Social acquaintance based routing in Vehicular Social Networks

Accepted Manuscript Social acquaintance based routing in Vehicular Social Networks Azizur Rahim, Tie Qiu, Zhaolong Ning, Jinzhong Wang, Noor Ullah, Amr Tolba, Feng Xia

PII: DOI: Reference:

S0167-739X(17)30432-6 http://dx.doi.org/10.1016/j.future.2017.07.059 FUTURE 3589

To appear in:

Future Generation Computer Systems

Received date : 16 March 2017 Revised date : 3 July 2017 Accepted date : 23 July 2017 Please cite this article as: A. Rahim, T. Qiu, Z. Ning, J. Wang, N. Ullah, A. Tolba, F. Xia, Social acquaintance based routing in Vehicular Social Networks, Future Generation Computer Systems (2017), http://dx.doi.org/10.1016/j.future.2017.07.059 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Social Acquaintance Based Routing in Vehicular Social Networks Azizur Rahima , Tie Qiua,∗, Zhaolong Ninga , Jinzhong Wanga,b , Noor Ullaha , Amr Tolbac,d , Feng Xiaa a School

of Software, Dalian University of Technology, Dalian 116620, China. of Sport Information Technology, Shenyang Sport University, Shenyang 110102, China. c Riyadh Community College, King Saud University, Riyadh 11437, Saudi Arabia d Mathematics and Computer Science Department, Faculty of Science, Menoufia University, Egypt b Department

Abstract The concept of Internet of Things (IoT) provides us the opportunity to interconnect different objects with the communication and processing capabilities for a diverse range of applications. Recently, Vehicular Social Networks (VSNs) have been introduced through the combination of relevant concepts from two primary disciplines, i.e., social networks and Vehicular Ad-hoc Networks (VANETs). Inspired from the social acquaintance in our daily life, we present a Social Acquaintance based Routing Protocol (SARP) for VSNs, which collectively consider three social feature metrics to make a forwarding decision. Proposed protocol aims to reduce End-to-End delay and improve packet delivery ratio in VSNs. Additionally, SARP overcomes the shortcoming of topology based routing and optimum local situation of geographically based routing protocols by considering the global and local community acquaintance of nodes. We performed extensive simulations under constant node density with different mobility speed and constant speed with varying node density to study the effect of node mobility speed and density on end-to-end delay and packet delivery ratio. The simulation results show that SARP outperforms GPSR by 22% and 26% in terms of end-to-end delay and packet delivery ratio respectively. Also, ∗ Corresponding

author Email address: [email protected] (Tie Qiu)

Preprint submitted to Journal of FGCS

July 3, 2017

SARP outperforms AOVD in terms of end-to-end delay. Keywords: Vehicular Social Networks, Cyber-Physical Systems, Internet of Things, Intelligent Transport Systems, Communication Architecture, Reliable Data Delivery

1. Introduction Vehicular Ad-hoc Network (VANET) is a unique form of ad-hoc networks, created by vehicles with restricted mobility on roads and streets with data processing and wireless communication capabilities. These vehicles either com5

municate directly or through Road Side Units (RSUs) utilizing different communication infrastructures, including 3G/4G, WiFi, etc. Network resources are used to access and obtain data from other networks. VANETs have the capabilities to be deployed in different operational environments for a diverse range of applications. In other words, VANETs are sparse ad-hoc networks formed by

10

vehicles to communicate opportunistically on contact. Due to its easy to deploy nature and diverse possible range of applications in a pervasive environment, VANETs have attracted the attention of not only academicians but also from industrial leaders. Typical applications of VANETs include intelligent traffic control systems, collision warnings, active navigation systems, and passenger

15

entertainment/comfort services. Quite recently, the research community in the field of communication and technology has been greatly attracted by Social Network Analysis (SNA) and its applications to design new algorithms and protocols for socially aware networking, such as Mobile Social Networks (MSNs), Vehicular Social Networks

20

(VSNs), and Internet of Things (IoT). The motivation behind the inheritance of social networks in communication networks is that all entities have interdependencies which relate them to each other in one way or another. These interdependencies include physical contact, mutual interest, similarity, group participation, and much more. SNA not only helps us to define these inter-

25

dependencies but can also be exploited to correlate them. Social networking

2

is the grouping of entities based on their social interdependencies to improve the effectiveness and efficiency of network services [1, 2]. Some algorithms and protocols have been proposed so far, based on the social interaction of entities (nodes, devices, people, and systems), which has shown that social interaction30

s of nodes can be exploited to achieve high efficiency and effectiveness within mobile communication systems [3]. The inheritance of social networks has a lot of potentials which can be exploited in a vehicular environment. Vehicular mobility and density are two of the main factors that significantly influence the communication in the vehicular

35

environment. Mobility directions and speed limitations on public roads restrict vehicular mobility. However, vehicular density is affected by the number of vehicles moving on a specific route within a particular period of the day/week. During rush hour, traffic jams result in higher vehicular density. Density variations during different times of day/week characterize vehicular dynamic network

40

topology, making data sharing and communication a challenging task in the vehicular environment. Besides, vehicular mobility depends on drivers’ interests, behaviors, and routine. For examples, on weekdays, the commuters often repeat the same path at the same time to and from the same destinations, such as work offices, universities, schools, etc. Similarly, different destinations, includ-

45

ing parks, cinemas, shopping malls, etc., are the most visited destinations for commuters on weekends. These vehicles are controlled by humans exhibiting some social behaviors. However, unlike Mobile Social Networks (MSNs), vehicular mobility is restricted to roads and streets. These vehicles communicate on opportunistic encounter within each other’s communication range.

