected]; [email protected]).

L. Qi is with the School of Information Science and Engineering, QufuNormal University, Qufu 276826, China (e-mail: [email protected]).

Digital Object Identifier 10.1109/TCSS.2019.2909137

are outsourced to the recruited human workers who regularlycollect information, process data, and execute the transmittedtasks for requestors, thus increasing the service quality [1]–[4].The crowdsourcing requestors usually divide the transmittedtasks into substantial atomic tasks which are offloaded to dif-ferent providers to make the large tasks be executed as quicklyas possible [5]. Each provider has to sacrifice battery capacityfor the implementation of the transmitted tasks, thus generatingenergy consumption, and therefore, the requestors are obligedto compensate for the providers in forms of monetary rewards,reputation approval, and so forth [2]. Technically, instead ofdepending on specific networks, crowdsourcing is fertilized bydevice-to-device (D2D) communication in a great deal.

To maintain the installed applications running, mobiledevices sense the surroundings and generate big data, whichmakes the volume of data traffic in the network witness greatincrease. The dense coverage of the cellular network andWi-Fi support the ever-increasing data traffic to some extent.However, telecom operators charge a lot for the servicesof cellular networks, and the insufficient bandwidth fails tosustain the service requests with large data volume. On theother hand, the unstable radio intensity and infinite coverageof Wi-Fi mean that mobile users enjoy the Wi-Fi services onlywhen they are within the scope [6]. Furthermore, with theincrease in data traffic, wireless networks can hardly satisfy themobile users’ quality of experience (QoE), thus asking for theimproved mobile data traffic [7], [8]. As a novel and powerfultechnique for improving the network performance, D2D com-munication enables mobile users to exchange data directly viaD2D links rather than the base station nearby [9]. Generally,the orthogonal and the nonorthogonal spectrum sharing aretwo kinds of D2D link access. The orthogonal spectrumenables the cellular and D2D pairs to employ different spectra.On the contrary, in the nonorthogonal spectrum, the D2D linksshare the cellular links. The D2D communication improvesthe throughput and shares the traffic volume of the corenetwork [10].

Despite the efficiency of D2D communication, throughdirect communication, mobile users’ privacy might beleaked [11]. When a participant forwards a service request oroffloads a task for execution to another participant, the formeris ignorant of whether the one answering the request orexecuting the offloaded task is a normal participant or anattacker which tries to catch and tamper with the participants’private information. Hence, in crowdsourcing, sound mobileauthentication techniques are introduced for the D2D scheme,

2329-924X © 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

Page 2: A Blockchain-Powered Crowdsourcing Method With Privacy ...static.tongtianta.site/paper_pdf/88073aae-e052-11e9-9b27-00163e08… · Xiaolong Xu , Qingxiang Liu, Xuyun Zhang , Jie Zhang,

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

2 IEEE TRANSACTIONS ON COMPUTATIONAL SOCIAL SYSTEMS

that is, each participant has attached the credits that are theevaluation of participant’s reputation, and visible for otherparticipants [12]. Each requestor prefers the service providerswith high credits. However, the credits are just the assessmenton service quality but not the guarantee that requestors’ requestinformation is not falsified. Thus, there remains a challengeon the safeguarding of privacy and ensuring a trust servicemechanism.

Blockchain, a decentralized and intelligent paradigm, is pro-posed to solve the privacy and data tempering problemsefficiently [13]. The data information is inspected by theminers, instead of being verified by the centralized controller.Then the information is encapsulated into a block, and theblocks are linked in the order of time [14]. What’s more,the information is encrypted using hash algorithms, includingthe hash value of the former block, the requestor informa-tion, the provider information, and so on. All participantsexcept the requestor and the provider compete for recordingthe services. A participant obtains the recording rights andtransmits the generated block to other participants to make surethe authenticity of the contained information, thus keeping thedata security [15]. As any changes in the blocks are verifiedby all participants, any tempering is bound to be recognized.

B. Motivation

To protect the participants’ privacy and data integrity,blockchain is applied in mobile crowdsourcing environment.Requestors transfer service descriptions, and with the incen-tives of rewards from requestors, idle participants contend toprovide corresponding services. However, due to the batterylimit of mobile terminals, if responding to requests coststoo much energy, the battery capacity might not sustain theparticipants to generate the blocks, which decreases the per-formance of the blockchain-powered mobile crowdsourcing.What’s more, when all of the providers respond to the samerequest, other requests might be shelved, affecting the QoEof requestors. Therefore, it remains a challenge to increasethe service time and reduce the energy consumption of theproviders. To address the challenge, a blockchain-poweredcrowdsourcing method (BPCM) with privacy preservation inmobile environment, named BPCM, is proposed.

C. Paper Contributions

In this paper, we make the following contributions.1) Propose a mobile crowdsourcing scheme powered by

blockchain to preserve the privacy as well as informationintegrity.

2) Adopt divide-and-conquer mechanism to cluster therequestors by density-based spatial clustering of appli-cations with noise (DBSCAN) and the requests in eachcluster are answered by a provider using improveddynamic programming (IDP) to generate the servicestrategies.

3) Evaluate the service strategies by normalizing the objec-tive functions using simple additive weighting (SAW)and selecting the optimal strategy based on multiplecriteria decision making (MCDM).

Fig. 1. Blockchain-powered mobile crowdsourcing framework with privacypreservation.

4) Conduct systematic experiment to validate the effective-ness and robustness of the proposed method BPCM.

The remainder of this paper is organized as follows.In Section II, we design the scheme of the mobile crowdsourc-ing powered by blockchain. Then, we formalize the servicetime, the profits, and the energy consumption of the providersin Section III. Furthermore, a BPCM is proposed in Section IV.Section V depicts the detailed experiments and evaluation.Section VI describes the related work, and we discuss theconclusion and future work in Section VII.

II. A BLOCKCHAIN-POWERED MOBILE CROWDSOURCING

FRAMEWORK WITH PRIVACY PRESERVATION

In this section, we propose a blockchain-powered mobilecrowdsourcing framework to preserve the participants’ privacyand data completeness. The framework is first elaborated andthen the mechanism of preserving privacy is designed.

A. Framework Establishment

To preserve the privacy of the crowdsourcing par-ticipants and maintain the data integrity, we design ablockchain-powered crowdsourcing framework in mobile envi-ronment as is illustrated in Fig. 1.

It is illustrated in Fig. 1 that there are N crowdsourcingrequestors and M providers. Let RR = {rr1, rr2, ..., rrN }be the set of the crowdsourcing requestors, where rri

(i = {1, 2, ..., N}) represents the i th crowdsourcing requestor.Let RP = {r p1, r p2, ..., r pM } be the crowdsourcing providerset, where r p j ( j = {1, 2, ..., M}) is the j th crowdsourcingprovider. rr2, r p j , and rrN manage a blockchain, respectively.r p1, r p j , and r pM are going to provide services for rr1, rr3,and rrN , respectively.

Definition 1 (Request specification of rri ): For rri , the spec-ification of the requestor is defined by a tuple with fiveattributes, denoted as rri = (< xri , yri >, Ai , tr s

i , tr fi ),

where < xri , yri > represents the location of rri , and Ai ,tr s

i , and tr fi represent the radius of valid range, the start time

of the request, and the finish time of the required services forthe rri , respectively.

Therefore, only when the provider in the valid range of therequestor, can the requestor is served. In addition, rri can beprovided with services in the time slot (tr s

i , tr fi ).

