Community-based Message Forwarding in Mobile Social Networks

Zhiming Chen

, Yang Xiang

School of Computer, Nanjing University of Posts and Telecommunications, Nanjing 210023, China

Jiangsu Key Laboratory of Big Data Security & Intelligent Processing, Nanjing 210023, Chi

Keywords: Delay tolerant networks, Community, Forwarding, Centrality, Movement direction.

Abstract: With the popularity of various smart devices and the application of sensor network technology, message

transmission using mobile devices is becoming widespread. This paper focuses on the forwarding in mobile

social network (MSN). The MSN is a special Delay Tolerant Network (DTN) consisting of mobile nodes. In

MSN, nodes move and share information with each other through carried short-range wireless

communication devices. Mobile nodes in the MSN typically access some building areas more frequently,

such as schools, companies, or apartments, while visiting other areas, such as the roads between buildings,

less frequently. The building areas that nodes frequently visit are called communities. To increase delivery

ratio and reduce transmission time in MSN, this paper proposes a novel zero-knowledge multi-copy routing

algorithm, Mixed Message Forwarding (MMF) which exploits and improves the metric, namely centrality.

Centrality reflects the importance of a node in the network. MMF improves copy diffusion by using

different directions of node movement as well. Special facilities called boundary boxes are added to the

network scenario. Boundary boxes are special throw boxes. Throw boxes are relays with large storage space

and fixed position. MMF is designed and evaluated, which utilizes the aforementioned boundary boxes to

reduce transmission delay. The simulation results show that the MMF can improve the delivery ratio and

reduce the transmission delay, compared with other algorithms.

1 INTRODUCTION

Delay tolerant networks (DTN) are a type of

challenged networks, wherein the contacts between

the communicating devices are intermittent.

Consequently, a contemporaneous end-to-end path

between the source and destination rarely exists. In

DTNs, the node is usually highly mobile and often

moves out of the ranges of other nodes, causing only

periodic connectivity throughout the network (Hom

J et al, 2017).

Mobile social networks (MSNs) are composed of

mobile users that move around and use their carried

wireless communication devices to share

information via online social network services, such

as Facebook, Twitter, etc (Xiao M et al, 2013).

Recently, the short-distance communication model

has also been adopted by encountered mobile users

in MSNs to share information, such as multimedia,

large-size files, etc., at a low cost. Such MSNs can

be seen as a special kind of delay tolerant network

(DTN).

Message forwarding is one of the most

challenging aspects of this network because of the

inherent intermittent connectivity. In this paper, we

seek to address this particular problem by employing

the theory of node centrality and movement

direction. We propose a novel forwarding strategy,

Mixed Message Forwarding (MMF), which exploits

special facilities such as delay to decrease

transmission delay.

The rest of this article is organized as follows:

Section 2 discusses related work. Section 3 is a brief

description of MMF. The forwarding process of the

MMF algorithm is presented in Section 4. Section 5

shows the performance of MMF through a number

of simulation experiments. We make a conclusion in

Section 6.

2 RELATED WORK

Epidemic routing (Vahdate A et al, 2000), which

indiscriminately floods the network with messages,

has the highest delivery ratio and delivery time but

Chen, Z. and Xiang, Y.

Community-based Message Forwarding in Mobile Social Networks.

DOI: 10.5220/0008097401970202

In Proceedings of the International Conference on Advances in Computer Technology, Information Science and Communications (CTISC 2019), pages 197-202

ISBN: 978-989-758-357-5

197

also the highest delivery cost. In order to reduce this

cost, researchers interested in social network

dynamics have utilized various social metrics to

select the relay node. Three influential social-based

protocols are SimBet (Daly E et al, 2007), Bubble

RAP (Pan H et al, 2011), and Friendship routing

(Bulut E et al, 2014). SimBet uses similarity and

betweenness centrality metrics to determine relay

nodes with higher probabilities of delivering the

message. Similarly, Bubble RAP uses centrality and

community to make forwarding decisions, and

friendship routing considers the relationships

between nodes by introducing a metric that measures

quality of friendship between nodes.

