You are on page 1of 10

IEEE INTERNET OF THINGS JOURNAL, VOL. 7, NO.

6, JUNE 2020 5481

A Reliable Trust Computing Mechanism Based


on Multisource Feedback and Fog Computing
in Social Sensor Cloud
Junbin Liang , Min Zhang , and Victor C. M. Leung , Fellow, IEEE

Abstract—Social sensor cloud (SSC) is combined with social Index Terms—Feedback trust, fog computing, social sensor
network, wireless sensor network, cloud computing, and fog com- cloud (SSC), trust computing.
puting, which is currently a new type of Internet of Things (IoT).
In order to provide a convenient, open, and highly reliable SSC
services, the devices of fog computing are distributed at the edge
of cloud computing. The devices of fog computing can indepen-
dently process and store data, and feedback more quickly in
SSC. The sensing layer of SSC faces different types of physi- I. I NTRODUCTION
cal attacks and communication attacks, such as message forgery, N RECENT years, the social sensor cloud (SSC) [1]–[6]
message tampering, reply attacks, hidden data attacks, etc., lead
to the lack of trust between social sensors and cloud data cen-
ters in SSC. Therefore, the trust evaluation between the sensing
I has appeared as a new type of sensor cloud system, grad-
ually generated and applied with the integration of wireless
layer and the network layer is necessary. However, computing sensor network, social network, cloud computing technolo-
the reliability of the social sensor data in cloud data centers will gies, and fog computing technologies, which has also become
generate a large amount of trust computing overhead, commu- a new hotpot of future research direction in the field of Internet
nication overhead, and communication delay, which hinder the
widespread application of SSC services. To combat this issue, of Things (IoT). The SSC computes and stores data from the
a reliable trust computing mechanism (RTCM) based on mul- social sensors by the cloud computing technology, realizes the
tisource feedback and fog computing fusion is proposed. First, integration of the virtual social network information world and
a new metric is designed for the trust of social sensor nodes, the real physical world, and provides the users with conve-
and multisource feedback trust value collection is performed at nient, economical, and highly reliable SSC services. The SSC
the sensing layer to improve the detection of malicious feed-
back nodes. Second, the trust feedback information of the sensing covers a wide range of various fields, such as online medi-
layer is collected by the devices of fog computing, and the rec- cal health [7], disaster detection [8], traffic surveillance [9],
ommendation trust calculation is performed, which reduces the military surveillance [10], etc.
communication delay and computing overhead. Third, a fusion Fog computing is distributed between the sensing layer and
algorithm is designed to aggregate different types of feedback the network layer, supporting the SSC to provide the con-
trust values, which overcomes the limitation of trust weights in
artificial weighting and subjective weighting in traditional trust venient and open SSC services. Actually, fog computing is
mechanisms. Theoretical analyses and simulation results show a new type of edge computing network architecture [11],
that the proposed trust computing mechanism has better com- distributed on social sensing devices and systems to collect
putational efficiency and higher reliability compared with existing multisource feedback, and is responsible for a part of tasks
methods. in computing, thereby extends cloud computing to the edge
of the grid [12], multisource feedback includes the feedback
between social sensor nodes, and the feedback between sen-
Manuscript received November 24, 2019; revised February 18, 2020; sor nodes and fog devices (FDs). In addition, fog computing
accepted March 2, 2020. Date of publication March 16, 2020; date of
current version June 12, 2020. This work was supported in part by the is a scalable and distributed computing paradigm, enabling
National Natural Science Foundation of China under Grant 61562005; timely deployment based on multisource feedback, avoiding
in part by the Natural Science Foundation of Guangxi Province under network congestion and protecting user privacy effectively,
Grant 2019GXNSFAA185042, Grant 2018GXNSFBA281169, and Grant
2018JJA170215; in part by the Guangxi Key Research and Development Plan enabling timely deployment based on service requests from
Project under Grant Guike AB19259006; and in part by the Thousands of the sensing layer.
Young and Middle-Aged Backbone Teachers Training Program for Guangxi However, SSC has a similarity as the social evolution of
Higher Education [Education Department of Guangxi (2017)] under Grant 49.
(Corresponding author: Min Zhang.) online social networks. With social factors, SSC is different
Junbin Liang and Min Zhang are with the Guangxi Key Laboratory of from the traditional wireless sensor network. In the process
Multimedia Communications and Network Technology, School of Computer of collecting social data, there are several difficulties and trust
and Electronics Information, Guangxi University, Nanning 530004, China
(e-mail: liangjb@gxu.edu.cn; 15388951803@163.com). problems as follows.
Victor C. M. Leung is with the College of Computer Science and Software 1) The social sensor data are in large amount [13], and
Engineering, Shenzhen University, Shenzhen 518060, China, and also with its structure type is complex [14], social sensors have
the Department of Electrical and Computer Engineering, University of British
Columbia, Vancouver, BC V6T 1Z4, Canada (e-mail: vleung@ece.ubc.ca). limited storage capacity and heterogeneity devices are
Digital Object Identifier 10.1109/JIOT.2020.2981005 difficult to be compatible [15].
2327-4662 
c 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.

Authorized licensed use limited to: Istinye Universitesi. Downloaded on April 21,2023 at 01:21:43 UTC from IEEE Xplore. Restrictions apply.
5482 IEEE INTERNET OF THINGS JOURNAL, VOL. 7, NO. 6, JUNE 2020