50

In a vehicular environment, commuters encounter other vehicles/passengers on their trajectories with a similar profile moving on the same street and facing the same traffic conditions. These commuters can share valuable information along the roads and may include traffic information, personal information, and vehicle information. Socially-aware techniques are applied to exploit the so-

55

cial similarity of nodes to improve data delivery services and connectivity in a vehicular environment. Inheritance of social networks into vehicular network3

s has been identified as one of the most efficient solutions for a diverse range of applications. These applications can be mainly divided into four categories; (a) Safety-based applications, (b) Convenience-based Applications, (c) Comfort60

based applications, and (d) Entertainment-based applications.

Social Ties

Vehicular Social Networks

Dynamic Network Topology

Centralized Communication

Road Side Units

Social Networks

Vehicular Ad-hoc Networks

Online Social Behaviors

Oprtunistic Contacts Social Services

DSRC Wi-Fi Bluetooth

Figure 1: Vehicular Social Networks

After homes and offices, regular citizens spend considerable time in vehicles based on their daily schedules. Different from MSNs, where human beings interact with each other using their smart devices, network entities in VSNs are heterogeneous including OBUs, RSUs, as well as drivers’ and passengers’ smart 65

devices. As shown in Fig. 1, VSNs incorporate relevant features and concepts from two different fields namely VANETs and social networks. VANETs provide the underlying communication network infrastructure, whereas social networks contribute to the social knowledge of entities. VSNs can be deployed in different scenarios such as urban and highway, where it can be either centralized

70

or distributed in nature. Based on the underlying communication architecture, three types of communication relations are found in VSNs: human-to-human, humans and machines, and machines-to-machines. Similarly, unique characteristics of VSNs and particular applications environment distinguish VSNs from traditional MSNs and VANETs.

75

Recently, many efforts have been made to propose different algorithms and architectures that incorporate the concepts of social networks to support a va-

4

riety of mobile social applications [4, 5, 6, 7]. However, these solutions have mainly focused on Delay Tolerant Networks (DTNs) and MSNs. Quite recently, a few works addressed the incorporation of social networks in vehicular net80

works [8, 9, 10, 11, 12, 13, 14, 15, 16]. It has been demonstrated that the incorporation of social networks into vehicular networks positively influence the services and communication. However, some technical challenges arise that need to be addressed. Dynamic network topology and high vehicular mobility make application development in VSNs a challenging task. Similarly, opportunistic

85

and short-term communication contacts in VSNs demand for efforts to develop new algorithms to accommodate different applications. Data dissemination, routing, mobility modeling, simulation, privacy, and security are other issues which need to be considered. Social relationships are relatively stronger and stable as compared to com-

90

munication links between mobile nodes, which can be exploited to enhance data transmission [17]. However, some questions arise considering the vehicular environment. Does it has social properties? Is there any permanence of social behaviors in vehicular mobility? How effective is to examine social behavior in the vehicular environment? Some works exist in literature to answer these

95

questions [18, 19, 20]. Authors in [19] analyzed real data sets and found that vehicular mobility shows small world phenomenon. Similarly, the study also indicates the existence of communities with similar interests. Social ties and social community are widely used concepts to enhance data dissemination in socially-aware networks. Community members are expected to communicate

100

more frequently as compared to other nodes out of the same community. However, for inter-community communication, global knowledge of popularity of nodes is required. Similarly, for intra-community data dissemination and forwarding, social ties can be exploited to enhance data delivery. Improving QoS in VSNs is one of the challenging tasks due to higher mobil-

105

ity and dynamic network topology. Information sharing and prediction based algorithms in VSNs is vulnerable to packet drop and privacy attacks. Similarly, resources utilization and traffic congestion are also the key factors to consider 5

while developing algorithms/protocols for VSNs. In short, to overcome these challenges efforts are needed with appropriate mechanism to deal with highly 110

dynamic nature of VSNs and intermittent connectivity to enhance data delivery. In this paper, we consider social characteristics of nodes in the vehicular environment and propose a novel protocol for data dissemination in VSNs named as SARP (Social-Acquaintance based Routing Protocol). This protocol exploits the social metrics in a fully distributed manner using characteristics of ad-hoc

115

and delay tolerant networks. This work differs from existing works as follows. First, community acquaintance is defined to quantify the global and local importance of intermediate nodes. Second, community acquaintance is jointly considered with node centrality and activeness with minimum possible overhead to select the next relay node. Third, we propose a new protocol (SARP to

120

enhance data delivery in VSNs with minimized end-to-end delay and improved packet delivery ratio. Finally, extensive simulations are performed to evaluate the performance of proposed protocol in VSNs. The rest of the paper is organized as follows. Related work on socially-aware routing and dissemination protocols are presented in Section 2 with motivation

125

for our proposed model. In Section 3, we provide an overview of proposed protocol and social features which are considered to develop the forwarding decision. Section 4 describes the simulation setup. Before concluding remarks in Section 6 with future work, simulation results are presented in section 5.

2. Related Works and Motivation 130

Data dissemination in VSNs is one of the main challenges due to highly dynamic nature and intermittent connectivity in the vehicular environment. Unlike wireless sensor networks, where sensor nodes directly transmit sensed data to the base station [21], in VSNs, an end-to-end path does not exist, and data is forwarded in store-carry-and-forward fashion. Different features and parame-

135

ters may be considered to enhance data dissemination in VSNs. In [22] authors have profoundly studied different data dissemination approaches in VSNs. In

6

the study, it is shown that social behaviors and mobility pattern of nodes1 are being exploited to design content dissemination protocols. Similarly, inextricable properties of mobiles devices and their users (i.e., their mobility pattern140

s, interdependent social behaviors) are being utilized in other communication networks, such as Pocket Switched Networks (PSNs) [23], Delay Tolerant Networks (DTNs) [24], Socially Aware networking (SAN) [25], and Opportunistic Networks (OppNets) [26]. SAN provides a ground to inherit social properties of nodes into the vehicular

145

environment, where a group of individuals having common interests may share information along the roads. Socially based protocols help to identify sociallysimilar nodes based on common interest, community affiliation, similar route, and destination. Stability of social ties and less frequent variation in social relationships are the key initiatives to inherit social properties in the vehicular

150

environment to enhance data dissemination. Recently, nodes’ social properties are extensively analyzed and considered by the research community to design new routing protocols for socially-aware networks [27, 28, 25, 29]. Quite recently, Xia et al. in [25] have presented a novel interest-based forwarding scheme for socially aware networks. The proposed protocol, BEEINFO

155

(Artificial BEE Colony inspired INterest-based FOrwarding), exploits food foraging behavior of bees to record information of different communities passing through. Bees’ awareness capability has been introduced in VSNs considering three distinct areas, i.e., shopping mall, hospital, and school. Vehicles along their routes collect community information and estimate community density.