The protocol proposed in Huijuan and Kai

(Huijuan Z et al, 2015) expands on the work in Kim

et al. (Kim CM et al, 2014) by additionally

considering endpoint-biased expanded ego

betweenness centrality. Contact Frequency Based

Approach (CFBA) and Contact Duration Based

Approach (CDBA) are two very similar routing

protocols proposed in a single paper (Gondaliya N et

al, 2016). They both separate the nodes into

communities using the k-clique method based on

contact duration, and use centrality, a metric that

represents the connectivity of the network, to select

the relay node. Social-Based Single Copy Routing,

or SBSCR, (Gondaliya N et al, 2016) is a

community-based routing mechanism in which

routing decisions are made based on a calculated

social based utility (SBU) that considers similarity

and friendship values. The two protocols proposed

in Chen and Lou (Chen H et al, 2015), Expected

Encounter Routing (EER) and Community Aware

Routing (CAR), are based on metrics determined

through history of interaction between nodes. IRS

(Singh AK et al, 2018) is an incentive based routing

strategy. In this approach intermediate nodes can

participate and earn incentives for sacrificing their

selfishness. Choksatid et al. propose the protocol

SED (Choksatid T et al, 2016), which is the

renovation of Epidemic Routing scheme. (Igarashi Y

et al, 2018) proposed by Igarashi et al. controls

message forwarding in each terminal using

parameters named “Community” and “Centrality.”

3 MMF ALGORITHM

OVERVIEW

In order to solve these problems, this paper proposes

a community-based opportunity network algorithm

MMF algorithm. The MMF algorithm is divided into

four stages: internal forwarding, external forwarding,

roaming and acquisition. To more intuitively see the

entire delivery process, Fig. 1 is presented a static

scene. Certainly, the nodes in our simulated scene

are mobile.

Figure 1: Network model.

(1) In the internal forwarding stage, nodes

transfer the message in the community to which the

source belongs until a node encounters a boundary

box. Nodes are more inclined to send the message to

boundary boxes.

(2) In the external forwarding stage, when a

boundary box receives the copy of the messages, it

will spread the messages to the neighboring

communities’ nodes within its transmission area.

(3) In the roaming stage, the node forwards the

messages priority to boundary boxes. We use a

multi-copy approach based on node movement

direction to spread information to other communities.

(4) In the acquisition stage, the destination will

extract message from any first-time carrier. The

device carrying the message can be a node or a

boundary box.

In the above scheme, the four phases do not

necessarily follow the order, which is determined by

the location of the source and the destination. The

source and the destination are in one community, so

it is possible to go directly to the fourth phase. Or, if

the destination is near the source node’s community,

then the third phase will not be executed.

CTISC 2019 - International Conference on Advances in Computer Technology, Information Science and Communications

198

4 MMF ALGORITHM

In this section, we presents the MMF algorithm in

detail. This article focuses on only message transfer,

and it can send messages as long as each carrier has

enough cache space and the link has enough

bandwidth. In addition, the communication time

between any two nodes and between the node and

the boundary box is independent.

4.1 Internal Forwarding Stage

The central utility value can be used to measure the

importance of a node in message transmission. The

central utility value defined in this paper consists of

two parameters, the number of node forwardings and

the number of neighbor node changes. The central

utility value is calculated as follows:







is the number of node  forwardings in a

certain time slot 





reflects the number of the node

 acts as a relay in a certain time slot. The higher 





the higher the degree to which a node is available in

that time slot.

The standard calculation method for the 





is as

follows:

 











(1)

where  













is the average of the 





from the

beginning to the time slot  , and  



 





 







is the dispersion.







is the number of neighbor node changes for

node  at a certain time slot 





reflects the number

of the node  encounters other nodes and the number

of changes in neighbor nodes in the certain time slot

. The higher 





means that the node meets more

new nodes in the time slot . In a time slot 





defined as:







 









 











  









 













(2)

where 









is the neighbor nodes of the current time

slot , and 











is the neighbor nodes of the

previous time slot 



The standardized calculation method of the 





as follows:

 











(3)

where  













is the average of the 





from the

beginning to the time slot t,  



 





 







the dispersion.

The node have highter  and  values and more

likely to encounter the destination node. So we

combine these two parameters by the following

formula to define the central utility value 



of node

:





  ,

(4)

where  are the set weights.

When the node receives the central utility 



the neighbor nodes, this node compares its value to

the maximum one. If this node’s value is lower, then

it will forward a message to the node with the largest

utility value.

4.2 Spread Stage

In the spread phase, when a boundary box receives a

copy of the message, the boundary box spreads

message to nodes with no message outside of the

community to which this copy sender belongs.

4.3 Roaming Stage

In the roaming stage, in order to speed up

transmission, we use a multi-copy method. The node

 carries the number of the belonging community,

the current position coordinates 







 and the

previous time slot position coordinates, such as









. These two position coordinates can be

used to calculate the current slot movement direction

of the node .

Suppose that in the scene, there is a node  with

copy, and there is a node  without copy in the

transmission range of . If the cosine of the current

slot movement direction of the two nodes satisfies

within a certain threshold range,  sends a copy to .

The range of the cosine value indicates that the arc

of the angle  between the two moving directions of

 and  is around





. That is to say, in the

transmission range of the node  having message,

there is a node  having no message, and if the angle

of movement direction of a and b is in the arc of







󰒮 , 󰒮 



 







, the node  transmits a copy to

. The cosine value is calculated as follows:

































































































































 (5)

where  .