2) Based on the open social cloud environment, there are security and Quality of Service (QoS) of SSC services. The
a large number of social sensors that may refuse to for- main contributions of this article are as follows.
ward messages [16] or propagate a large number of false 1) A fog-based SSC high-reliability trust computing archi-
messages [17], wasting network transmission resources. tecture is established, enabling the social sensor data
3) Social sensors have random mobility, their perceived to be effectively layered and processed in three stages
data are prone to failure and cannot assess data of collection, transmission, and calculation, which
authenticity [18], resulting in unnecessary computational improves the efficiency of trust computing and reduced
overhead and unreliable SSC services. communication latency and network load.
Malicious service providers provide untrustworthy 2) A lightweight trust computing mechanism combined
services [19]. For example, a large number of social users with multisource feedback and fog computing is
posted and disseminated false messages on social media designed. The trust computing is completed by the
sites during a major news event. First, false messages have sensing layer and the fog layer. Social sensors of the
a negative influence in the respect of social communica- sensing layer are monitored by FD of the fog comput-
tion; second, the limited energy of the node is wasted and ing in real time, which meets the requirements of low
the service provided is unreliable from the perspective of latency. FD undertakes part of the task of trust comput-
SSC [20]. In the mobile ad hoc networks, untrusted nodes ing and storage in the cloud data center and reduces the
may adversely affect the quality and reliability of the sensor computational overhead and transmission cost of cloud
data, thereby harm the security of the service [31]. Trust computing.
and reputation systems play an important role in decision 3) A new trust algorithm that aggregates different types of
making on the Internet [30]. Therefore, it is very important feedback is proposed. By weighting the trust values of
to analyze and detect the SSC data and progress the relia- different types of feedback, weight distribution is per-
bility of service. In this article, a reliable trust computing formed according to different trust factors, which over-
mechanism (RTCM) based on multisource feedback and fog comes the limitation of weight assignment of traditional
computing in SSC is proposed, combining with the trust trust schemes.
computing of feedback information in SSC, and consider- The organization of this article is as follows. Section I
ing the rapidity, real time, validity, and accuracy of SSC introduces the trust computing scheme existing in SSC, and
service. Section II gives an overview of related work about trust mech-
First, a fog-based SSC highly reliable trust computing archi- anisms. Section III describes the SSC architecture with trust
tecture is established, in order to improve the security and mechanism and Section IV gives the trust analysis the fog
quality of SSC service better. SSC consists of cloud data cen- computing of the SSC architecture. Based on the trust analy-
ters, FDs, and sensors. The data collected by social sensors sis of fog computing, Section V gives a detailed description
are uploaded to the fog layer for preliminary processing and of RTCM based on multisource feedback and fog computing.
then collected through the cloud data centers. Data processing Section VI analyzes the theory and the reliability of RTCM.
and storage based on the cloud computing platform, to pro- Section VII summarizes this article and proposes the direction
vide high-reliability and low-latency sensor cloud services for of future research.
different social users.
Second, the multisource feedback trust computing is per- II. R ELATED W ORK
formed between the sensing layer and the fog layer and does
Social sensing cloud service applications have been widely
not need to be uploaded to the cloud data centers, in order to
used, more researchers are paying attention to the social
reduce the overhead of trust computing. Social sensor nodes
security cloud-related security issues, and propose different
of the sensing layer and devices of the fog layer feedback trust
solutions based on the trust mechanism. In order to research
value and trust computing is carried out on the fog platform.
SSC security deeply, this section will provide insight into the
Therefore, real-time validity and low overhead of the proposed
latest research advances in this area.
trust mechanism are guaranteed, which is combined with fog
computing.
In addition, a new trust algorithm that aggregates differ- A. Trust Based on Social Evolution
ent types of feedback is proposed by aggregating feedback Chang et al. studied the user trust in social network services
trust values of different devices, since it is necessary to estab- (SNSs), designed an improved model, conducted quantita-
lish a trust computing mechanism with high-reliability SSC tive empirical research, explored impacts of social influence,
by setting new trust metrics based on feedback from different privacy issues, and perceived risks from the user groups in
nodes and servers, we propose a trust computing algorithm that Facebook and LinkedIn. However, the study has certain lim-
aggregates different types of feedback trust values from dif- itations, SNS includes many types, and each has different
ferent devices, the target node, and the intermediate node, and user group characteristics [32]. Lee et al. [25] proposed an
matrices are established to perform multidimensional aggrega- approach to measure the trust of IoT social media environ-
tion on the trust value, and weights according to different types ment service based on ontology. A trust ontology method
of feedback values to aggregate the trust value of the social based on the social content semantic library is constructed,
sensor node, then the trust value of the social sensor node extended, and personalized [26], by establishing novel trust, it
is evaluated. The accuracy of data transmission by measuring provides reliable trust metrics for the users and the trust-aware

Authorized licensed use limited to: Istinye Universitesi. Downloaded on April 21,2023 at 01:21:43 UTC from IEEE Xplore. Restrictions apply.
LIANG et al.: RELIABLE TRUST COMPUTING MECHANISM BASED ON MULTISOURCE FEEDBACK AND FOG COMPUTING IN SOCIAL SENSOR CLOUD 5483

services with considering the dynamic trust of the social


media. However, the approach only examines simple trust
metrics from users to social content, without including the
effectiveness and usability in real-world trust-aware environ-
ments.
Considering the influence of social factors in SSC,
Guler et al. [21] proposed the optimal communication strategy
scheme based on the principle of the homophily, which is with
positive and negative trust relationships in the social sensor
network. The authors integrate social awareness into network
communication under the influence-centered constraints [22].
The social-aware sensors collect and process the data avail-
able in the social media, and predict the type of friendship
or hostile relationship of social sensors through links. The
analysis shows that the scheme reduces the end-to-end delay
and propagation costs in social communication effectively. Fig. 1. Reliable trust computing architecture of SSC based on multisource
However, the scheme only verifies the influence of different feedback and fog computing.
nodes in the network through the simple summation and aver-
aging method of binary evaluation, neither involves the trust
problems of different levels in a large system such as SSC nor C. Trust Based on Edge Device Feedback
involves the trust computing overhead and communication cost Due to the characteristics of social evolution in SSC,
problems. using the feedback trust computing in the social network
can further improve the reliability of the trust mechanism.
Combined with the idea of edge computing, Yuan and Li [28]
B. Trust Combined With Fog Computing
proposed a reliable and lightweight trust mechanism of multi-
Pahl et al. [23] proposed an architecture pattern in the for- source feedback information fusion based on IoT edge devices.
mat of a state machine that describes the basic processing First, the lightweight trust evaluation mechanism is used to
of the orchestration activities, supported by a blockchain to coordinate the IoT edge devices, and the global trust cal-
enable trusted orchestration management (TOM). The archi- culation is performed through multisource feedback, which
tecture coordinates the communication between the sensors reduces the overhead of trust computing; second, a feedback
and the cloud services by introducing the fog and edge information fusion algorithm based on objective information
architecture and processes the data from sensors and clouds. entropy theory is proposed, which overcomes the shortcomings
Li et al. [24] used the blockchain technology to solve the secu- in traditional methods. The experimental results show that the
rity problems of the edge device, and analyzed the identity of mechanism has better computational efficiency and reliability.
device entity, sensor data source, and data transaction pro- In summary, enhancing trust is undoubtedly a reliable mea-
cessing, to prevent this information from being tampered. The sure to ensure the security of SSC services. However, most
experimental results show that the blockchain-based solution of the current research programs only focus on social fac-
can map trust to the IoT edge cloud architecture while con- tors affecting communication, or subjective methods are used
sidering the robustness and robustness of trust. However, the to assign weights to trust decision factors, which reflects the
feasibility while storing the sensor data in a large amount in lack of adaptability to global trust aggregation computing. On
the blockchain is not considered. the other hand, these schemes do not take into account the
In order to ensure the security of SSC data storage, large scale in SSC, which cannot be applied to solve a large
Zhang et al. [27] studied the hidden data attacks in sensor number of trust computing, the reliability of trust computing is
clouds, and designed a detection system based on fog comput- reduced. Therefore, it is necessary to design a low-overhead,
ing and trust evaluation mechanism. First, the trust evaluation high-reliability trust mechanism for SSC.
mechanism composes of three layers: 1) the direct trust layer in
WSN; 2) the preliminary decision layer of the underlying FD;
and 3) the data analysis layer based on the fog platform. The III. A RCHITECTURE OF SSC W ITH T RUST M ECHANISM
main functions of the underlying FD include collecting sensor In this section, we first present the fog-based SSC architec-
trust value and topology information, making major decisions, ture, second, we also analyze the function of each layer in the
and passing messages from the fog platform. A hierarchi- architecture.
cal trust evaluation mechanism can resist more attacks with As shown in Fig. 1, RCTM based on multisource feedback
lower network resource consumption costs. Second, the author and fog computing in SSC, in which fog computing under-
defines three types of scenarios (redundant data, parametric takes a part of data analysis and management tasks from the
curve features, and data validation) for verification, and gives cloud data center, and bring kinds of potential advantages. For
three detection schemes. The experimental results show that example, the devices of fog computing provide a wide range of
the proposed mechanism has a certain reliability in detecting services for a large number of heterogeneous sensors and pro-
hidden data attacks. cess the sensor data rapidly, thereby reducing network latency