Definition 2 (Provision specification of r p j ): For r p j ,the provider is defined by a tuple of four attributes, denoted

Page 3: A Blockchain-Powered Crowdsourcing Method With Privacy ...static.tongtianta.site/paper_pdf/88073aae-e052-11e9-9b27-00163e08… · Xiaolong Xu , Qingxiang Liu, Xuyun Zhang , Jie Zhang,

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

XU et al.: BPCM WITH PRIVACY PRESERVATION IN MOBILE ENVIRONMENT 3

Fig. 2. Process of service request and provision with privacy preservation.

as r p j = (< x p j , yp j >, t psj , t p f

j ), where < x p j , yp j >,

t psj , and tp f

j represent the position, the start, and finish timeinstant of the j th requestor.

From Definition 2, r p j is enabled to provide services forthe requestors in the time duration of (t ps

j , t p fj ).

B. Privacy Protection Mechanism

In this section, we consider preserving the privacy ofparticipants as well as the data integrity from the perspectivesof crowdsourcing and blockchain.

The framework proposed earlier allows participants to regis-ter with fake information, that is, the identities of participantsin the framework do not represent their authentic information.Also, each participant has attached credits of his reputationassessing the service quality.

1) Crowdsourcing-Based Privacy Protection: We define theprocess of service and provision in the blockchain-basedmobile crowdsourcing environment. As is illustrated in Fig. 2,there are three situations. In the first situation, the requestorrejects service provision of the provider. After getting theidentity of the provider, the requestor knows the provider’sreputation. If the provider’s credits are low, the requestorrejects the service request or waits for other providers. Second,if the requestor is satisfied with the provider’s reputation,the location information is transmitted to the provider. Theprovider refuses to supply service, if the requestor is faraway. Third, both the requestor and the provider are pleasedwith each other. Then the provider responds to the requestorand provides service. The fake identities of the participantsand the multiverification have guaranteed the security of theparticipants’ real information.

2) Blockchain-Powered Privacy Protection: The partici-pants in the mobile crowdsourcing all act as the “miners.” Thatis, each participant maintains a blockchain that is accessible

Fig. 3. Example for block generation with four steps. (a) Step 1. (b) Step2. (c) Step 3. (d) Step 4.

to other participants. The blocks encapsulate the service infor-mation.

When a provider responds to a requestor, the provider infor-mation, the requestor information, and the task descriptionare broadcast in the mobile crowdsourcing environment. Otherparticipants contend to generate the head of blocks based onthe requestor and provider information, the task description,hash values of the last blocks in the blockchain they maintainand the current time instant. The winner participant whoobtains the recording right broadcasts the generated blockto other participants to make sure the authenticity of theinformation inside. Only when other participants have checkedthe block, can the winner participant add the block to theblockchain he/she maintains. The blocks are the reference forthe providers to make sure whether they are going to serve tosome certain requestors.

Fig. 3 illustrates an example of block generation.In Fig. 3(a), after r p1 catches the task description from rr2,the service record is forwarded, including the information ofr p1 as well as rr2 and the detailed task description. rr1, rr3,rr4, rr5, and r p2 competes for recording the service. Further-more, in Fig. 3(b), rr5 obtains the right and generates the blockcontaining the service information. Then, rr5 transmits theblock to other participants (i.e., rr1, r p2, rr3, and rr4) for theverification of the block shown in Fig. 3(c). If the participantsmake sure the authenticity of the block, rr5 will add the blockto the blockchain he/she sustains, as shown in Fig. 3(d).

While other participants compete to record the serviceand generate blocks, the provider travels to the requestor’slocation. Then the requestor transmits the task description tothe provider again. The provider visits the generated blockand checks whether the latter task description is authenticby comparing it with the former inside the block. If twotask descriptions are the same, then the provider suppliesservices for the requestor. In this way, the integrity of thetask description is preserved.

Page 4: A Blockchain-Powered Crowdsourcing Method With Privacy ...static.tongtianta.site/paper_pdf/88073aae-e052-11e9-9b27-00163e08… · Xiaolong Xu , Qingxiang Liu, Xuyun Zhang , Jie Zhang,

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

4 IEEE TRANSACTIONS ON COMPUTATIONAL SOCIAL SYSTEMS

TABLE I

KEY NOTATIONS AND REPRESENTATIONS

Since each block records the hash value of the former blockand the hash value of the whole blockchain is based on theblocks, thus once the data in the block are falsified, the hashvalues of all the latter blocks are changed and the blockchainis broken. Furthermore, any changes in the blocks should beadmitted by all of the participants. Therefore, it is impossiblefor the attackers to temper with the information. For example,in Fig. 1, if an attacker wants to falsify the information insidethe blocks r p j maintains, rri (∀i = {1, 2, ..., N}) and r pk

(∀k = {1, 2, ..., k − 1, k + 1, ..., M}) all need to allow themodification.

III. SYSTEM MODEL AND PROBLEM FORMULATION

In this phase, a system model is proposed to formalizethe service time, the profits and the energy consumption forthe providers. Key notations and representations are listedin Table I.

A. Service Time Analysis

In the mobile crowdsourcing environment, requestors for-ward service requests and providers respond to the requests.Let I j,i indicate whether r p j provides service for rri , whichis calculated as

I j,i =�

1, r p j provides service for rri

0, otherwise.(1)

Let st j,i and f t j,i be the actual time of r p j startingand finishing serving to rri , respectively. at j,i represents theservice time of r p j for rri which is calculated as

at j,i = f t j,i − st j,i . (2)

Let T be the service time of all the crowdsourcing providerswhich is calculated as

T =M�

j=1

N�i=1

I j,i · at j,i . (3)

B. Profit Analysis

There are incentives for crowdsourcing providers whenresponding to the requests, i.e., the payment from the corre-sponding requestors. When r p j provides services for rri , rri

needs to pay for the services. The profit relates to the datavolume and the service time.

Let v j,i be the data volume of services r p j provides for rri ,which is calculated as

v j,i =� f t j,i

st j,i

dr j (t)dt (4)

where dr j (t) denotes the processing rate of r p j at the timeinstant t .

PD represents the data transmission profits of providerswhich is calculated as

PD =M�

j=1

N�i=1

I j,i · ev j · v j,i (5)

where ev j denotes the profit factor based on the transmissionvolume.

Let PT be the service-providing profits of providers, whichis calculated as

PT =M�

j=1

N�i=1

I j,i · et j · at j,i (6)

where et j represents the payment factor on the time.The profits of the crowdsourcing providers, denoted as P ,

is calculated as

P = PD + PT. (7)

C. Energy Consumption Analysis

Crowdsourcing providers have to sacrifice the battery capac-ity for providing services, including the data processing andtransmission. The energy consumption of r p j for processingdata, denoted as E D j , is calculated as

E D j =N�

i=1

I j,i · (pdt j · at j,i + pdv j · v j,i ) (8)

where pdt j and pdv j represent the time-related and datavolume-related power consumption for r p j , respectively, whileprocessing data.

Let ETj be the generated energy while r p j transmits datato crowdsourcing requestors. ETj is calculated as

ETj =N�

i=1

I j,i · (ptt j · t t j,i + ptv j · v j,i ) (9)

where ptt j and ptv j denote the time-related and volume-related power consumption of r p j , respectively, while trans-mitting data for crowdsourcing requestors. t t j,i representsthe transmission time between r p j and rri . Since the highbandwidth of cellular networks and little distance from therequestors to the providers, t t j,i is negligible.

Thus, the consumed energy, denoted E , is calculated as

E =M�

j=1

(E D j + ETj ). (10)

Page 5: A Blockchain-Powered Crowdsourcing Method With Privacy ...static.tongtianta.site/paper_pdf/88073aae-e052-11e9-9b27-00163e08… · Xiaolong Xu , Qingxiang Liu, Xuyun Zhang , Jie Zhang,

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

XU et al.: BPCM WITH PRIVACY PRESERVATION IN MOBILE ENVIRONMENT 5

D. Problem Formulation

In this paper, we aim at optimizing the service time,increasing the profits, and reducing the energy consumptionfor the providers. The formalized problem is defined as

max T, max P, min E (11)

s.t.N�

i=1

B j,i(t) ≤ 1 (12)

B j,i (t) ≤ 1 − αi, j (t). (13)

In the multiobjective optimization problem, the constraintsare given in (12) and (13), respectively.