160

Community density is defined as the number of nodes in a particular community. Individuals with similar and shared interests build communities, and it is understood that members of the same community meet more often than members out of the community. LABEL [30] is one of the well-known routing protocols based on community concepts to deliver messages only to the member 1 With

the term “nodes” we mean vehicles and human beings participating in data dissem-

ination in VSNs

7

165

nodes of destination community. Similarly, following the same idea, BUBBLE RAP [31] enhance routing performance considering central nodes with community detection, resulting in high cost for maintenance and construction of socio-aware overlay. To reduce end-to-end delay and achieve higher delivery ratio is a challenging

170

goal to be achieved in VSNs. Gu et al. in [32] proposed a socially-aware routing protocol with fuzzy logic to achieve this aim. This fuzzy logic algorithm not only depends on traditional greedy approach but also exploits the social behaviors of nodes, i.e., centrality. Cunha et al. in [33] considered the daily variation of traffic flow and social ties among vehicles to propose a data dissemination protocols in

175

vehicular networks. The social metrics considered by authors include clustering coefficient and node degree to select the best node for data dissemination in the vehicular environment. Besides, some other works in the literature [34, 35, 36, 37, 38, 39] combine community awareness with other social metrics, i.e., similarity, node centrality, and betweenness to enhance inter-community and

180

intra-community data delivery. These protocols consider historical encounter of nodes to predict the contact probability for data forwarding. Apart from social network analysis, based on underlying communication architecture of VSNs, the traditional performance metrics including delivery ratio, bandwidth usage, and data delivery delay affect the QoS of VSNs. However,

185

some applications, i.e., traffic, safety, and emergency information dissemination require short data delivery delay. On the other side, some applications demand higher data delivery as compared to reduced end to end delay such as entertainment applications. However, exploring the efficient modeling of socially aware metrics in VSNs is a challenging task. Efforts are still needed to exploit

190

application-oriented social metrics to design data dissemination protocols for VSNs.

8

3. Overview of SARP Reliable and efficient data dissemination in the vehicular environment is one of the most extensively studied issues in literature. Higher mobility and dy195

namic network topology are the key factors making data dissemination in the vehicular environment as one of the challenging tasks and have attracted the research community. Traditional routing protocols for VANETs, MANETs, and other large-scale networks often use the geographical information or network topology information for making a routing decision [40, 41]. Routing protocols

200

based on network topology can’t be applied to VSNs due to its dynamic network topology. Similarly, routing protocols based on geographical information can be helpful in highway scenarios but might end up in local optimum in urban areas. On the other hand, the social relationships among nodes in the vehicular environment are relatively stronger and stable as compared to node mobility.

205

In this paper, we propose a multi-dimensional routing protocol, which exploits the nodes social characteristics such as community acquaintance, node centrality, and activeness to route a message from a source to destination in VSNs. Nodes belong to different communities based on their interest and geographical location. Compared to existing social-aware routing protocols, SARP not on-

210

ly considers the current values of parameters but also examines the historical values for decision making which makes it more flexible and reliable. Similarly, global importance of relay nodes is considered to overcome the local optimum problem of geographically based routing protocols. We consider an application scenario for information sharing and develop a

215

mechanism to calculate a priority value of next relay node. The message is forwarded in store-and-carry fashion from a source node to the destination node within the same or different community. People with the same interests or from the same community usually meet more often and have the greater probability to meet in future as compared to other members of communities or with different

220

interests. We consider three research groups (Group A, Group B, and Group C) with different research direction with mutual collaboration, where the members

9

of the same research group have greater interaction and probability to meet as compared to members of other research groups. If a member of Group A wants to deliver a file to someone in the same research group, he/she may directly hand 225

over the desired file personally or can deliver the file to a person who has greater community acquaintance as compared to him/her. Why don’t we consider node centrality to deliver the message/file as most of the previous work do? A person or node can have higher node centrality, but it may not be sufficient to deliver a message to the destination node as compared to the one who has greater inter-

230

action within the same community at the same time. Similarly, does it guaranty that a message will be delivered to the destination for sure by choosing a relay node with high community acquaintance and node centrality? To increase the probability of message delivery, we also consider node activeness. Consequently, a person who has higher community acquaintance, the high degree of centrality

235

and most active in the community can be the most suitable relay candidate to deliver a message. Similarly, a person who is active and has more connection within Group A but no or less acquaintance with members of Group B or Group C can’t be a suitable relay for information delivery from Group A to Group B or Group C. The global community acquaintance is required to route a message

240

from source to destination community during inter community communication. As shown in Fig. 2, if the source node s wants to send a packet to a destination node d outside community A, it can choose one of the nodes i, j, and k as a relay node. Degree centrality for node i, j, and k is the same (i.e., 5); however, the neighbors of j remain the same while the neighbor list of node j and k

245

include some transit nodes. This indicates that node i and k are more active as compared to node j. Besides, the interaction of node i is limited to members of community A; however, node k has an interaction with nodes from community C, which may increase the probability of message delivery towards destination community. In other words, node k has higher global community acquaintance

250

as compared to node i and j. Following this, we consider these three social feature metrics to define priority value for data forwarding to choose next relay node. The following subsections describe how we estimate these social feature 10

Figure 2: Community acquaintance

metrics. 3.1. Community Acquaintance 255

Community acquaintance is the global and local measurement of node’s popularity in a network. Higher global community acquaintance increases the probability of inter-community packet forwarding, and higher local community acquaintance increases the probability of intra-community forwarding. In our proposed method we use the following equation to estimate nodes’ community

260

acquaintance.