Authorized licensed use limited to: Istinye Universitesi. Downloaded on April 21,2023 at 01:21:43 UTC from IEEE Xplore. Restrictions apply.
5484 IEEE INTERNET OF THINGS JOURNAL, VOL. 7, NO. 6, JUNE 2020

and responding faster to service requests from the sensing fog layer can shorten the response time effectively, reduce the
layer, based on huge data resources of cloud computing, fog delay by effective communication, and save the network data
computing, and cloud data center work in an integrated man- transmission cost.
ner to solve the resource and QoS requirements of large-scale In SSC, considering the interaction between different social
IoT systems. sensors, social sensors, and FDs, we establish a trust mecha-
The SSC reliable trust computing architecture based on mul- nism based on the feedback of multiple social sensors on FDs
tisource feedback and fog computing is divided into three systematically.
layers: 1) the network layer; 2) the fog layer; and 3) the sensing The trust feedback involves two sets of entities. The first
layer. Their composition and function are as follows. is the set of social sensors (S = {s1 , s2 , . . . , si , . . . , sn }), i
1) The sensing layer is composed of social sensors. represents the ID of the social sensor, and n represents the
The social sensors transmit data by WIFI, 433 MHZ, total number of social sensors, the second is the set of FD
ZigBee (2.4G), and other technologies, and communi- (D = {dl , d2 , . . . , dw , . . . , dm }), w represents the ID of the
cate through wireless communication base stations or FD, and m represents the total number of FDs. There are two
gateways. Before completing data interaction, the social basic trust feedback relationships between the two entities: the
sensors send request messages to FDs for ensuring the first one is the feedback relationship between social sensors,
reliability of social sensors being interacted with. which is the basic trust relationship when the social sensors
2) The fog layer consisting of FDs and fog platforms. communicate, the second one is the feedback relationship of
Due to the open social sensor network, there may exist FDs to social sensors, it is a special indirect trust relationship
some malicious social sensors, which generate malicious in this article, and it is also a key factor to reduce the sensor
feedback, resulting in unreliable trust computing of the malicious feedback, which is extremely important for RTCM.
fog layer. Therefore, we extend the traditional feed- Next, with reference to the methods in [27] and [28], we
back mechanism. The FDs are deployed to monitor the present the definition of the trust relationship involved in this
interaction behavior of the social sensors and summa- article.
rize feedback of the social sensors. The fog platforms 1) Definition 1 (S-to-S Direct Trust): S-to-S direct trust
initially analyze the feedbacks of the FD to the social refers to the ability value of the social sensor node
sensors, and send the analysis result to the cloud data to complete the task, and the record of trust degree is
center. Moreover, RCTM proposed in this article is dif- obtained through the direct interaction history between
ferent from the traditional feedback mechanisms, the the two devices.
trust aggregation computing based on multisource feed- 2) Definition 2 (D-to-S Recommendation Trust): D-to-S
back information through FD instead of the cloud data recommendation trust is the trust value objectively cal-
center, which not only reduces the transmission cost and culated by the fog platform for the social sensors. The
delay but also improves the computing efficiency and social sensor feeds its direct trust to other sensors to
reliability. the FD, and the FD is assembled. Different trust values
3) The network layer is supported by cloud computing. are transmitted to the fog platform for calculation, and
Since the cloud data center contains central servers that the fog platform generates recommendation trust for the
manage databases, and cloud computing provides the social sensors.
infrastructure, platform, and software services for the 3) Definition 3 (S-to-S Global Trust): S-to-S global trust
large-scale IoT systems, enabling various industries and refers to the quantitative value of the ability of the social
organizations to obtain more security data and access sensor to request the service object to complete the task.
to information and interfaces easier. Consequently, we The global trust is computational integrated based on
assume that the cloud data center is always trustworthy. S-to-S direct trust and D-to-S recommendation trust.
According to Definitions 1–3, S-to-S global trust is the result
of aggregation computing by multiple trust factors. Since tradi-
IV. T RUST A NALYSIS BASED ON F OG C OMPUTING IN SSC tional schemes only consider the subjective S-to-S direct trust,
In the proposed RTCM combining multisource feedback and the credibility of the sensor cannot be correctly assessed when
fog computing, the trust computing is completely completed malicious attacks occur, resulting in the risk of data leakage.
by the fog layer and the sensing layer, and there is no demand In order to solve this problem, RCTM involves direct trust
to upload tasks to the central server of the cloud. and recommendation trust and involves the social evolution
At the sensing layer, the social sensors generate more feed- of social sensors. We use the objective trust method to com-
back. The feedback is uploaded to the fog layer, collected pute integrated trust between the sensors and overcome the
by the FD, and analyzed and managed by the fog platform, limitations of traditional trust mechanisms.
to increase the efficiency and effectiveness of trust comput-
ing all the feedback information is uploaded to the cloud data
V. M ULTISOURCE F EEDBACK AND T RUST C OMPUTING
center for processing and storage by the central server, the
response time of the trust service request will be extended, A. Definition of Trust Computing
the data transmission cost will also increase, then SSC will As shown in Fig. 2, the social sensing layer generates direct
face extreme challenges in the network load and reliability. trust, and the direct trust are fed back to the fog layer. The
Therefore, trust computing through the sensing layer and the fog layer generates recommendation trust for the sensing layer