In this problem, B j,i(t) represents whether the j th crowd-sourcing provider r p j serves to the i th crowdsourcingrequestor rri at the time instant t , and B j,i(t) is calculated as

B j,i (t) =�

1, I j, i = 1

0, otherwise.(14)

Therefore, the constraint (12) describes that at a certain timeinstant, the crowdsourcing provider can only serve to at mostone crowdsourcing requestor.

Let α j,i (t) represent whether r p j is in the valid domainof rri and is calculated as

α j,i (t) =�

1, disi, j ≥ Ai

0, otherwise(15)

where disi, j denotes the distance from rri to r p j and iscalculated as

disi, j =�

(xri − x p j )2 + (yri − y p j )

2. (16)

The constraint (13) shows that a crowdsourcing providercannot serve for a crowdsourcing requestor unless it is in thevalid domain of rri .

IV. METHOD DESIGN

In this section, DBSCAN is employed to cluster therequestors for optimizing the performance of the crowdsourc-ing. What’s more, based on IDP, each provider conducts theselection of specific requestors to generate service solutions.Finally, SAW and MCDM are utilized to select the optimalsolution.

A. Requestor Clustering Using DBSCAN

In the phase, DBSCAN is used to cluster the requestors.Then each provider answers the service requests in a clusterto improve the quality of services for the requestors and thecrowdsourcing experience for the providers. As is depictedin (11), this paper aims to optimize the service time, increasethe profits, and reduce the energy consumption. Therefore,if a provider is far away from a requestor, and the requesttime slot of the requestor is short, that is, the provider justsupplies short-time services to the requestor, for much timeis spent on the travel. The provider may not need to respondto such requestors. DBSCAN is an efficient algorithm thathelps cluster closely packed points and mark out the outliers

Fig. 4. Clustering example with 17 requestors and 2 providers.

in low-density regions. Hence, DBSCAN is applied to clusterthe requestors.

In DBSCAN, � represents the distance threshold of therequestors If the distance from rri to rr j is less than �, then rr j

is in the �-range of rri , denoted as (rri , �). Min Pts notes thethreshold of the number of the requestors in the �-range of arequestor. Therefore, (�, Min Pts) describes the compactnessof the requestors.

Fig. 4 illustrates a clustering example of 17 requestorsand 2 providers in the mobile crowdsourcing environment.The requestors are scattered irregularly in the scope and theparameter Min Pts is set as 4. After DBSCAN is utilized tocluster the requestors, the requestors are divided into threetypes: noise requestors, border requestors, and core requestors.For a requestor rri , if the number of the requestors in (rri , �) isover or equal to Min Pts, then rri is a core requestor. If therequestor rrk is not a core requestor, but it is included in(rri , �), then rrk is a border requestor. If a requestor is neithera core requestor nor a border requestor, it is a noise requestor.

Fig. 4 depicts the 14 of 17 requestors are divided into2 clusters, and the 3 noise requestors are excluded from theclusters. Each provider answers the requestors of a cluster anddismisses the noise requestors.

Algorithm 1 elaborates how to cluster the requestorsbased on DBSCAN. The algorithm inputs �, Min Pts, andRR and outputs the cluster set, denoted as RC . For eachunchecked requestor, the number of requestors in its �-rangeis evaluated to determine whether it is a noise requestor(Lines 5–11). If the requestor is a core requestor, then therequestor is added into the requestor cluster RC and so arethe requestors in its �-range (Lines 12–15). Furthermore,the noise requestors are assessed whether they are borderrequestors, if not, the requestors in their �-range are addedinto RC (Lines 16–31). The process is continued until all ofthe requestors are checked and the requestor cluster RC isoutput (Line 34).

B. Request Selection Based on IDP

In the above, the requestors are divided into severalclusters using DBSCAN, and in this section, whether aprovider answers a certain request is considered. Each provideris responsible for a cluster and provides services for therequestors of the cluster. Each requestor forwards the servicerequest, containing the start time and finish time of therequested services. For the providers, to maximize the servicetime, IDP is applied to generate the service strategies.

Page 6: A Blockchain-Powered Crowdsourcing Method With Privacy ...static.tongtianta.site/paper_pdf/88073aae-e052-11e9-9b27-00163e08… · Xiaolong Xu , Qingxiang Liu, Xuyun Zhang , Jie Zhang,

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

6 IEEE TRANSACTIONS ON COMPUTATIONAL SOCIAL SYSTEMS

Algorithm 1 Clustering Requestors Using DBSCANRequire: �, Min Pts, RREnsure: RC1: Set all the requestors unchecked2: for i = 1 to N do3: Mark rri checked4: n=05: for rr j in RR do6: if rr j in (rri , �) then7: n=n+18: end if9: end for

10: if n < Min Pts then11: Mark rri as noise requestor12: else13: RC = RC + {nextcluster}14: RC = RC + {rri }15: RR = RR − {rri }16: for rrk in (rri , �) do17: RC = RC + {rrk}18: RR = RR − {rrk}19: for rr j in RR do20: n=021: if rr j in (rrk, �) then22: n = n + 123: end if24: end for25: if n ≤ Min Pts then26: for rr j in (rrk, �) do27: RC = RC + {rr j }28: RR = RR − {rr j }29: end for30: end if31: end for32: end if33: end for34: ReturnRC

We first sort the service requests by the finish time inthe ascending order first. Then the actual response time ofeach requestor is determined. Finally, the service strategiesare generated by IDP.

1) Service Request Sorting: Considering that there areseveral requestors in a cluster, therefore, when a requestorforwards a service request, the provider might serve to anotherrequestor, which makes it difficult for the provider to deter-mine the request-answer order to prolong the service time.In this section, the service requests are sorted according tothe finish time instant in the ascending order. The provideranswers the service request in the order.

Fig. 5 shows an example of request sorting with fourrequestors. The first request starts from 1 min and finishesat 3 min and rr2, rr3, and rr4 request service from 6, 2, and10 min to 9, 5, and 14 min, respectively. To maximize theservice time of the provider r p1, we sort the requests by thefinish time instant and r p1 provides services in the order ofrr1, rr3, rr2, and rr4.

Fig. 5. Example of request sorting with four requestors.

2) Response Time Reconfirmation: In this section, whenthe provider responds to the requests is reconfirmed. Fig. 5illustrates an example of request sorting with four requestors.It is shown that r p1 provides service for rr1, rr3, rr2, and rr4sequentially. The service for rr1 finishes at 3 min, but theservice request of rr3 starts at 2 min. Therefore, after sortingthe requests by the finish time instant, it remains a challengeto determine when the provider starts to serve to a certainrequestor.

The response to rri is conducted after the requestors whoserequested services finish earlier. RFj represents the set ofrequestors responded before rr j . If rri ∈ RFj , mti, j (i �= j)represents the transferring time of r pk from rri to rr j .In addition, the actual finish time of each service f t j,k isequal to the requested time tr f

j . The actual start time st j,k

is calculated as

st j,k = max(tr sj , f ti,k + mti, j )(rri ∈ RFj ). (17)

Fig. 6 depicts the specific response time of the 4 shownin Fig. 5. In Fig. 6(a), r p1 spends 1 min transferring to rr1,and therefore, the start time of rr1, st1,1 is 2 min. Then,in Fig. 6(b), after r p1 finishes serving for rr1, r p1 spends1 min transferring to rr3. Therefore, st3,1= 4 min. At 5 min,the services of rr3 are finished, and r p1 travels to rr2. Thetraveling time is 1 min and, therefore, st2,1= 6 min in Fig. 6(c).Furthermore, in Fig. 6(d), r p1 spends 2 min transferring fromrr2 to rr4, and st4,1= 11 min.