CAcqni = αCAcqlocalni + (1 − α)CAcqglobal ni

(1)

Where α is a control variable, and the values are used in our proposed model are 1 and 0 for intra-community and inter-community packet forwarding respectively. Global community acquaintance is required for inter-community forwarding, but once a packet reaches destination community, the local com265

munity acquaintance is considered for packet forwarding. Node encounter is measured when two nodes are within the communication range of each other, and a hello message is successfully exchanged. 11

CAcqlocalni =

N X

Encounterni ,nj

(2)

j=1

Equation 2 is used to calculate the local community acquaintance of node ni . Where Encounterni ,nj = 1, if a hello message is successfully exchanged between 270

ni and nj and ni and nj belong to the same community. N is the total number of nodes in the same community. Similarly, equation 3 is used to calculate the global community acquaintance of node ni . Where Encounterni ,nk = 1, if a hello message is successfully exchanged between ni and nk and ni and nk do not belong to the same community.

275

M is the total number of nodes across all communities (total number of nodes in the network).

CAcqglobalni =

M X

Encounterni ,nk

(3)

k=1

Individuals with similar and shared interests build communities. In this paper, we assign community numbers to nodes based on their interest and mobility similarity, which is similar to LABEL [30]. Initially, community acquaintance of 280

node ni (global and local) remains 1 and increases once a node encounters other nodes. Values of global and local community acquaintance depend upon the number of communities and number of members in each community respectively. For example, in a network topology with 100 nodes, four communities, and 25 members in each community, the highest possible value for global and local

285

community acquaintance can be 4 and 25 respectively. When a node encounters a node from the same community, excluding repetition, local community acquaintance is increased, but global community acquaintance remains the same. On the hand, if a node encounters a node from the different community then its global acquaintance increases.

12

290

3.2. Social Activeness The network topology of VSNs is highly dynamic, and nodes’ neighbor list frequently alters with time. As a general concept from social life, a person who meets more new people in his/her daily routine is considered to be more socially active as compared to one who keeps his/her interaction to a group of limited

295

people. Following this concept, the social activeness of node ni at time t can be calculated as

SActni (t) = 1 −

Nni (t − △t) ∩ Nni (t) Nni (t − △t) ∪ Nni (t)

(4)

where Nni (t) and Nni (t − △t) represent the number of current neighbors of node ni at time t and previous number of neighbors of node ni at time t − △t respectively. Value of △t is not constant; it depends upon the current and 300

previous value of t at which the value SActni (t) is calculated.

SActni ← βSActni (t − △t) + (1 − β)SActni (t)

(5)

Where β is a smoothing factor to consider the current value of SActni at time t and the previous value of SActni at time t−△t. In our proposed method, the value of β is set to 0.5 to consider the equal significance of current and the previous value of SActni . A greater value for SActni indicates that ni is more 305

active in the network probability to meet new members is high. Consequently, increases the probability of packet delivery. 3.3. Degree Centrality Centrality is the relative measurement of nodes’ importance in a social network and can be measured with different methods observed from social network

310

analysis. However, we consider degree centrality of a node to measure its capability of direct links with its neighbors in VSNs. A Higher value of degree centrality of nodes indicates the stronger capability of interaction with other nodes in VSNs and increases the probability of packet delivery. Gu et al. in 13

[32] consider a similar concept of centrality and node activeness; however, our 315

consideration and use of these metrics are entirely different. For a node ni , degree centrality at time t is calculated as

DCni (t) =

N X

Eni ,nj

(6)

k=1

Where Eni ,nj = 1 if there exists a direct communication link between nodes ni and nj . Similar to social activeness of a node, degree centrality is periodically updated with a smoothing factor β = 0.5 so that the current and previous value 320

of DCni are equally weighted.

DCni (t) ← βDCni (t) + (1 − β)DCni (t − △t)

(7)

3.4. SARP Nodes within their communication range communicate in pair-wise fashion and can exchange date packets. Nodes on the encounter with other nodes update their neighbor list and count their direct contacts with other nodes to update 325

values of SAct and DC. Thus, nodes on the encounter with other nodes also keep track of their community acquaintance (CAcq). Our proposed protocol, SARP, collectively combines the above three social metrics to calculate the priority value using the following equation.

P riorityni = 1 −

CAcqni

1 + SActni + DCni (t)

(8)

If node ni encounters node nj , it compares the P riority values, and if the 330

P riority of node nj is greater than node ni , node ni will forward the message to node nj otherwise it will buffer and carry the message until a new node with higher P riority is encountered.

14

4. Simulation setup In this section, we present the simulation setup to evaluate the performance 335

of SARP. We used VanetMobiSim to generate vehicular mobility traces and widely-adopted network simulator NS2 to assess the performance of our proposed protocol. We compare our proposed protocol with two widely accepted routing protocols for VANETs and MANETs, i.e., GPSR and AODV. For mobility traces generation, we considered, and area of 2000 x 3000 m2 area near

340

Xinghai Square

2

as shown in the Fig. 3. Other parameters that we used for

our simulation are illustrated in Table 1. For implementation and simulation, we made the following assumption in this paper. • All nodes are fully cooperative and cooperate in data forwarding; • Nodes are categorized into different communities based on interest and 345

mobility pattern.

Figure 3: Simulation area

2 Xinghai

Square is a city square in Dalian City,

https://en.wikipedia.org/wiki/Xinghai_Square

15

Liaoning Province,

China.