Community-based Message Forwarding in Mobile Social Networks

199

4.4 Acquisition Stage

In the extraction phase, the destination gets the

message when it encounters the message carrier.

This message carrier may be in the internal

forwarding stage, external forwarding stage or

roaming stage. In the worst case, the message is

spread to each community, then the destination just

gets the message.

5 SIMULATION

In this section, we conducted a number of

simulations to evaluate the performance of MMF.

The comparison of the algorithms, evaluation

methods, parameter settings and results are shown

below.

In this paper, we only focus on zero-knowledge

multi-copy routing algorithm in MSN. In order to

obtain a fair performance comparison, we only

compare our algorithm with the existing zero-

knowledge multi-copy routing algorithm: Epidemic

algorithm, Prophet algorithm and HS algorithm.

Epidemic, Prophet and HS algorithms all transfer

messages through replication. The Epidemic

algorithm can achieve the best delivery delay in all

routing algorithms. The Prophet algorithm is a

utility-based multi-probability routing algorithm. In

the HS algorithm, the message carrier uses a binary

method to send a message to the encountered node

or throwbox.

5.1 Parameter Settings

The scenes simulated in this paper are relatively

large, the nodes are assumed to move randomly, and

the initial positions of the nodes are also randomly

generated. The boundary box is generated at

initialization time and has a fixed location. In

addition, we can alter the parameter value as needed,

so that we can observe the impact of each parameter

value on the result and bring out the optimal results,

which is beneficial to compare with other algorithms

and evaluate the pros and cons of the algorithm.

In the simulation of this paper, the scene of the

model is a large rectangle. To simplify the

simulation, we set the length and width of the scene

to be the same. The node's transmission radius is set

to 15, and the number of communities is fixed at a

value of 9. The experiment is roughly divided into

three parts in terms of the average delivery rate, the

average transmission time and the influence of the

angle cosine value on the average transmission time.

In the first experiment, the community length and

width  of the four algorithms were adjusted to 50,

60 and 70, and then we compare the experimental

results. In the second experiment, the nodes’ number

 of the four algorithms was adjusted to 1000, 2000,

and 3000, and then we compare the experimental

results. The evaluation parameters are shown in

Table 1.

Table 1: Parameter Settings.

Parameter Name

Range

Experimental Area

( )

150*150/180*180/210*210

Number of Nodes

()

1000/2000/3000

Node Transmission

Radius

Number of

Communities

()

Community Length

And Width ()

50/60/70

Duration ()

240/600/6000

Angle cosine ()

0. 1/0. 5/0. 8/1

The metrics evaluated in this simulation are the

average delivery rate and the average transmission

time. The average delivery rate is the ratio of the

number of successful deliveries to the total number

of messages. The average transmission time is the

delivery time of the first copy to its destination.

5.2 Simulation Results

We conduct three sets of simulation experiments to

evaluate the impact of various parameters on

performance. In the first set of simulations, we

adjusted the community length and width  of the

four algorithms to 50, 60, and 70, and set N=2000,

θ=0. 5. In the second set of simulations, we adjust

the number of nodes N of the four algorithms to

1000, 2000, and 3000, and set D=50, θ=0. 5. In

order to observe the difference between the

algorithms more specifically, in these two sets of

experiments, we will observe the trend of each

algorithm under the condition of t=240, t=600,

t=6000, as shown in Figure 2 and Figure 3.

CTISC 2019 - International Conference on Advances in Computer Technology, Information Science and Communications

200

(a) t=240.

(b) t=600.

Figure 2: Comparison of average delivery ratio.

Fig. 2 shows that the average delivery rate of the

four algorithms decreases as the community length

and width increase. In contrast, the HS algorithm has

the worst average delivery rate at D=70. Prophet's

probabilistic selection causes the message to spread

continuously close to the destination, but since

Prophet only relies on nodes to forward messages, it

has moderately lower performance. MMF is mainly

due to the external forwarding function of boundary

boxes, which enables the message to spread rapidly

to other communities, so it performs better than the

two algorithms above. Epidemic's message number

and number of forwarding nodes are not limited, so

the diffusion speed is very fast in a small scenario,

but as the entire scene becomes larger, spreading

messages requires more transmission time.

(a) t=240.

(b) t=600.

Figure 3: Comparison of average transmission time.