Algorithm 2 describes the process of determining thespecific response time. The algorithm inputs the requestorcluster of the t th provider r pt and outputs the reachabilitymatrix, denoted as RE . First, for each requestor in therequestor cluster, we find out the requestors which can reachit (Lines 3–8). Then the actual start time is calculated basedon (17) (Lines 10–17). The process is continued until theservice time of the requestors has been determined.

3) Service Strategy Generation: After sorting the requestsand determining the service time of each requestor, in thissection, we apply IDP to generate the service strategies. Likethe normal dynamic programming, when evaluating the servicestrategies, we record the service time of the provider r p j

for the i th requestor at the t time instant by a binary arrayET [i ][t]. ET is updated as

ET [i ][t] =⎧⎨⎩

0, i = 0

maxrrk∈RE j

{γ jk,i , ET [k][t]}, i > 0 (18)

Page 7: A Blockchain-Powered Crowdsourcing Method With Privacy ...static.tongtianta.site/paper_pdf/88073aae-e052-11e9-9b27-00163e08… · Xiaolong Xu , Qingxiang Liu, Xuyun Zhang , Jie Zhang,

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

XU et al.: BPCM WITH PRIVACY PRESERVATION IN MOBILE ENVIRONMENT 7

Fig. 6. Response procedure of the four requestors shown in Fig. 5.(a) Step 1. (b) Step 2. (c) Step 3. (d) Step 4.

Algorithm 2 Reconfirmation of Response TimeRequire: RCt

Ensure: RE1: RE = �2: for rri in RCt do3: for rr j in RFi do4: Calculate mt j,i

5: if mt j,i+tr fj < tr f

i then6: REi = REi + {rr j }7: end if8: end for9: sti,t = tr s

i10: for k = 1 to i do11: if rrk ∈ REi then12: temp = tr j

k + mtk,i

13: if temp > sti,t then14: sti,t = temp15: end if16: end if17: end for18: end for19: return RE

where γj

k,i is calculated as

γj

k,i = ET [k][t − mtk,i ] + atk, j . (19)

Based on (18), we can calculate the service time of theprovider at a certain time instant for any requestor. Binaryarrays are also utilized to record the profits and the energyconsumption, respectively. To get the maximum provider ser-vice time, we just need to search ET . However, this paper aimsto optimize the service time as well as the profits but minimizethe energy consumption. It goes beyond doubt that, whenthe providers’ service time reaches the maximum, the energyconsumption also arrives at the maximum. Hence, the total

Algorithm 3 Service Strategies Using IDPRequire: r p j , RC j , REEnsure: ET1: for i = 1 to | RC j | do2: for i = 1 to | t p f

j | do3: M AX=04: for k = 1 to | REi | do5: temp1 = ET [k][t]6: temp2 = ET [k][t − mtk,i ] + f tk, j − stk, j7: if M AX < temp1 and M AX < temp2 then8: if temp1 ≥ temp2 then9: M AX = temp1

10: else11: M AX = temp212: end if13: end if14: end for15: end for16: end for17: return ET

performance may not be the optimal, when the provider servicetime reaches the maximum.

Therefore, instead of selecting the strategy with the maxi-mum service time, the first n strategies with the largest servicetime are selected.

Algorithm 3 describes the procedure of generating ET .r p j , RC j , and RE are input and the algorithm outputs ET .The reachable requestors of each requestor are consideredto ensure the maximum service time at any time instant ischosen (Lines 1–10). The process is continued until all of therequestors have been checked and ET is output (Line 15).

C. Strategy Evaluation Using SAW and MCDM

The proposed method aims at achieving tradeoffs amongmaximizing the providers’ service time, improving the profits,and reducing the energy consumption. In the above, we selectthe fisrt n strategies with top service time. To select therelatively optimal strategy, SAW and MCDM are utilized toevaluate the strategies.

The longer the service time, the more efficient the strategyis, so the service time is a positive criterion. The service timeis normalized as

V (T ) =�

T −TminTmax−Tmin

, Tmax − Tmin �= 0

1, Tmax − Tmin = 0(20)

where Tmax represents the maximum service time of theproviders in the n strategies and Tmin represents the minimumone. Similarly, the profit is also a positive criterion, which isnormalized as

V (P) =�

P−PminPmax−Pmin

, Pmax − Pmin �= 0

1, Pmax − Pmin = 0(21)

where Pmax and Pmin stand for the maximum and minimumprofit of the n strategies, respectively.

Page 8: A Blockchain-Powered Crowdsourcing Method With Privacy ...static.tongtianta.site/paper_pdf/88073aae-e052-11e9-9b27-00163e08… · Xiaolong Xu , Qingxiang Liu, Xuyun Zhang , Jie Zhang,

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

8 IEEE TRANSACTIONS ON COMPUTATIONAL SOCIAL SYSTEMS

Different from the service time and the profit, we aim atreducing the energy consumption. Hence, the energy consump-tion is a negative criterion, which is normalized as

V (E) =�

Emax−EEmax−Emin

, Emax − Emin �= 0

1, Emax − Emin = 0(22)

where the maximum and minimum energy consumptions aredenoted as Emax and Emin, respectively.

After normalizing the three objective functions, we conductan overall consideration of the three objectives. The utilityvalue of each strategy is calculated as

V (S) = ω1 · V (T ) + ω2 · V (P) + ω3 · V (E) (23)

where V (S) represents the utility value of a strategy and ω1,ω2, and ω3 describe the weights of V (T ), V (P), and V (E),respectively.

Therefore, the problem with three objectives is transformedinto

max V (S) (24)

ω1 + ω2 + ω3 = 1. (25)

D. Method Overview

This paper targets at optimizing the service time, increas-ing the profit, and reducing the energy consumption for theproviders. First, we cluster the requestors based on DBSCANto exclude the noise requestors and regularize the irregularrequestors. Then each provider is responsible for a cluster.Furthermore, IDP is applied to generate the service strategies.Finally, SAW and MCDM are utilized to evaluate the servicestrategies and help select the optimal one.

Algorithm 4 elaborates the process of the proposed BPCM.We input the requestor set RR as well as the provider set RPand the algorithm outputs the best strategy, denoted as BS.First, the requestors are clustered based on Algorithm 1, andin this process, the noise requestors are excluded from theservice range (Lines 1–3). Then, the requestors in each clusterare sorted according to the requested service finish time in theascending order (Lines 4–6). Furthermore, the actual start timeof each requestor is determined based on Algorithm 2 (Line 7).In addition, IDP is utilized to generate the service strategiesby Algorithm 3 (Line 8). Finally, the objective functions arenormalized and the strategies are evaluated based on SAW andMCDM (Lines 9 and 10).

V. EXPERIMENTAL EVALUATION

In this section, a series of experiments is conducted tovalidate the performance of the proposed method BPCM. Firstof all, the setup of the experimental environment is describedin detail, including the parameter settings and the introductionof the comparison methods. Then, the utility values of thesolutions generated by BPCM are evaluated to select outthe most optimal solution at each provider–requestor scale.Finally, the effects of various provider–requestor scales on theperformance of the service time, the profits, and the energyconsumption performed by these crowdsourcing methods areevaluated and compared.