Authorized licensed use limited to: Istinye Universitesi. Downloaded on April 21,2023 at 01:21:43 UTC from IEEE Xplore. Restrictions apply.
LIANG et al.: RELIABLE TRUST COMPUTING MECHANISM BASED ON MULTISOURCE FEEDBACK AND FOG COMPUTING IN SOCIAL SENSOR CLOUD 5485

ωm represents the weighting factor of the ser-


 attrm , and meets the conditions of
vice attribute
0 ≤ ωm ≤ 1, M m=1 ωm = 1.

B. S-to-S Direct Trust


The S-to-S direct trust computing is obtained from the
records of past interactions of different social sensors directly
and does not need to be requested from a third sensor. Each
interaction between the social sensors generates S-to-S direct
trust. In wireless sensor networks, different types of evidence
can be collected, such as packet success rate, routing success
rate (interactive frequency), forwarding delay, data correctness
rate, etc. Considering the social factors of social sensors, the
social sensor proposes that the object of the service request
Fig. 2. Trust calculation based on multisource feedback and fog computing.
service has its own personality preference. Here, we use the
service quality rating as a factor of trust calculation, and the
personality preference between the social sensors and different
through computational fusion. For example, FD is responsible objects as weight.
for monitoring the service behavior of the social sensor. The Before the service provider sj receives the service request,
social sensor periodically feedbacks direct trust to FD. The FD its own service quality evaluation factor is
calculates the recommendation trust between different sensors    M
sj ,t
and global trust. If the social sensor si wants to obtain the SG Qself,t
sj = Q s j , t = ωm qattr m
. (1)
trust value of sj , then si sends a request to the FD, and the FD m=1
feeds back the result to si . After an interaction, according to the service attribute, the
First, we define specific attributes and parameters of the QoS factor obtained by the social sensor si (si ∈ S) that issues
relevant trust calculation. the service request to the service provider sj (sj ∈ S) is
1) Set of Service Attributes: A set of service attributes
represent a collection of quality of interaction   M
s ,s
between social sensors. Expressed with A T T R = Qtsi ,sj = Q si , sj , t = ωm qmi j . (2)
{attr1 , attr2 , . . . , attrm }, attrm represents the m attribute m=1
that affects the trust rating. There are many factors The difference between the service provider sj ’s own QoS
affecting the QoS. In this article, we use the packet size, rating factor and the QoS rating factor obtained by the social
forwarding delay, and route success rate as the attributes sensor si that issued the service request is
affecting the quality service. According to the weights  
 t self,t 
of different attributes, the normalized method is used to Q(s )
Q si ,sj
− Q sj

DQ sj ,s = . (3)
process each attribute value from 0 to 1. ( i j) t
Qsi ,sj
2) Factor of Self-Service Quality Rating: The factor rep-
resents the value of each service quality that a social Q(s )
If DQ(sji ,sj ) exceeds a certain threshold, it can be considered
sensor claims to other social sensors within a cer- that the service provided by the social sensor sj is unreliable
tain period of time (t), it is expressed as, Qs,t ATTR = and its trust value is lowered. The trust value is expressed as
sj ,t sj ,t sj ,t sj ,t
{qattr , qattr2 , . . . , qattrm }, 0 ≤ q attrm ≤ 1, indicates the ⎧ Q(sj )
1

⎪ 1, 0 ≤ DQ(si ,sj ) ≤ ς1
QoS attribute value m of service provider sj . We assume ⎪
that the social sensor can update its own quality rating   ⎨
Q(s )
Vk si , sj = 0.5, ς1 ≤ DQ sj ,s ≤ ς2 (4)
factor in time according to its service performance and ⎪
⎪ ( i j)
feedback from the service requester each time the task ⎪
⎩ 0, ς ≤ D ( j ) ≤ 1
Q s
2 Q(si ,sj )
is completed.
3) Factor of User’s Quality Rating: The QoS attribute value 0 ≤ ς1 ≤ ς2 ≤ 1, Vk (si , sj ) represents the QoS trust rating
obtained by the service provider’s social sensor si (si ∈ S) obtained by the social sensor si after social sensor si and sj
for the service provider si (si ∈ S) after each interaction, interact for the kth time.
si ,sj si ,sj si ,sj si ,sj
defined as QATTR = {qattr 1
, qattr 2
, . . . , qattr m
}, 0 ≤ According to (1)–(3), the direct trust value of the social
si ,sj
qattrm ≤ 1 indicates the evaluation value of the service sensor si to sj is expressed as SDk (si , sj ), and calculated as
provider si QoS attribute m for the provider sj after an follows:
interaction.  
SDk si , sj
4) Social Sensor Personality Preferences: When social sen-    
sors choose different objects to interact, they have σ Vk si ,sj ×ck si , sj , k = 1   
=
their own preferences for the weight of different ser- δSDk−1 si , sj + σ (1 − δ)Vk si , sj × ck si , sj , k ≥ 2.
vice attributes, expressed by ω = {ω1 , ω2 , . . . , ωM }, (5)

Authorized licensed use limited to: Istinye Universitesi. Downloaded on April 21,2023 at 01:21:43 UTC from IEEE Xplore. Restrictions apply.
5486 IEEE INTERNET OF THINGS JOURNAL, VOL. 7, NO. 6, JUNE 2020