Table 1: Simulation parameters

Parameter

Value

Network Simulator

ns-2.34

Simulation Time

300s

Number of Nodes

100, 150, 200, 250, 300

Simulation Area

2000*3000 m2

Trip Generation

Random trip generation

Traffic Source/Destination

Random

Packets Generation Rate

5 Packets

CBR interval

0.20 s

MAC Protocol

IEEE 802.11p

Speed of Vehicles

30-70 KM/H

Transmission Range

80 meters

5. Results and discussion In this paper, we investigate the average End-to- End Delay and Packet delivery ratio with variation in node density with constant speed and with constant node density with varying speed. This section presents the performance 350

analysis of SARP in comparison with AODV and GPSR. 5.1. End-to-end Delay The average end-to-end delay is one of the important factors to compare the performance of any routing protocol in communication networks. The average end-to-end delay is the measurement of all possible delays from a source node to

355

destination node, i.e., buffering, propagation, transmission, and retransmission delay. End-to-end delay is the time taken by data packets delivered successfully from a source node to destination node in a network. Average end-to-end delay can be calculated as the mean of end-to-end delays of all successfully delivered packets. We calculated the end-to-end delay as the time difference of the time

360

at which packet was transmitted at the source and the time at which it was 16

delivered at destination. In our analysis, we analyzed end-to-end delay for different node density and different node speed. 5.1.1. End-to-end delay vs. number of nodes Fig. 4 shows the average end-to-end delay vs. the number of vehicles with 365

constant node speed. We studied the end-to-end delay of our proposed protocol in comparison with AODV and GPSR. The result shows that end-to-end delay of AODV is much higher with the lower number of nodes in the network as compared to GPSR and SARP. The reasons for higher end-to-end delay of AODV are the frequent breaking of links and re-establishment of new connections.

370

However, with increasing number of nodes, end-to-end delay is reduced, but with further increase in the number of nodes, end-to-end delay starts increasing due to increased routing overhead. On the other side, end-to-end delay of GPSR is lower than AODV. However, end-to-end delay increases with increased packet overhead caused by hello beacons in dense networks. End-to-end delay

375

of SARP is greater than GPSR with lower node density; however, it decreases with increasing node density. With increasing node density, the probability of nodes encounters increases. Consequently, the probability of data delivery increases resulting in lower end-to-end delay. Thus, SARP outperforms GPSR and AODV in the dense network with minimized end-to-end delay.

380

5.1.2. End-to-end delay vs. Speed Fig. 5 shows the end-to-end delay vs. node speed with a constant number of nodes. We kept the number of nodes consistent to analyze the effect of node mobility on end-to-end delay. At lower speed with constant node density, SARP and GPSR outperform AODV. However, with increasing node speed

385

end-to-end delay of AODV starts decreasing but with the further increasing, it starts rising again. This degradation is caused by broken link with high node mobility. On the other hand, end-to-end delay of GPSR increases with increase node speed. To compare with AODV and GPSR, SARP outperforms concerning end-to-end delay. Initially, the probability to of node encounters

17

AODV

300

GPSR

275

SARP

250

End-to-End Delay (ms)

225 200 175 150 125 100 75 50 25 0

100

150

200

250

300

Number of Vehicles

Figure 4: End-to-end delay vs. number of nodes

390

increases with increasing node speed. Consequently, end-to-end delay starts decreasing. However, a further increase in speed results in increased end-to-end but still outperforms both AODV and GPSR. The reason for this increase is the link failure and re-establishment of links at high speed. 5.2. Packet delivery ratio

395

Packet delivery ratio is another important factor to measure the performance of protocols in networks. Packet delivery ratio depends on various parameters set for simulation which includes packet size, transmission range, the number of nodes, and node mobility. Packet delivery ratio is the ratio of the number of packets successfully delivered to a destination node to the number of packets

400

sent by a source node. The performance is considered to better if the packet delivery ratio is high. In our analysis, we compared the performance of SARP in terms of packet delivery ratio with different node density and varying node speed.

18

AODV

120

GPSR

110

SARP

End-t-End Dealy (ms)

100 90 80 70 60 50 40 30 20 10

30

40

50

60

70

Speed

Figure 5: End-to-end delay vs. node speed

5.2.1. Packet delivery ratio vs. node density 405

Packet delivery ratio of SARP in comparison with AODV and GPSR with respect to the number of nodes under constant speed is shown in Fig. 6. It is shown in the figure that with low node density, packet delivery ratio of SARP is better than GPSR but lower than AODV; however, packet delivery ratio of AODV starts decreasing with increasing number of nodes while packet delivery

410

ratio SARP increases with increasing number of nodes. As the number of nodes increases, the probability of node encounter increases resulting in increased probability of relay nodes with a high priority value. On the other hand, packet delivery ratio of AODV and GPSR decreases with increasing number of nodes as routing overhead increases in dense networks.

415

5.2.2. Packet delivery ratio vs. node speed Fig. 7 shows the effect of varying speed on packet delivery ratio with constant node density. As shown in the figure, increasing node speed has reversed effect on packet delivery ratio; however, SARP still outperforms AODV and GPSR. Initially, with the increase in speed performance of all protocols increases

420

in terms of packet delivery ratio, however, with further increase in speed, the 19

AODV

100

GPSR SARP

90

Packet Delivery Ratio

80 70 60 50 40 30 20 10 0

100

150

200

250

300

Nodes

Figure 6: Packet delivery ratio vs. number of nodes

performance of AODV and GPSR starts immediately degrading while SARP shows better performance in terms of packet delivery ratio. Link failure at high speed causes packet loss due to which AODV performance is degraded. Similarly, very short contact duration of nodes also results in packet loss degrading 425

the performance of GPSR and SARP. In Case if GPSR, sometimes packets may be forwarded in the wrong direction that leads to low data delivery ratio. 5.3. Impact of community acquaintance The three social metrics we considered for our proposed model are independent. However, all these social metrics show the importance of a node in

430

the network and play an important role in decision-making procedure to select an intermediate node for data delivery. We investigated the importance of community acquaintance in the network and performed some simulation to see the impact of community acquaintance on end-to-end delay and packet delivery ratio. We considered only degree centrality and social activeness for decision