Algorithm 4 BPCMRequire: RR, RPEnsure: BS1: for rri in RR do2: Cluster rri by Algorithm 13: end for4: for requestors in each cluster do5: Sort the requestors6: end for7: Determine the actual start time by Algorithm 28: Apply IDP to generate service strategies by Algorithm 39: Normalize T, P and E by (19 to 21)

10: Evaluate the strategies by (22 and 23)11: return BS

TABLE II

PARAMETER SETTINGS

A. Simulation Setup

In our simulation, there are multiple requestors who shouldbe clustered according to the number of the providers. Hence,we engaged four data sets of provider–requestor scales in ourexperiments (located at https://share.weiyun.com/5oIIV6C),and the number of the requestors is set to 20, 40, 80, and 160,whereas the number of the providers is set to 3, 6, 12, and 18.The number of strategies n is equal to 5. The parameters andthe corresponding values are specified in Table II.

In order to conduct the comparison analysis, three crowd-sourcing methods are adopted besides BPCM. The comparisonmethods are introduced as follows.

Dynamic Programming Based on Voronoi (VNDP): Therequestors are divided into groups, each of which has onecrowdsourcing service provider according to the Voronoimethod [16]. In each group, the normal dynamic pro-gramming is used to decide whether a provider respondsto a certain requestor. This method will repeat until allrequestors in every group are checked and the generatedsolution with the longest service time is selected as theoptimal strategy.Dynamic Programming Based on DBSCAN (DBDP): Inthis method, the requestors are clustered by the numberof providers according to the DBSCAN. In order toprovide the crowdsourcing services, the service strate-gies are generated using the normal dynamic program-ming. Then the strategy with the longest service time isselected as the optimal one.

Page 9: A Blockchain-Powered Crowdsourcing Method With Privacy ...static.tongtianta.site/paper_pdf/88073aae-e052-11e9-9b27-00163e08… · Xiaolong Xu , Qingxiang Liu, Xuyun Zhang , Jie Zhang,

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

XU et al.: BPCM WITH PRIVACY PRESERVATION IN MOBILE ENVIRONMENT 9

Fig. 7. Utility values performed by BPCM at different provider–requestorscales. (a) Provider–requestor scale is 3–20. (b) Provider–requestor scaleis 6–40. (c) Provider–requestor scale is 12–80. (d) Provider–requestor scaleis 18–160.

IDP Based on Voronoi (VIDP): Under the condition ofthis method, the requestors are grouped by the Voronoimethod. Then, the dynamic programming was engagedto figure out if the requestor is able to be served to.It repeats until all requestors are checked, respectively,and the first five solutions with the largest service timeare selected. Finally, the solution with the highest utilityvalue is evaluated as the optimal strategy.

These crowdsourcing methods are implemented on a desk-top PC with the Inter Core i7-4720 3.60-GHz processors and4-GB RAM. The corresponding assessment results will bedescribed in detail in Sections V-B and V-C.

B. Solution Election of BPCM

According to Section IV, BPCM generates a series ofsolutions and SAW as well as MCDM is employed to selectthe most optimal solution in our experiments. Fig. 7 showsthe utility values of the solutions generated by BPCM on dif-ferent provider–requestor scales. After statistics and analysis,the solution with the maximum utility value is employed as themost optimized solution. In Fig. 7, the most optimized solu-tions are solution 1, 2, 3, and 1 when the providers–requestorscale is 3–20, 6–40, 12–80, and 18–160, respectively.

C. Comparison Analysis

In this section, the performances of VNDP, DBDP, VIDP,and BPCM are evaluated and compared in detail. The servicetime, the profits, and the energy consumption are the mainmetrics to evaluate the performance of the crowdsourcingmethods. In addition, as the energy consumption is looselyrelated to the service time and the profits according toSection III, the relation between the energy consumption andthe service time along with the relevance between the energyconsumption and profits should be figured out. The evaluationresults are shown in Figs. 8–12, respectively.

Fig. 8. Comparison of the service time by VNDP, DBDP, VIDP, and BPCMat different provider–requestor scales.

Fig. 9. Comparison of the profits by VNDP, DBDP, VIDP, and BPCM atdifferent provider–requestor scales.

1) Comparison on Service Time: According to Section III,the service time describes how long the providers supply ser-vices for the requestors. From the perspective of the providers,longer service time means more rewards they obtain and themethod performs better. Fig. 8 shows the comparison of theservice time of the four different crowdsourcing methods.It can be indicated that when the provider–requestor scaleis small, BPCM and VIDP are close and both better thanVNDP and DBDP. When the scale is enlarged, the advan-tage of BPCM compared with VIDP is apparent. With theincrease in provider–requestor scale, the gap of the servicetime among BPCM and the comparative methods expands,which may represent our proposed method has a better per-formance on the condition of large quantities of crowdsourcingparticipants.

2) Comparison on Profits: The aim of the providers tosupply the crowdsourcing services is to earn the profits. Hence,the crowdsourcing method that brings more profits should beconsidered as the more efficient methods. Fig. 9 illustratesthe comparison of the profits generated by the comparativemethods and BPCM. It is intuitive that when the numberof the providers is 18 and the number of the requestors is160, the profits of BPCM are more than 450$ and betterthan any other crowdsourcing methods. However, when theprovider–requestor scale is not as large as 18–160, the differ-ence between BPCM and other crowdsourcing method is notobvious enough. The reason may be that the profits earnedby the providers of the crowdsourcing service depends on theservice time. Therefore, the proposed method BPCM may besuitable for large groups of the providers and the requestors.

Page 10: A Blockchain-Powered Crowdsourcing Method With Privacy ...static.tongtianta.site/paper_pdf/88073aae-e052-11e9-9b27-00163e08… · Xiaolong Xu , Qingxiang Liu, Xuyun Zhang , Jie Zhang,

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

10 IEEE TRANSACTIONS ON COMPUTATIONAL SOCIAL SYSTEMS

Fig. 10. Comparison of the energy consumption by VNDP, DBDP, VIDP,and BPCM at different provider–requestor scales.

3) Comparison on Energy Consumption: When the crowd-sourcing service provider supplies services to crowdsourcingrequestors, the energy is consumed for processing data andtransmitting data. From the perspective of the providers,the energy consumption is considered as a kind of cost.Therefore, more profits the providers earn with less energyconsumption they generate make more interests of the crowd-sourcing services. Fig. 10 shows the comparison of the energyconsumption by those crowdsourcing methods. In Fig. 10,the energy consumption of the proposed method BPCM is alittle more than other method because the energy consumptionis closely related to the service time. When the service timeincreases, the energy consumption is hard to reduce. However,BPCM does not surpass the other three methods too much,which is acceptable.

4) Comparison on Energy Consumption Per Minute: Fromthe perspective of the providers, more service time means moreprofits. However, the energy consumption increases with theservice time, which results in a conflict between the servicetime and the energy consumption. Therefore, the relationbetween the service time and the energy consumption shouldbe figured out to compare the ratio of the energy consumptionand the service time. Fig. 11 depicts the comparison of theenergy consumption per minute by VNDP, DBDP, VIDP, andBPCM at different provider–requestor scales. It shows thatwhen the number of the provider is 3 and the number ofthe requestor is 20, the energy consumption per minute ofthe proposed method BPCM is little more than VIDP butmuch better than VNDP and DBDP. In addition, with theincrease in the requestors and the providers, BPCM is betterthan any other comparative methods. According to Section III,only the energy consumption in the processing data is relatedto the service time. This means our proposed method costsless energy consumption per minute with the increase in theprovider–requestor scales.

5) Comparison on Energy Consumption Per Dollar: Thecrowdsourcing providers aim at earning the interests from thesupplements. After the services are provided, the providerssacrifice the battery capacity for the profits from the requestorsand generate some energy consumption. In this situation, moreprofits with less energy consumption mean more interestsfor the providers. However, the profit along with the energyconsumption increases along with the service time length-ening, which makes it hard to get high profits with low

Fig. 11. Comparison of the energy consumption per minute by VNDP, DBDP,VIDP, and BPCM at different provider–requestor scales.