μ is the number of successful interactions between the Algorithm 1: Recommended Trust Algorithm Feed Back
two sensors si and sj , and v is the total number of inter- by FD
actions. δ is defined as the historical factor, indicating the Input : the normalized matrix fdw →sj (t)
weight of the direct trust value when the interaction is rel- Output: DB(si , sj )(t)
ative to the historical interaction, δ(0 ≤ δ ≤ 1). Ck (si , sj ) 1 Begin
2 for i= 1 to n do
indicates the correct rate of data transmission during this 3 for j= 1 to n do
interaction. 4 According to the definition of trust matrix, calculate trust
value of each data in fdw →sj (t) ;
5 SD(si ,sj ) (t) = δSDk−1 (si , sj ) + σ (1 − δ)Vk (si , sj ) × ck (si , sj );
C. D-to-S Recommended Trust 6 if i = j then
7 set SD(si ,sj ) (t) = 0 ;
In a certain social sensing area, we assume that there are
8 end
n social sensors (S = {S1 , S2 , . . . , Si . . . , Sn }) communicating 9 end
with the FD, and the FD dw periodically sends the request 10 end
packets to the social sensor. 11 According to the definition of vector: 
12 u(j) = (u1 , u1 , . . . , uj , . . . , un ), (when j = i, uj = 0 ;
In response, all social sensors that accept the request Calculate the
13
 weight of each index  through information entry:

will upload their interaction history with other social sen- 14
 η1 , η2 , . . . , ηj , . . . , ηn , (when
hη (j) =  j = i, ηj = 0 where
sors, including direct trust values fed back to the FD ηj = uj / u1 + u2 + . . . + uj + . . . + un (j = 1, 2, . . . , n) ;
15 DBs ,s  (t) = hη (j) ∗ fdw →sj (t);
and the FD saves these direct trust values in matrix i j
End
fdw →sj (t) 16

fdw →sj (t)


⎛ ⎞
SD(s1 ,s1 ) (t) ··· SD(si ,s1 ) (t) ··· SD(sn ,s1 ) (t) according to the following equation:
⎜SD(s1 ,s2 ) (t) ··· SD(si ,s2 ) (t) ··· SD(sn ,s2 ) (t)⎟ SG(si→ sj ) (t) = α(t) ∗ SDk (si ,sj ) (t)
⎜ ⎟
⎜ ··· ··· ··· ··· ··· ⎟
=⎜ ⎜SD(s ,s ) (t)
⎟. + (1 − α(t)) ∗ DB(si ,sj ) (t) (9)
⎜ 1 j
··· SD(si ,sj ) (t) ··· SD(sn ,sj ) (t)⎟⎟
⎝ ··· ··· ··· ··· ··· ⎠ where α(t) and 1 − α(t) are the weights of SDk (si , sj ) and
SD(s1 ,sn ) (t) ··· SD(si ,sn ) (t) ··· SD(sn ,sn ) (t) DB(si ,sj ) , respectively. We believe that sensors with social fac-
(6) tors, if more successful interactions, the two will trust each
other more.
Since the recommendation trust value should be a nor- Therefore, the weight of direct trust is higher. Traditional
malized matrix, to eliminate the influence of the physical schemes generally assign weights manually, resulting in a lack
dimension. When the FD feeds the recommendation trust to of flexibility in the weighting of trust factors.
the social sensor si , in order to avoid affecting the objectivity, Based on the low latency and flexibility of the fog com-
the social sensor si ’s own evaluation is discarded. puting, we use the FD to monitor the requests of different
Aggregating the recommendation trust of different social social sensors in a certain period of time, and set the weight
sensors to sj , considering the historical success rate (repre- of the trust factor according to the frequency requested
sented by η) of the social sensor si interacting with other by the social sensor sj for si within the specified period.
social sensors as a weighting factor: ηj = uj /(u1 + u2 + · · · + α(t) = 1 − vn (0 ≤ v ≤ 1), n is the number of the times that
uj + · · · + un ), uj represents the number of successful interac- sensor sj is requested to interact with si . Algorithm 2 details
tions between the jth sensor and si , when j = i, uj = 0, the the global trust computing of the social sensor.
weighting factor of the recommended trust value is represented
by vector hη (j) as VI. T HEORETICAL A NALYSIS AND S IMULATION
   A. Theoretical Analysis
hη (j) = η1 , η2 , . . . , ηj , . . . , ηn j = i, ηj = 0 . (7)
1) Space Complexity: The traditional trust models usually
According to (6) and (7), the recommendation of the FD use the broadcast policy to collect the feedback trust value of
based on feedback to the social sensor si can be expressed as the entire cluster, which increases the communication overhead
and communication delay of the mechanism. In our proposed
DB(s ,sj ) (t) = hη (j) ∗ fdw →sj (t). (8) RTCM, social sensors do not share trust information with
i

The social sensors send feedback trust to the FD, and the others, but calculate according to feedback generated by the
recommendation trust generated by the FD calculated is as collection of FDs. We assume that there are m fog platforms in
shown in Algorithm 1. the SSC, and each fog platform serves n social sensor nodes.
In the specified time interval, the maximum trust calculation
is γ times, so the maximum communication overhead of trust
D. S-to-S Global Trust information transmission is
After obtaining the D-to-S recommendation trust of S-to-S
totalcommunication (t) = m ∗ (n + 2n) ∗ γ = 3mnγ .
direct trust and FD feedback, according to Definition 3, global
trust is the result of the fusion of S-to-S direct trust and D-to-S Proof: When calculating the global trust, the social sensor
recommendation trust, so global trust should be calculated node si sends a feedback request to the FD dw , and receives

Authorized licensed use limited to: Istinye Universitesi. Downloaded on April 21,2023 at 01:21:43 UTC from IEEE Xplore. Restrictions apply.
LIANG et al.: RELIABLE TRUST COMPUTING MECHANISM BASED ON MULTISOURCE FEEDBACK AND FOG COMPUTING IN SOCIAL SENSOR CLOUD 5487