435

making (SARP-DC). As shown in the Fig. 8, SARP-DC performs better than AODV and GPSR in terms of end-to-end delay. However, it’s end-to-end delay is greater than SARP. 20

SARP

100

GPSR AODV

90

Packet Delivery Ratio

80 70 60 50 40 30 20 10 0

30

35

40

45

50

55

60

65

70

Speed

Figure 7: Packet delivery ratio vs. node speed

In this case, the possible reason for this delay is that data packets are forwarded to nodes with higher degree centrality but low or no community acquaintance. 440

On the other side, in SARP data packets are sent to the node with possible community acquaintance resulting in lower end-to-end delay. Community acquaintance does not only help to improve the performance in terms of end-to-end delay but also increases the probability of packet delivery. As it is shown in Fig. 9, the packet delivery ratio of SARP-DC is degraded

445

as compared to AODV, GPSR and SARP. Some of the packets are forwarded to nodes with low or no community acquaintance and are dropped as the TTL expires before reaching the destination community. Consequently, from this analysis, it is concluded that community acquaintance not only helps to improve performance in terms of end-to-end delay but also data delivery ratio.

450

6. Conclusion VSNs being a bridge between vehicular networks and social networks have attracted the research community due to its diverse range of applications. In this paper, we have presented a novel routing protocol exploiting social feature 21

AODV

300

GPSR

275

SARP SARP-DC

End-to-End Delay (ms)

250 225 200 175 150 125 100 75 50 25 0

100

150

200

250

300

Number of Vehicles

Figure 8: End-to-end delay with Degree Centrality

metrics, opportunistic encounters and mobility pattern for collaborative data 455

delivery in VSNs. Traditional routing protocols based on network topology do not suit VSNs due to its highly dynamic nature. Similarly, geographically based routing protocols may result in the local optimum. The proposed protocol considers the opportunistic encounters of nodes with social feature metrics to quantify the global and local community acquaintance of nodes for selection of

460

relay nodes. We used VanetMobiSim for mobility trace generation considering a local area of 2000 * 3000 m2 . Extensive simulations were performed using vehicular mobility traces using NS2 to analyze the effect of node density and speed of vehicles on end-to-end delay and packet delivery ratio. We compared the per-

465

formance of our proposed protocol, SARP, with two widely accepted routing protocols, i.e., AODV and GPSR. Results have shown that overall SARP outperforms AODV and GPSR in terms of packet delivery ratio and end-to-end delay. However, packet delivery ratio of SARP decreasing with increasing speed but still outperforms AODV and GPSR.

470

The mobility pattern of real data sets may not be the same as generated

22

Packet Delivery Ratio

100 95 90 85 80 75 70 65 60 55 50 45 40 35 30 25 20 15 10 5 0

AODV GPSR SARP SARP-DC

100

150

200

250

300

Nodes

Figure 9: Packet delivery with Degree Centrality

by mobility generation tools. We intend to consider real data sets of vehicular mobility to analyze the performance as a future work. Besides, emergency warning systems in VSNs demand reduced end-to-end delay and to prioritize and schedule emergency messages. Thus, we will focus on packet scheduling 475

in VSNs. Furthermore, to enhance packet data delivery and reduce end-to-end delay we will focus on socially-aware multi-casting in VSNs with real vehicular mobility traces.

7. Acknowledgment This work is supported by National Natural Science Foundation of China 480

(61572106, 61502075 and 61672131), Dalian Science and Technology Planning Project (2015A11GX015 and 2015R054).

References [1] N. Kayastha, D. Niyato, P. Wang, E. Hossain, Applications, architectures, and protocol design issues for mobile social networks: A survey, Proceed485

ings of the IEEE 99 (12) (2011) 2130–2158. 23

[2] A. M. Ahmed, T. Qiu, F. Xia, B. Jedari, S. Abolfazli, Event-based mobile social networks: Services, technologies, and applications, IEEE Access 2 (2014) 500–513. [3] R. Fei, K. Yang, X. Cheng, A cooperative social and vehicular network and 490

its dynamic bandwidth allocation algorithms, in: 2011 IEEE Conference on Computer Communications Workshops (INFOCOM WKSHPS),, IEEE, 2011, pp. 63–67. [4] J. Li, H. Wang, S. U. Khan, A semantics-based approach to large-scale mobile social networking, Mobile Networks and Applications 17 (2) (2012)

495

192–205. [5] H. Zhou, J. Chen, J. Fan, Y. Du, S. K. Das, ConSub: incentive-based content subscribing in selfish opportunistic mobile networks, IEEE Journal on Selected Areas in Communications 31 (9) (2013) 669–679. [6] B. Jedari, L. Liu, T. Qiu, A. Rahim, F. Xia, A game-theoretic incentive

500

scheme for social-aware routing in selfish mobile social networks, Future Generation Computer Systems 70 (2017) 178 – 190. [7] F. Xia, L. Liu, B. Jedari, S. K. Das, PIS: a multi-dimensional routing protocol for socially-aware networking, IEEE Transactions on Mobile Computing 15 (11) (2016) 2825–2836.

505

[8] I. Lequerica, M. G. Longaron, P. M. Ruiz, Drive and share: efficient provisioning of social networks in vehicular scenarios, IEEE Communications Magazine 48 (11) (2010) 90–97. [9] A. Kchiche, F. Kamoun, Access-points deployment for vehicular networks based on group centrality, in: 3rd International Conference on New Tech-

510

nologies, Mobility and Security, IEEE, 2009, pp. 1–6. [10] N. Abbani, M. Jomaa, T. Tarhini, H. Artail, W. El-Hajj, Managing social networks in vehicular networks using trust rules, in: IEEE Symposium on Wireless Technology and Applications (ISWTA), IEEE, 2011, pp. 168–173. 24

[11] I. Lequerica, M. G. Longaron, P. M. Ruiz, Drive and share: efficient pro515

visioning of social networks in vehicular scenarios, IEEE Communications Magazine 48 (11) (2010) 90–97. [12] S. Smaldone, L. Han, P. Shankar, L. Iftode, RoadSpeak: enabling voice chat on roadways using vehicular social networks, in: Proceedings of the 1st Workshop on Social Network Systems, ACM, 2008, pp. 43–48.