Fig. 12. Comparison of the energy consumption per dollar by VNDP, DBDP,VIDP, and BPCM at different provider–requestor scales.

energy consumption. Hence, the energy consumption per dol-lar of the four crowdsourcing methods are compared in Fig. 12.It is intuitive that BPCM consumes less energy than other com-parative methods when each dollar is earned. Moreover, whenthe requestors and the providers become more, the energyconsumption of each dollar by the four crowdsourcing methodstends to be stable. This means that our proposed crowdsourc-ing method BPCM is able to bring the service providers moreinterests and be more environment-friendly than the othercrowdsourcing methods.

VI. RELATED WORK

In recent years, we have witnessed the widespread develop-ment of the crowdsourcing, which has undergone an enor-mous revolution as a new computing paradigm. Moreover,the blockchain, a novel and efficient technology to solve theprivacy problem, has been applied in many applications [17].

Mobile crowdsourcing is an important technology for gath-ering the data of various types such as collecting loca-tion related data for many applications [18]. In addition,the mobile crowdsourcing is a novel paradigm where workersare employed to solve the computation tasks that softwarecannot easily exceed [19].

Gupta et al. [20] developed a new mobile crowdsourcingplatform mClerk that is available to be deployed in the devel-oping regions. In the designed mClerk platform, everyone witha low-end mobile device can access and utilize mClerk sinceit receives and sends tasks via short message service (SMS).In addition, the mClerk enables the distribution of graph-ical tasks by leveraging a protocol to send small imagesvia SMS. Duan et al. [21] investigated the task allocation

Page 11: A Blockchain-Powered Crowdsourcing Method With Privacy ...static.tongtianta.site/paper_pdf/88073aae-e052-11e9-9b27-00163e08… · Xiaolong Xu , Qingxiang Liu, Xuyun Zhang , Jie Zhang,

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

XU et al.: BPCM WITH PRIVACY PRESERVATION IN MOBILE ENVIRONMENT 11

and price determine issues in multicircuit substation (MCS)and they introduced a VickreyClarkeGroves-based auctionmechanism and a suboptimal auction mechanism for thecontinuous working pattern and the discontinuous workingpattern, respectively. Zhang et al. [22] designed a participantcoordination framework. In this framework, if the systemserver does not know the trajectories of participants, the frame-work allows the system server to provide optimal quality ofinformation (QoI) for sensing tasks. In order to protect theprivacy of participants and ensure the QoI of collected data,a punishment mechanism, a cooperative data aggregation, andan incentive distribution method are proposed in their paper.Miao et al. [23] proposed a budget-aware task alloca-tion approach for spatial crowdsourcing, aiming to helpthe requestors make key decisions about task allocation.Sun et al. [24] presented a privacy-preserving object-findingand secure system called SecureFind via mobile crowdsourc-ing. In this system, one unique Bluetooth is attached toeach valuable object, and the owner who loses the objectsubmits request for finding the object to other mobile users.The SecureFind ensures excellent object security so that onlythe owner of the object can obtain the location of the lostobject. In addition, Shah-Mansouri et al. [25] proposed aProfit Maximizing Truthful auction mechanism for the mobilecrowdsourcing systems (ProMoT) based on the greedy mech-anism. The goal of ProMoT is to increase the profit of theplatform while offering satisfying rewards to the smartphoneusers.

With the rapid development of mobile and communicationtechnologies, the mobile devices are equipped with powerfulprocessors, fast wireless communication modules, large mem-ories, and so on [26]. Normally, the mobile devices such astablets and smartphones are carried by human beings. There-fore, the mobile devices are applicable in the crowdsourcingenvironment to solve the large-scale data today [27].

Zhao et al. [28] introduced a novel trustworthy devicepairing (TDP) scheme that achieves reliable device selectionsfor the trustworthy D2D communications and user-transparentsecure D2D connections. In order to improve the energyefficiency for the mobile devices, Wang et al. [29] developedan iterative combinatorial auction algorithm correspondingly,where the cellular network is considered as an auctioneer,and the D2D users are regarded as bidders. Zhang et al. [30]designed an effective social-aware approach to optimize D2Dcommunication by taking advantages of the physical wirelessnetwork layer and the social network layer. In addition, basedon so-called Indian Buffet Process, they announce to simulatethe content distribution in the online networks. Xu et al. [31]coped with the delivering content to optimize the peer dis-covery as well as the resource allocation by integrating thephysical and social layer information. Subsequently, on con-dition that the quality of the service requirements for theD2D links and cellular should be guaranteed, a 3-D iterativematching algorithm is proposed to increase the rate of D2Dpairs which is weighted by the strength of social relationships.Han et al. [32] proposed an algorithm based on the reducedsolution space to minimum the cost of crowdsourcing whilesatisfying the coverage probability.

Recently, blockchain, the backbone technology of Bitcoin,has become an efficiency technology for the data securityproblems [33]. In addition, the blockchain has been applied tomany applications successfully such as IoT, Ethereum, and soon [34], [35]. Using the blockchain technology, high securityand transparency can be achieved in the system [36].

In order to support the mobile blockchain, Jiao et al. [37]considered to deploy the edge computing service. An auction-based edge computing resource market is proposed for theedge computing service provider. Zyskind et al. [38] imple-mented a protocol to turn the blockchain into an automatedaccess-control manager which did not need the trust from athird party. Xiong et al. [39] proposed a lightweight infrastruc-ture of the proof for the work-based blockchains, and thecomputation-intensive part during the process of the consensusis offloaded to the cloud platform. Kim et al. [40] inves-tigated a secure authentication management human-centricscheme (SAMS), aiming to attest the mobile devices byemploying the blockchain for trusting the resource informa-tion from the mobile devices participated in the multiplereaction monitoring (MRM) resource pool. Lee et al. [41]proposed a new firmware update scheme that takes advan-tage of the blockchain technology to download the latestfirmware for embedded devices and ensure the security ofa firmware version and the correctness of the firmware.Xia et al. [42] proposed a system MeDShare based on theblockchain to address the problem of sharing medical dataamong the medical big data custodians in the few-trustenvironment.

To the best of our knowledge, few existing works havefocused on the investigation of the blockchain-based crowd-sourcing problem in mobile environment, which aims at max-imizing the service time, increasing the profits, and reducingthe energy consumption, while preserving the privacy of theparticipants and data integrity. For this purpose, a BPCMwith privacy preservation in mobile environment is designedaccordingly.

VII. CONCLUSION

In the past few years, the crowdsourcing is being proposedas a prevalent technology to guarantee the direct communica-tion between the participants. Based on the mechanism of thecrowdsourcing, the throughput of the network is expanded andthe service performance of the crowdsourcing is guaranteed.To preserve the participants’ privacy and data completenessin crowdsourcing, blockchain is applied to the mobile crowd-sourcing environment. In order to maximize the service time,increase the profits, and reduce the energy consumption forthe providers, a blockchain-powered crowdsourcing method,named BPCM, is proposed in this paper. First, we designthe mobile crowdsourcing paradigm based on the blockchain.Then DBSCAN and IDP are utilized to generate the servicestrategies. Furthermore, SAW and MCDM are employed toevaluate the service strategies and select the optimal one.Finally, the comprehensive experiments and evaluations areconducted to validate the high performance of the proposedmethod.

Page 12: A Blockchain-Powered Crowdsourcing Method With Privacy ...static.tongtianta.site/paper_pdf/88073aae-e052-11e9-9b27-00163e08… · Xiaolong Xu , Qingxiang Liu, Xuyun Zhang , Jie Zhang,

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

12 IEEE TRANSACTIONS ON COMPUTATIONAL SOCIAL SYSTEMS

In the future work, the proposed method will be adjustedand expanded according to the real-world scenario. On theother hand, the blockchain-powered mobile crowdsourcingwill consult the actual blockchain and crowdsourcing to makethe proposed more precise.