TABLE I
Algorithm 2: Global Trust Algorithm for Social Sensors PARAMETERS AND P OSSIBLE VALUES
Input : a social sensor set S = {s1 , s2 , . . . , si , . . . , sn }, a fog device set
D = {d1 , d2 , . . . , dw , . . . , dm } , S-to-S direct trust
SDs ,s  (t), D-to-S feedback trust DBs ,s  (t) and the time
i j i j
interval t for trust calculation;
Output: SGs →s  (t)
i j
1 Begin
2 Social sensor si send a request message to fog device dw for si ’s
feedback trust DBs ,s  (t);
i j
3 if (t > 0) then
4 for ((z=1 to n, and Sz ∈ S) do
5 sz send the S-to-S direct trust SDs ,s  (t) to their fog device
i j
dw ;
6 end NetLogo simulator for the experiment. PSM is based on the
7 dw aggravates these feedbacks to obtain DBs ,s  (t) and sent it cosine similarity to measure observing conditions by describ-
i j
8 to si ;
9  
si computes sj ’s S-to-S direct trust SD s ,s (t) according to sj ’s
ing how much the traveling status of rater is similar to that
i j of ratee, and the fixed weight of similarity is preset for high-
10 behavior;
11 si aggravates SD s ,s (t) and DB s ,s (t) and SGs ,s  (t);
    accuracy reputation computation. DRM is proposed by setting
i j i j i j
12 end distributed local authorities (LAs) for detecting node to update
13 End the reputation values, then LAs comprehensively weigh the
reputation values of service providers and optimize service
resource allocation. In simulations, we deploy two types of
at most one feedback response from the FD. Therefore, the social sensor nodes and FD. There are two types of social
total communication overhead of the recommendation trust of sensor nodes, one is honest sensor node (HS) and it always
the requesting FD is 2nγ . Next, si will use the direct trust provides correct feedback, the other is malicious and provides
information to sj to feedback to the FD, and the total commu- feedback that is opposite to the actual data. In the simulator,
nication overhead of the feedback report information is nγ . it is assumed that the feedback provided by the fog layer is
Therefore, if there are m FDs, the maximum communication always reliable because the fog layer is managed by the cloud
overhead for completing the trust calculation is 3mnγ . data center.
2) Time Complexity: In the proposed trust computer The parameters and values used in the simulation are shown
system, the maximum time complexity of S-to-S global trust is in Table I. The computer used in the simulation is con-
  figured as CPU 3.2 GB, memory 16 GB, and hard disk
totaltime (t) = O n2 . 1 TB. The simulator has 1000 social sensor nodes and 40
FDs. The simulation running time step is 600 s, and the
Proof: It is assumed that there are m FDs in the SSC and trust calculation interval is 50 s. In the simulation, the trust
each FD serves n social sensor nodes. The time complexity of value of the node ranges from 0 to 1 (4), we set the trust
S-to-S global trust is determined by the number of executions threshold to 0.5, and HSs are trusted node with an ini-
of the algorithm in a specified time interval t. tial trust value greater than 0.5. After successful interaction,
In Algorithm 1, the recommendation trust fed back to the the trust value is promoted by the FD (9), the more inter-
FD by the social sensor is calculated, and the number of cycles actions, the trust value will be promoted higher. FD is a
is n2 , in Algorithm 2, the recommendation trust is requested collaborative device that is responsible for global trust cal-
by the social sensor that sends the request and the global trust culation. We assume that feedbacks of FD are always reliable,
value calculation is performed on the other social sensor in the greater the number of participating collaborations with
combination with the direct trust and the number of cycles the higher computing efficiency. The social sensor honesty
is n, so the time complexity is totaltime (t) = O(n2 ). ratio in the three simulation scenarios is 90%, 80%, and
60%. The FD collaboration ratio is divided into 75%, 50%,
B. Simulations and 25%, which means the SSC system correspondingly are
1) Simulation Design: In order to verify and analyze the idle, busy, and highly busy. The simulated device entity
reliability of the proposed trust computing mechanism, we use indicates that the underlying SSC is in different network
the NetLogo to conduct simulations. Netlogo is a simulator states.
implemented in the Java language, and it can simulate the Scene 1: The node is honest and the FD is idle (HS accounts
process of interaction between social sensor entities and FDs for 90% and FD accounts for 25%).
in the SSC. Scene 2: The node is dishonest and the FD is busy (HS
In the trust computing mechanism based on multisource accounts for 80% and FD accounts for 50%).
feedback aggregation, the main threat comes from the feed- Scene 3: The node height is dishonest and the FD is highly
back of malicious social sensor nodes in the sensing layer. busy (HS accounts for 60% and FD accounts for 75%).
In this experiment, we design three scenes for simulation. We obtain the task success rate of social sensors in different
For comparison, a current typical personal similarity mea- scenes, perform trust evaluation for social sensor nodes, and
sure mechanism (PSM) [28] and a distributed reputation compare and analyze the computational efficiency and reli-
management mechanism (DRM) [29] are added to the ability of the trust computer system proposed. For a better

Authorized licensed use limited to: Istinye Universitesi. Downloaded on April 21,2023 at 01:21:43 UTC from IEEE Xplore. Restrictions apply.
5488 IEEE INTERNET OF THINGS JOURNAL, VOL. 7, NO. 6, JUNE 2020

Fig. 3. Task success ratio in scene 1. Fig. 5. Task success ratio in scene 3.

DRM when the nodes are relatively dishonest and the FD is


busy.
In Fig. 5, it shows that in scene 3, at the beginning of the
simulation, RTCM takes the task success rate for 64%, PSM
takes the task success rate for 48%, and DRM takes the task
success rate for 38%. With the changing of time, task success
rate in the three scenes is gradually improved. It shows that
with the state of node dishonesty and FD, RTCM has better
reliability than PSM and DRM.
3) Analysis of Simulation Results (Trust Evolution): In the
three scenes, social sensor node accepts the data request, and
completes data interaction, then the trust value of social sen-
sor can be further increased and the interaction record is
submitted to the FD for saving and updating. If the node
Fig. 4. Task success ratio in scene 2.
does not complete the interaction, then the interaction record
will also be submitted to the FD, and updated according to
comparison, PSM and DRM are added to the NetLogo for Algorithms 1 and 2, and the weights of S-to-S direct trust
simulation. and D-to-S recommended trust are adjusted according to the
2) Analysis of Simulation Results (Reliability With Different number of interactions and the success rate of the interaction,
Percent of HS): In the simulation, we compute task success thereby overcomes the limitation of traditional trust computing
rates to reflect the reliability of RTCM. According to the task mechanisms to manually set the weight.
success rate of the social sensor which changes with time in In the three scenes, the average trust value of social sensor
different scenes, the reliability of the proposed trust comput- nodes are separately counted. The initial global trust SG of
ing mechanism is reflected. The higher the task success rate, the honest social sensor node is randomly distributed from 0.5
the higher the reliability of the trust computing mechanism. to 1, and the malicious social sensors are randomly distributed
Figs. 3–5 show the task success rate in the three scenes. In from 0 to 0.5. Figs. 6–8 show the average trust value of social
these simulations, we assume that the feedback of FD is always sensors in scenes 1–3. We can draw a conclusion that RTCM
trustworthy. has higher values as the interaction time increasing than PSM
Fig. 3 shows that in scene 1, at the beginning of the simula- and DRM.
tion, the trust mechanism we proposed takes the task success The above simulation results reflect that when the node
rate for 83%, significantly higher than PSM and DRM, and the is dishonest and the network state of the FD is busy, it
task success rate of the three mechanisms over time tends to will affect the data interaction between social sensor nodes,
a higher level. It shows that the three trust computing mech- and decrease the efficiency of trust computing in SSC bot-
anisms have higher reliability when the nodes are relatively tom, and the worse network state, the more significant of
honest and the FD is idle. impact. The trust calculation of RTCM is combined of
In Fig. 4, it shows that in scene 2, at the beginning of feedback by FD and dynamically adjusts the weight of
the simulation, RTCM takes the task success rate for 82%, S-to-S direct trust and D-to-S recommended trust accord-
PSM takes the task success rate for 75%, and DRM takes ing to feedback. Compared with PSM and DRM, it has a
the task success rate for 67%. With the changing of time, higher task success rate in different scenarios. Therefore, the
the success rate of the three mechanisms becomes stable. proposed trust computing mechanism has higher reliability and
It shows that RTCM has higher reliability than PSM and flexibility.