Next, we also perform three sets of simulations

to evaluate the performance of the above algorithm

in terms of average transmission time. The results in

Fig. 3 show that as the number of nodes increases,

the average transmission time all decreases. The

results show that at t=240 and t=600, the Prophet

has the longest transmission time at n=1000. When

t=6000, the transmission time of HS is more than

the other three algorithms. Overall, the longer t is,

the more unstable HS is. Because the role of the

boundary box in the HS is only to spread the

messages in the community, the spread of the

messages between the communities only depends on

the single copy of the node, so the number of nodes

has a great impact on it and the transmission time is

long. MMF uses the external forwarding function of

boundary boxes and relatively few nodes, so that the

messages can quickly spread to the other

communities, therefore it has a shorter average

transmission time among the three algorithms except

Epidemic.

Community-based Message Forwarding in Mobile Social Networks

201

Figure 4: Effect of MMF angle cosine on average

transmission time.

Next, we modify the angle cosine value K used

by MMF in the roaming phase for four times. We set

N=2000, D=50, t=240. As shown in Fig. 4, we

compared the average transmission time at K=0.1,

K=0.5, K=0.8, and K=1. The result shows that the

closer K is to 1, the smaller the average transmission

time is, the more similar it is to flooding, but this

will cause massive copies, unnecessary energy

consumption and useless bandwidth occupation.

6 CONCLUSIONS

In this paper, we study a special mobile social

network, in which the running scenario includes

some nodes, communities and boundary boxes, and

propose a zero-knowledge multi-copy routing

algorithm called MMF. MMF set a higher priority

for boundary boxes to help spread information

quickly. Theoretical analysis and simulation results

show that boundary boxes play an important role in

the message dissemination process. By using

boundary boxes, MMF achieves better performance

than several existing zero-knowledge MSN routing

algorithms.

ACKNOWLEDGMENTS

This research is supported by National Natural

Science Foundation of China under Grant Nos.

61872191, 41571389, 61872193.

REFERENCES

Hom J, Good L, Yang S. A survey of social-based routing

protocols in Delay Tolerant Networks. International

Conference on Computing. Santa Clara: IEEE, 2017:

788-792.

Xiao M, Wu J, Huang L. Homing spread: Community-

home-based Multi-copy Routing in Mobile Social

Networks. IEEE Infocom.Turin: IEEE, 2013: 2319-

2327.

Vahdate A, Becker D. Epidemic routing for partially

connected ad hoc networks. Technique Report, CS-

2000-06, Department of Computer Science, Duke

University, Durham, NC, 2000.

Daly E, Haahr M. Social Network Analysis for Routing in

Disconnected Delay-Tolerant MANETs. Acm

Interational Symposium on Mobile Ad Hoc, 2007.

Pan H, Crowcroft J, Yoneki E. Bubble rap: social-based

forwarding in delay tolerant networks. IEEE

Transactions on Mobile Computing, 2011, 10(11):

1576-1589.

Bulut E, Szymanski B. Exploiting Friendship Relations for

Efficient Routing in Mobile Social Networks. IEEE

Transactions on Parallel and Distributed Systems,

2014, 23(12), 2254-2265.

Huijuan Z, Kai L. A Routing Mechanism Based on Social

Networks and Betweenness Centrality in Delay

Tolerant Networks. International Journal of Computer

Science & Information Technology, 2015, 7(6), 107-

116.

Kim CM, Han YH, Youn JS, Jeong YS. A Socially Aware

Routing Based on Local Contact Information in

Delay-Tolerant Networks. The Scientific World

Journal, 2014.

Gondaliya N, Kathiriya D. Community detection using

inter contact time and social characteristics based

single copy routing in delay tolerant networks.

International Journal of Ad hoc, Sensor & Ubiquitous

Computing, 2016, 7(1), 21-35.

Gondaliya N, Kathiriya D. Community detection using

inter contact time and social characteristics based

single copy routing in delay tolerant networks.

International Journal of Ad hoc, Sensor & Ubiquitous

Computing, 2016, 7(1), 21-35.

Chen H, Lou W. Contact expectation based routing for

delay tolerant networks. Ad Hoc Networks, 2015,

36(1), 244-257.

Singh AK, Pamula R. IRS: Incentive Based Routing

Strategy for Socially Aware Delay Tolerant Networks.

2018 5th International Conference on Signal

Processing and Integrated Networks, 2018, 343-347.

Choksatid T, Narongkhachavana W, Sumet Prabhavat. An

efficient spreading Epidemic Routing for Delay-

Tolerant Network. 2016 13th IEEE Annual Consumer

Communications & Networking Conference, 2016,

473 – 476.

Igarashi Y, Miyazaki T. A DTN routing algorithm

adopting the “Community” and “Centrality”

parameters used in social networks. 2018 International

Conference on Information Networking, 2018, 211 –

216

CTISC 2019 - International Conference on Advances in Computer Technology, Information Science and Communications

202