REFERENCES

[1] K. Mao, L. Capra, M. Harman, and Y. Jia, “A survey of the useof crowdsourcing in software engineering,” J. Syst. Softw., vol. 126,pp. 57–84, Apr. 2017. doi: 10.1016/j.jss.2016.09.015.

[2] Y. Zhang, C. Jiang, L. Song, M. Pan, Z. Dawy, and Z. Han, “Incentivemechanism for mobile crowdsourcing using an optimized tournamentmodel,” IEEE J. Sel. Areas Commun., vol. 35, no. 4, pp. 880–892,Apr. 2017.

[3] H. Garcia-Molina, M. Joglekar, A. Marcus, A. Parameswaran, andV. Verroios, “Challenges in data crowdsourcing,” IEEE Trans. Knowl.Data Eng., vol. 28, no. 4, pp. 901–911, Apr. 2016.

[4] B. Morschheuser, J. Hamari, J. Koivisto, and A. Maedche,“Gamified crowdsourcing: Conceptualization, literature review, andfuture agenda,” Int. J. Hum.-Comput. Stud., vol. 106, pp. 26–43,Oct. 2017.

[5] Y. Tong, L. Chen, Z. Zhou, H. V. Jagadish, L. Shou, and W. Lv,“SLADE: A smart large-scale task decomposer in crowdsourcing,”IEEE Trans. Knowl. Data Eng., vol. 30, no. 8, pp. 1588–1601,Aug. 2018.

[6] J. Wu, B. Cheng, C. Yuen, Y. Shang, and J. Chen, “Distortion-awareconcurrent multipath transfer for mobile video streaming in heteroge-neous wireless networks,” IEEE Trans. Mobile Comput., vol. 14, no. 4,pp. 688–701, Apr. 2015.

[7] G. Ma, Z. Wang, M. Zhang, J. Ye, M. Chen, and W. Zhu, “Under-standing performance of edge content caching for mobile video stream-ing,” IEEE J. Sel. Areas Commun., vol. 35, no. 5, pp. 1076–1089,May 2017.

[8] P. Zhao, W. Yu, X. Yang, D. Meng, and L. Wang, “Buffer data-drivenadaptation of mobile video streaming over heterogeneous wirelessnetworks,” IEEE Internet Things J., vol. 5, no. 5, pp. 3430–3441,Oct. 2018.

[9] H. H. Yang, J. Lee, and T. Q. Quek, “Heterogeneous cellular networkwith energy harvesting-based D2D communication,” IEEE Trans. Wire-less Commun., vol. 15, no. 2, pp. 1406–1419, Feb. 2016.

[10] T. D. Hoang, L. B. Le, and T. Le-Ngoc, “Energy-efficient resourceallocation for D2D communications in cellular networks,” IEEE Trans.Veh. Technol., vol. 65, no. 9, pp. 6972–6986, Sep. 2016.

[11] M. Wang and Z. Yan, “A survey on security in D2D communications,”Mobile Netw. Appl., vol. 22, no. 2, pp. 195–208, Apr. 2017.

[12] J. Sun, R. Zhang, and Y. Zhang, “Privacy-preserving spatiotemporalmatching for secure device-to-device communications,” IEEE InternetThings J., vol. 3, no. 6, pp. 1048–1060, Dec. 2016.

[13] J. Kang et al., “Blockchain for secure and efficient data sharing invehicular edge computing and networks,” IEEE Internet Things J., to bepublished.

[14] A. Kosba, A. Miller, E. Shi, Z. Wen, and C. Papamanthou, “Hawk:The blockchain model of cryptography and privacy-preserving smartcontracts,” in Proc. IEEE Symp. Secur. Privacy (SP), May 2016,pp. 839–858.

[15] M. Li et al., “Crowdbc: A blockchain-based decentralized frame-work for crowdsourcing,” IEEE Trans. Parallel Distrib. Syst., to bepublished.

[16] Y. Han, T. Luo, D. Li, and H. Wu, “Competition-based participantrecruitment for delay-sensitive crowdsourcing applications in D2D net-works,” IEEE Trans. Mobile Comput., vol. 15, no. 12, pp. 2987–2999,Dec. 2016.

[17] X. Liang, S. Shetty, D. Tosh, C. Kamhoua, K. Kwiat, and L.Njilla, “Provchain: A blockchain-based data provenance architecturein cloud environment with enhanced privacy and availability,” inProc. 17th IEEE/ACM Int. Symp. Cluster, Cloud Grid Comput., 2017,pp. 468–477.

[18] Y. Wang, X. Jia, Q. Jin, and J. Ma, “Mobile crowdsourcing: Framework,challenges, and solutions,” Concurrency Comput., Pract. Exper., vol. 29,no. 3, 2017, Art. no. e3789.

[19] K. Yang, K. Zhang, J. Ren, and X. Shen, “Security and privacy in mobilecrowdsourcing networks: Challenges and opportunities,” IEEE Commun.Mag., vol. 53, no. 8, pp. 75–81, Aug. 2015.

[20] A. Gupta, W. Thies, E. Cutrell, and R. Balakrishnan, “mClerk: Enablingmobile crowdsourcing in developing regions,” in Proc. SIGCHI Conf.Hum. Factors Comput. Syst., 2012, pp. 1843–1852.

[21] Z. Duan, M. Yan, Z. Cai, X. Wang, M. Han, and Y. Li, “Truthful incen-tive mechanisms for social cost minimization in mobile crowdsourcingsystems,” Sensors, vol. 16, no. 4, p. 481, 2016.

[22] B. Zhang et al., “Privacy-preserving QoI-aware participant coordina-tion for mobile crowdsourcing,” Comput. Netw., vol. 101, pp. 29–41,Jun. 2016.

[23] C. Miao, H. Yu, Z. Shen, and C. Leung, “Balancing quality andbudget considerations in mobile crowdsourcing,” Decision Support Syst.,vol. 90, pp. 56–64, Oct. 2016.

[24] J. Sun, R. Zhang, X. Jin, and Y. Zhang, “Securefind: Secure andprivacy-preserving object finding via mobile crowdsourcing,” IEEETrans. Wireless Commun., vol. 15, no. 3, pp. 1716–1728, Mar. 2016.

[25] H. Shah-Mansouri and V. W. S. Wong, “Profit maximization in mobilecrowdsourcing: A truthful auction mechanism,” in Proc. IEEE Int. Conf.Commun. (ICC), Jun. 2015, pp. 3216–3221.

[26] O. Bello and S. Zeadally, “Intelligent device-to-device communicationin the Internet of Things,” IEEE Syst. J., vol. 10, no. 3, pp. 1172–1182,Sep. 2016.

[27] Y. Liu, L. Wang, S. A. R. Zaidi, M. Elkashlan, and T. Q. Duong,“Secure D2D communication in large-scale cognitive cellular networks:A wireless power transfer model,” IEEE Trans. Commun., vol. 64, no. 1,pp. 329–342, Jan. 2016.

[28] C. Zhao, S. Yang, X. Yang, and J. A. McCann, “Rapid, user-transparent,and trustworthy device pairing for D2D-enabled mobile crowdsourc-ing,” IEEE Trans. Mobile Comput., vol. 16, no. 7, pp. 2008–2022,Jul. 2017.

[29] F. Wang, C. Xu, L. Song, and Z. Han, “Energy-efficient resourceallocation for device-to-device underlay communication,” IEEE Trans.Wireless Commun., vol. 14, no. 4, pp. 2082–2092, Apr. 2015.