Authorized licensed use limited to: Istinye Universitesi. Downloaded on April 21,2023 at 01:21:43 UTC from IEEE Xplore. Restrictions apply.
LIANG et al.: RELIABLE TRUST COMPUTING MECHANISM BASED ON MULTISOURCE FEEDBACK AND FOG COMPUTING IN SOCIAL SENSOR CLOUD 5489

FD, effectively improve the computing efficiency, reduce the


communication delay and transmission cost, and enhance the
service quality and security of the SSC.
The simulation results show that compared with the exist-
ing trust mechanisms, RCTM has obvious advantages in terms
of computational efficiency and reliability. However, there are
still many fields for improvement in the current trust com-
puting system based on fog computing. The next step of the
research work is to consider how to adjust the weights of dif-
ferent trust factors more flexibly according to the feedback
of social sensor nodes, to increase the reliability of the trust
computing mechanism.

R EFERENCES
Fig. 6. Average trust value in scene 1. [1] C. Zhu, V. C. M. Leung, J. J. P. C. Rodrigues, L. Shu, L. Wang,
and H. Zhou, “Social sensor cloud: Framework, greenness, issues, and
outlook,” IEEE Netw., vol. 32, no. 5, pp. 100–105, Sep./Oct. 2018.
[2] Q. Xu, Z. Su, S. Yu, and Y. Wang, “Trust based incentive scheme to
allocate big data tasks with mobile social cloud,” IEEE Trans. Big Data,
early access, doi: 10.1109/TBDATA.2017.2764925.
[3] T. Aamir, A. Bouguettaya, H. Dong, A. Erradi, and R. Hadjidj, “Social-
sensor cloud service selection,” in Proc. IEEE Int. Conf. Web Services
(ICWS), 2017, pp. 508–515.
[4] W. Tian, L. Yang, J. Weijia, W. Guojun, and P. Shaoliang, “Research
progress of sensor-cloud security,” J. Commun., vol. 39, no. 3,
pp. 35–52, 2018.
[5] A. G. Neiat, A. Bouguettaya, T. Sellis, and S. Mistry, “Crowdsourced
coverage as a service: Two-level composition of sensor cloud services,”
IEEE Trans. Knowl. Data Eng., vol. 29, no. 7, pp. 1384–1397, Jul. 2017.
[6] R. J. R. Reyes, F. F. de Mendonca, and K. L. Dias, “A service-
oriented architecture with data virtualization support for cloud-based
wireless sensor networks,” in Proc. VII Brazil. Symp. Comput. Syst. Eng.
(SBESC), 2017, pp. 199–204.
[7] C. Chang, S. N. Srirama, and M. Liyanage, “A service-oriented mobile
cloud middleware framework for provisioning mobile sensing as a
service,” in Proc. IEEE 21st Int. Conf. Parallel Distrib. Syst. (ICPADS),
Fig. 7. Average trust value in scene 2.
2015, pp. 124–131.
[8] S. Rani, S. H. Ahmed, R. Talwar, and J. Malhotra, “Can sensors collect
big data? An energy efficient big data gathering algorithm for WSN,”
IEEE Trans. Ind. Informat., vol. 13, no. 4, pp. 1961–1968, Aug. 2017.
[9] K. Nakashima, M. Yokoyama, Y. Taniyama, T. Yoshihisa, and T. Hara,
“S3 system: A system for sharing social sensor data and analytical
programs,” in Proc. Adjunct Int. Conf. Mobile Ubiquitous Syst. Comput.
Netw. Services, 2016, pp. 147–152.
[10] J. D. Zeng et al., “Research progress of sensor-cloud,” J. Comput. Res.
Develop., vol. 54, no. 5, pp. 925–939, 2017.
[11] I. Petri, J. Diaz-Montes, O. Rana, M. Punceva, I. Rodero, and
M. Parashar, “Modelling and implementing social community clouds,”
IEEE Trans. Services Comput., vol. 10, no. 3, pp. 410–422,
May/Jun. 2015.
[12] S. Chatterjee, R. Ladia, and S. Misra, “Dynamic optimal pric-
ing for heterogeneous service-oriented architecture of sensor-cloud
infrastructure,” IEEE Trans. Services Comput., vol. 10, no. 2,
pp. 203–216, Mar./Apr. 2015.
[13] T. Aamir, H. Dong, and A. Bouguettaya, “Trust in social-sensor
cloud service,” in Proc. IEEE Int. Conf. Web Services (ICWS), 2018,
pp. 359–362.
Fig. 8. Average trust value in scene 3. [14] L. F. Bilecki and A. Fiorese, “A trust reputation architecture for cloud
computing environment,” in Proc. IEEE 14th ACS Int. Conf. Comput.
Syst. Appl. (AICCSA), 2017, pp. 614–621.
[15] S. Bhatt and V. Krishnamurthy, “Controlled information fusion with risk-
VII. C ONCLUSION averse CVaR social sensors,” in Proc. IEEE 56th Annu. Conf. Decis.
Control (CDC), 2017, pp. 2605–2610.
This article proposed an RTCM that combines multisource [16] M. Rezvani, A. Ignjatovic, E. Bertino, and S. K. Jha, “A trust assessment
feedback and fog computing in SSC. First, the trust value framework for streaming data in WSNs using iterative filtering,” in Proc.
IEEE 10th Int. Conf. Intell. Sensors Sensor Netw. Inf. Process. (ISSNIP),
of the social node is evaluated. Second, the social sensor 2015, pp. 1–6.
needs to compute the global trust value through the feed- [17] K. Chard, S. Caton, O. F. Rana, and K. Bubendorfer, “Social cloud:
back of other nodes and FDs. At the same time, an algorithm Cloud computing in social networks,” in Proc. IEEE 3rd Int. Conf. Cloud
Comput., 2010, pp. 99–106.
for dynamically adjusting the weights of different trust fac- [18] S. Madria, V. Kumar, and R. Dalvi, “Sensor cloud: A cloud of virtual
tors is proposed, which can optimize the SSC system with sensors,” IEEE Softw., vol. 31, no. 2, pp. 70–77, Mar./Apr. 2013.