520

[13] Y. Zhang, J. Zhao, G. Cao, Roadcast: a popularity aware content sharing scheme in vanets, ACM SIGMOBILE Mobile Computing and Communications Review 13 (4) (2010) 1–14. [14] D. Popovici, M. Desertot, S. Lecomte, T. Delot, A framework for mobile and context-aware applications applied to vehicular social networks, Social

525

Network Analysis and Mining 3 (3) (2013) 329–340. [15] J. Qin, H. Zhu, Y. Zhu, L. Lu, G. Xue, M. Li, Post: Exploiting dynamic sociality for mobile advertising in vehicular networks, IEEE Transactions on Parallel and Distributed Systems 27 (6) (2016) 1770–1782. [16] Z. Ning, F. Xia, N. Ullah, X. Kong, X. Hu, Vehicular Social Networks:

530

Enabling Smart Mobility, IEEE Communications Magazine 55 (5) (2017) 49–55. [17] M. R. Schurgot, C. Comaniciu, K. Jaffres-Runser, Beyond traditional DTN routing: social networks for opportunistic communication, IEEE Communications Magazine 50 (7) (2012) 155–162.

535

[18] F. D. Cunha, A. C. Vianna, R. A. F. Mini, A. A. F. Loureiro, How effective is to look at a vehicular network under a social perception?, in: 2013 IEEE 9th International Conference on Wireless and Mobile Computing, Networking and Communications (WiMob), 2013, pp. 154–159. [19] F. D. Cunha, A. C. Vianna, R. A. Mini, A. A. Loureiro, Is it possible to

540

find social properties in vehicular networks?, in: 2014 IEEE Symposium on Computers and Communications (ISCC), IEEE, 2014, pp. 1–6. 25

[20] X. Liu, Z. Li, W. Li, S. Lu, X. Wang, D. Chen, Exploring social properties in vehicular ad hoc networks, in: Proceedings of the Fourth Asia-Pacific Symposium on Internetware, ACM, 2012, p. 24. 545

[21] G. Han, A. Qian, J. Jiang, N. Sun, L. Liu, A grid-based joint routing and charging algorithm for industrial wireless rechargeable sensor networks, Computer Networks 101 (2016) 19 – 28, industrial Technologies and Applications for the Internet of Things. [22] F. Mezghani, R. Dhaou, M. Nogueira, A.-L. Beylot, Content dissemina-

550

tion in vehicular social networks: taxonomy and user satisfaction, IEEE Communications Magazine 52 (12) (2014) 34–40. [23] P. Hui, A. Chaintreau, R. Gass, J. Scott, J. Crowcroft, C. Diot, Pocket Switched Networking: Challenges, feasibility and implementation issues, in: Workshop on Autonomic Communication, 2005, pp. 1–12.

555

[24] K. Fall, A delay-tolerant network architecture for challenged internets, in: Proceedings of the 2003 conference on Applications, technologies, architectures, and protocols for computer communications, ACM, 2003, pp. 27–34. [25] F. Xia, L. Liu, J. Li, J. Ma, A. V. Vasilakos, Socially aware networking: A survey, IEEE Systems Journal 9 (3) (2015) 904–921.

560

[26] L. Pelusi, A. Passarella, M. Conti, Opportunistic networking: data forwarding in disconnected mobile ad hoc networks, IEEE Communications Magazine 44 (11) (2006) 134–141. [27] Y. Zhu, B. Xu, X. Shi, Y. Wang, A survey of social-based routing in delay tolerant networks: positive and negative social effects, IEEE Communica-

565

tions Surveys & Tutorials 15 (1) (2013) 387–401. [28] N. Vastardis, K. Yang, Mobile social networks: Architectures, social properties, and key research challenges, IEEE Communications Surveys & Tutorials 15 (3) (2013) 1355–1371.

26

[29] H. Jie, Y. Lieliang, L. Hanzo, Mobile social networking aided content 570

dissemination in heterogeneous networks, China Communications 10 (6) (2013) 1–13. [30] P. Hui, J. Crowcroft, How small labels create big improvements, in: Pervasive Computing and Communications Workshops, 2007. PerCom Workshops’ 07. Fifth Annual IEEE International Conference on, IEEE, 2007,

575

pp. 65–70. [31] P. Hui, J. Crowcroft, E. Yoneki, Bubble rap: Social-based forwarding in delay-tolerant networks, IEEE Transactions on Mobile Computing 10 (11) (2011) 1576–1589. [32] X. Gu, L. Tang, J. Han, A social-aware routing protocol based on fuzzy logic

580

in vehicular ad hoc networks, in: High Mobility Wireless Communications (HMWC), 2014 International Workshop on, IEEE, 2014, pp. 12–16. [33] F. D. Cunha, G. G. Maia, A. C. Viana, R. A. Mini, L. A. Villas, A. A. Loureiro, Socially inspired data dissemination for vehicular ad hoc networks, in: Proceedings of the 17th ACM international conference on Modeling,

585

analysis and simulation of wireless and mobile systems, ACM, 2014, pp. 81–85. [34] F. Li, J. Wu, LocalCom: a community-based epidemic forwarding scheme in disruption-tolerant networks, in: 2009 6th annual IEEE communications society conference on sensor, mesh and ad hoc communications and net-

590

works, IEEE, 2009, pp. 1–9. [35] M. Musolesi, P. Hui, C. Mascolo, J. Crowcroft, Writing on the clean slate: Implementing a socially-aware protocol in Haggle, in: World of Wireless, Mobile and Multimedia Networks, 2008. WoWMoM 2008. 2008 International Symposium on a, IEEE, 2008, pp. 1–6.