[30] Y. Zhang, E. Pan, L. Song, W. Saad, Z. Dawy, and Z. Han,“Social network aware device-to-device communication in wirelessnetworks,” IEEE Trans. Wireless Commun., vol. 14, no. 1, pp. 177–190,Jan. 2015.

[31] C. Xu, C. Gao, Z. Zhou, Z. Chang, and Y. Jia, “Social network-basedcontent delivery in device-to-device underlay cellular networks usingmatching theory,” IEEE Access, vol. 5, pp. 924–937, 2016.

[32] Y. Han and H. Wu, “Minimum-cost crowdsourcing with coverageguarantee in mobile opportunistic D2D networks,” IEEE Trans. MobileComput., vol. 16, no. 10, pp. 2806–2818, Oct. 2017.

[33] Z. Xiong, Y. Zhang, D. Niyato, P. Wang, and Z. Han, “When mobileblockchain meets edge computing,” IEEE Commun. Mag., vol. 56, no. 8,pp. 33–39, Aug. 2018.

[34] A. Dorri, S. S. Kanhere, R. Jurdak, and P. Gauravaram, “Blockchain forIoT security and privacy: The case study of a smart home,” in Proc. IEEEInt. Conf. Pervasive Comput. Commun. Workshops (PerCom Workshops),Mar. 2017, pp. 618–623.

[35] V. Buterin, Ethereum: A Next-Generation Cryptocurrency andDecentralized Application Platform. Nashville, TN, USA: BitcoinMagazine.

[36] K. Suankaewmanee, D. T. Hoang, D. Niyato, S. Sawadsitang,P. Wang, and Z. Han, “Performance analysis and application of mobileblockchain,” in Proc. Int. Conf. Comput., Netw. Commun. (ICNC),Mar. 2018, pp. 642–646.

[37] Y. Jiao, P. Wang, D. Niyato, and Z. Xiong, “Social welfare max-imization auction in edge computing resource allocation for mobileblockchain,” in Proc. IEEE Int. Conf. Commun. (ICC), May 2018,pp. 1–6.

[38] G. Zyskind, O. Nathan, and A. Pentland, “Decentralizing privacy: Usingblockchain to protect personal data,” in Proc. IEEE Secur. PrivacyWorkshops (SPW), May 2015, pp. 180–184.

[39] Z. Xiong, S. Feng, W. Wang, D. Niyato, P. Wang, and Z. Han,“Cloud/fog computing resource management and pricing for blockchainnetworks,” IEEE Internet Things J., to be published.

[40] H.-W. Kim and Y.-S. Jeong, “Secure authentication-management human-centric scheme for trusting personal resource information on mobilecloud computing with blockchain,” Hum.-Centric Comput. Inf. Sci.,vol. 8, no. 1, p. 11, 2018.

[41] B. Lee and J.-H. Lee, “Blockchain-based secure firmware update forembedded devices in an Internet of Things environment,” J. Supercom-put., vol. 73, no. 3, pp. 1152–1167, 2017.

[42] Q. Xia, E. B. Sifah, K. O. Asamoah, J. Gao, X. Du, and M. Guizani,“Medshare: Trust-less medical data sharing among cloud serviceproviders via blockchain,” IEEE Access, vol. 5, pp. 14757–14767,2017.

Page 13: A Blockchain-Powered Crowdsourcing Method With Privacy ...static.tongtianta.site/paper_pdf/88073aae-e052-11e9-9b27-00163e08… · Xiaolong Xu , Qingxiang Liu, Xuyun Zhang , Jie Zhang,

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

XU et al.: BPCM WITH PRIVACY PRESERVATION IN MOBILE ENVIRONMENT 13

Xiaolong Xu received the Ph.D. degree fromNanjing University, Nanjing, China, in 2016.

From 2017 to 2018, he was a Research Scholarwith Michigan State University, East Lansing, MI,USA. He is currently an Assistant Professor withthe School of Computer and Software, NanjingUniversity of Information Science and Technology,Nanjing. He has authored or coauthored more than50 peer review papers in international journals andconferences, including the IEEE TRANSACTIONSON CLOUD COMPUTING, the IEEE TRANSAC-

TIONS ON BIG DATA, the IEEE TRANSACTIONS ON COMPUTATIONAL

SOCIAL SYSTEMS, the IEEE TRANSACTIONS ON EMERGING TOPICS IN

COMPUTATIONAL INTELLIGENCE, SPE, Journal of Network and ComputerApplications (Elsevier), Future Generation Computing Systems, Concurrencyand Computation: Practice and Experience (Wiley), The Journal of Com-munity Informatics, ICWS, ICSOC, and so on. His current research interestsinclude mobile computing, edge computing, Internet of Things (IoT), cloudcomputing, and big data.

Dr. Xu was a recipient of the Best Paper Award of the IEEE CBD 2016.

Qingxiang Liu is currently pursuing the B.S. degreein computer science and technology with the Schoolof Computer and Software, Nanjing Universityof Information Science and Technology, Nanjing,China.

He has authored or coauthored a conference paperat IEEE Big data Security 2018 and a journal paperat Future Generation Computer Systems. His currentresearch interests include big data and mobile cloudcomputing.

Xuyun Zhang received the B.S. and M.E. degrees incomputer science from Nanjing University, Nanjing,China, in 2008 and 2011, respectively, and thePh.D. degree from the University of TechnologySydney (UTS), Ultimo, NSW, Australia, in 2014.

He was a Post-Doctoral Fellow with theMachine Learning Research Group, Data61, CSIRO,Canberra, ACT, Australia. He is currently a Lecturerwith the Department of Electrical and ComputerEngineering, The University of Auckland, Auckland,New Zealand. His current research interests include

Internet of Things (IoT) and smart cities, big data, cloud computing, scalablemachine learning and data mining, data privacy and security, and Web servicetechnology.

Jie Zhang received the B.Sc. and M.Sc. degreesin computer science from the Nanjing Universityof Information Science and Technology, Nanjing,China, in 2010 and 2013, respectively. She iscurrently pursuing the Ph.D. degree with NanjingUniversity, Nanjing.

She is currently an Engineer with the ShanghaiMeteorological Information and Technical SupportCenter, Shanghai, China. She has authored or coau-thored more than ten peer review papers in interna-tional journals and conferences. Her current research

interests include mobile computing, edge computing, Internet of Things (IoT),cloud computing, and big data.

Lianyong Qi received the Ph.D. degree from theDepartment of Computer Science and Technology,Nanjing University, Nanjing, China, in 2011.

He is currently an Associate Professor with theSchool of Information Science and Engineering,Chinese Academy of Education Big Data, QufuNormal University, Qufu, China. He has authoredor coauthored more than 50 papers including theIEEE Journal on Selected Areas in Communications,the IEEE TRANSACTIONS ON CLOUD COMPUTING,the IEEE TRANSACTIONS ON BIG DATA, Future

Generation Computing Systems, the IEEE Journal of Computer and SystemSciences, Concurrency and Computation: Practice and Experience (Wiley),ICWS, ICSOC, and so on. His current research interests include servicescomputing, big data, and Internet of Things (IoT).

Wanchun Dou received the Ph.D. degree inmechanical and electronic engineering from theNanjing University of Science and Technology,Nanjing, China, in 2001.

In 2005 and from 2008 to 2009, he was a VisitingScholar with the Department of Computer Scienceand Engineering, The Hong Kong University ofScience and Technology, Hong Kong. He is currentlya Full Professor with the State Key Laboratoryfor Novel Software Technology, Nanjing University,Nanjing, China. He has chaired three National

Natural Science Foundation of China projects. He has authored or coauthoredmore than 60 research papers in international journals and internationalconferences. His current research interests include workow, cloud computing,and service computing.