Authorized licensed use limited to: Istinye Universitesi. Downloaded on April 21,2023 at 01:21:43 UTC from IEEE Xplore. Restrictions apply.
5490 IEEE INTERNET OF THINGS JOURNAL, VOL. 7, NO. 6, JUNE 2020

[19] Z. Aiko, K. Nakashima, T. Yoshihisa, and T. Hara, “A social sensor visu- Min Zhang received the B.E. degree from Central
alization system for a platform to generate and share social sensor data,” South University, Changsha, China, in 2016. She is
in Proc. IEEE 42nd Annu. Comput. Softw. Appl. Conf. (COMPSAC), currently pursuing the M.S. degree in software engi-
Feb. 2018, pp. 628–633. neering with Guangxi University, Nanning, China.
[20] X. Yi, A. Bouguettaya, D. Georgakopoulos, A. Song, and J. Willemson, Her research interests focus on wireless sensor
“Privacy protection for wireless medical sensor data,” IEEE Trans. networks and cloud computing.
Depend. Secure Comput., vol. 13, no. 3, pp. 369–380, Feb. 2015.
[21] B. Guler et al., “Using social sensors for influence propagation in
networks with positive and negative relationships,” IEEE J. Sel. Topics
Signal Process., vol. 9, no. 2, pp. 360–373, Mar. 2015.
[22] F. H. Bijarbooneh, W. Du, E. C.-H. Ngai, X. Fu, and J. Liu, “Cloud-
assisted data fusion and sensor selection for Internet-of-Things,” IEEE
Internet Things J., vol. 3, no. 3, pp. 257–268, Jun. 2016.
[23] C. Pahl, N. E. Ioini, S. Helmer, and B. Lee, “An architecture pattern for
trusted orchestration in IoT edge clouds,” in Proc. 3rd Int. Conf. Fog
Mobile Edge Comput. (FMEC), 2018, pp. 63–70.
[24] F. Li, X. Wang, H. Chen, K. Sharif, and Y. Wang, “ClickLeak: Keystroke
leaks through multimodal sensors in cyber-physical social networks,”
IEEE Access, vol. 5, pp. 27311–27321, 2017.
[25] O.-J. Lee, H. L. Nguyen, J. E. Jung, T.-W. Um, and H.-W. Lee, “Towards
ontological approach on trust-aware ambient services,” IEEE Access,
vol. 5, pp. 1589–1599, 2017.
[26] T. Wang et al., “A comprehensive trustworthy data collection approach
in sensor-cloud system,” IEEE Trans. Big Data, early access.
[27] G. Zhang, T. Wang, G. Wang, A. Liu, and W. Jia, “Detection of hid-
den data attacks combined fog computing and trust evaluation method
in sensor-cloud system,” Concurrency Comput. Pract. Exp., to be
published.
[28] J. Yuan and X. Li, “A reliable and lightweight trust computing mecha- Victor C. M. Leung (Fellow, IEEE) received the
nism for IoT edge devices based on multi-source feedback information B.A.Sc. (Hons.) degree in electrical engineering
fusion,” IEEE Access, vol. 6, pp. 23626–23638, 2018. from the University of British Columbia (UBC),
[29] X. Huang, R. Yu, J. Kang, and Y. Zhang, “Distributed reputation Vancouver, BC, Canada, in 1977, and was awarded
management for secure and efficient vehicular edge computing and the APEBC Gold Medal as the Head of the grad-
networks,” IEEE Access, vol. 5, pp. 25408–25420, 2017. uating class with the Faculty of Applied Science,
[30] A. Jøsang, R. Ismail, and C. Boyd, “A survey of trust and reputation and the Ph.D. degree in electrical engineering from
systems for online service provision,” Decis. Support Syst., vol. 43, no. 2, UBC in 1982, through a Canadian Natural Sciences
pp. 618–644, 2007. and Engineering Research Council Postgraduate
[31] K. Govindan and P. Mohapatra, “Trust computations and trust dynamics Scholarship.
in mobile adhoc networks: A survey,” IEEE Commun. Surveys Tuts., He is currently a Distinguished Professor of com-
vol. 14, no. 2, pp. 279–298, 2nd Quart., 2011. puter science and software engineering with Shenzhen University, Shenzhen,
[32] S. E. Chang, A. Y. Liu, and W. C. Shen, “User trust in social networking China. He is also an Emeritus Professor of electrical and computer engineering
services: A comparison of Facebook and LinkedIn,” Comput. Human and the Director of the Laboratory for Wireless Networks and Mobile Systems
Behav., vol. 69, pp. 207–217, Apr. 2017. with the University of British Columbia (UBC), Vancouver, BC, Canada. He
has coauthored more than 1300 journal/conference papers and book chapters.
His research is in the broad areas of wireless networks and mobile systems.
Dr. Leung received the IEEE Vancouver Section Centennial Award,
the 2011 UBC Killam Research Prize, the 2017 Canadian Award for
Telecommunications Research, and the 2018 IEEE TCGCC Distinguished
Junbin Liang received the B.E. and M.S. degrees Technical Achievement Recognition Award. He coauthored papers that won
from Guangxi University, Nanning, China, in 2000 the 2017 IEEE ComSoc Fred W. Ellersick Prize, the 2017 IEEE Systems
and 2005, respectively, and the Ph.D. degree from Journal Best Paper Award, the 2018 IEEE CSIM Best Journal Paper Award,
Central South University, Changsha, China, in 2010. and the 2019 IEEE TCGCC Best Journal Paper Award. He is named in
He is currently a Professor with Guangxi the current Clarivate Analytics list of “Highly Cited Researchers.” He is
University. He is currently a Visiting Professor with serving on the Editorial Boards for the IEEE T RANSACTIONS ON G REEN
the University of British Columbia, Vancouver, BC, C OMMUNICATIONS AND N ETWORKING, the IEEE T RANSACTIONS ON
Canada. His research interests include sensor-cloud C LOUD C OMPUTING, IEEE ACCESS, IEEE N ETWORK, and several other
systems, fog computing, and distributed computing. journals. He is a Fellow of the Royal Society of Canada, the Canadian
Academy of Engineering, and the Engineering Institute of Canada.

Authorized licensed use limited to: Istinye Universitesi. Downloaded on April 21,2023 at 01:21:43 UTC from IEEE Xplore. Restrictions apply.

You might also like