595

[36] M. Musolesi, S. Hailes, C. Mascolo, Adaptive routing for intermittently connected mobile ad hoc networks, in: Sixth IEEE International Sympo27

sium on a World of Wireless Mobile and Multimedia Networks, IEEE, 2005, pp. 183–189. [37] E. M. Daly, M. Haahr, Social network analysis for routing in disconnect600

ed delay-tolerant manets, in: Proceedings of the 8th ACM international symposium on Mobile ad hoc networking and computing, ACM, 2007, pp. 32–40. [38] E. M. Daly, M. Haahr, Social network analysis for information flow in disconnected delay-tolerant MANETs, IEEE Transactions on Mobile Com-

605

puting 8 (5) (2009) 606–621. [39] C. Boldrini, M. Conti, A. Passarella, Exploiting users social relations to forward data in opportunistic networks: The HiBOp solution, Pervasive and Mobile Computing 4 (5) (2008) 633–657. [40] G. Han, Y. Dong, H. Guo, L. Shu, D. Wu, Cross-layer optimized routing in

610

wireless sensor networks with duty cycle and energy harvesting, Wireless Communications and Mobile Computing 15 (16) (2015) 1957–1981. [41] G. Han, J. Jiang, C. Zhang, T. Q. Duong, M. Guizani, G. K. Karagiannidis, A survey on mobile anchor node assisted localization in wireless sensor networks, IEEE Communications Surveys Tutorials 18 (3) (2016) 2220–

615

2243.

28

  Azizur Rahim received the B.Sc. degree from University of Engineering and Technology,  Peshawar and MS degree in Electrical Engineering from COMSATS, Islamabad. He is  currently a PhD scholar under the supervision of Prof. Feng Xia at Mobile and Social  Computing Laboratory (MSCLab), School of Software, Dalian University of Technology,  Dalian, China. His research interests include Mobile and Social Computing, Ad‐hoc  Networks, VANETs, Mobile Social Networks, Vehicular Social Networks.    Tie Qiu received B.Sc from Inner Mongolia University of Technology, M.Sc and Ph.D from  Dalian University of Technology (DUT), China, in 2003, 2005 and 2012, respectively. He is  currently Associate Professor at School of Software, Dalian University of Technology. He  was a visiting professor at electrical and computer engineering at Iowa State University in  U.S. (Jan. 2014‐Jan. 2015). He serves as an Associate Editor of IEEE Access  Journal, Computers & Electrical Engineering (Elsevier journal) and Human‐centric  Computing and Information Sciences (Springer Journal), an Editorial Board Member of Ad  Hoc Networks (Elsevier journal) and International Journal on AdHoc Networking Systems, a  Gest Editor of Future Generation Computer Systems (Elsevier journal). He serves as  General Chair, PC Chair, Workshop Chair, Publicity Chair, Publication Chair or TPC Member  of a number of conferences. He has authored/co‐authored 8 books, over 60 scientific  papers in international journals and conference proceedings. He has contributed to the  development of 4 copyrighted software systems and invented 15 patents. He is a senior  member of China Computer Federation (CCF) and a Senior Member of IEEE and ACM.   Zhaolong Ning received the M.Sc. and Ph.D. degrees from Northeastern University, China, in  2011 and 2014, respectively. He was a Research Fellow at Kyushu University, Japan, from 2013  to 2014. He is currently an Assistant Professor at the Dalian University of Technology. His  current research interests include cloud computing, network optimisation, and social  network. He has authored over 40 papers in the above areas. He is a member of the ACM.      Jinzhong Wang received the B.S. degree in computer education from Anshan Normal  University, Anshan, China, in 2002, and the M.Sc. degree in computer application technology  from Liaoning University, Shenyang, China, in 2005. Since 2005, he has been with  Shenyang Sport University, Shenyang, China. He is currently pursuing the Ph.D. degree with the  Mobile and Social Computing Laboratory, School of Software, Dalian University of Technology,  Dalian, China. His current research interests include the analysis of big trafc data and mobility  pattern prediction by trajectory data.   

  Noor Ullah received his Bachelor’s and Master’s degrees in Electrical Engineering with major in  Communication and Electronics from University of Engineering and Technology Peshawar,  Pakistan. He is currently a PhD candidate in Software Engineering, at School of Software, Dalian  University of Technology, China. His research interests include, but not limited to Vehicular  Socials Networks and Intelligent Transportation System design.    Amr Tolba received the BSc and Ph.D. degrees from Mathematics and Computer Science  Department of Menoufia University, Egypt. He is currently Assistant Professor at Computer  Science, King Saud University, Saudi Arabia. His main research interests include Social Network  Analysis, and Internet of Things.    Feng Xia received the B.Sc. and PhD degrees from Zhejiang University, Hangzhou, China. He was  a Research Fellow at Queensland University of Technology, Australia. He is currently a Full  Professor in School of Software, Dalian University of Technology, China. He is the (Guest) Editor  of several international journals. He serves as General Chair, PC Chair, Workshop Chair, or  Publicity Chair of a number of conferences. Dr Xia has published 2 books and over 200 scientific  papers in international journals and conferences. His research interests include social  computing, computational social science, big data, and mobile social networks. He is a Senior  Member of IEEE and ACM.       

Azizur Rahiim  

       

 

Tie Qiu

               

 

Zhaolong Niing 

                 

     

Jinzhong Wang     

               

     

Noor Ullah 

             

       

Amr Tolbaa   

               

     

Feng Xia  

 

Highlights     

This paper presents a novel protocol based on social acquaintance for data forwarding in VSNs  Social feature metrics are considered for decision‐making.   Reduced End‐to‐End delay and improve packet delivery ratio  Simulations were performed to evaluate the proposed protocol under different scenarios