Directional-IPeR: Enhanced Direction and Interest Aware

PeopleRank for Opportunistic Mobile Social Networks

Yosra S. Shahin, Soumaia A. Al Ayyat and Sherif G. Aly

Computer Science and Engineering Department, The American University in Cairo, Cairo, Egypt

Keywords: Mobile Networks, Opportunistic Networks, Delay Tolerant Networks, Social Networks, PeopleRank,

Interest-Aware Forwarding, IPeR, MIPeR, Socialcast.

Abstract: Network infrastructures are being continuously challenged by virtue of increased demand, resource-hungry

applications, and at times of crisis when people need to work from homes such as the current Covid-19

epidemic situation, where most of the countries applied partial or complete lockdown and most of the people

worked from home. Opportunistic Mobile Social Networks (OMSN) prove to be a great candidate to support

existing network infrastructures. However, OMSNs have copious challenges comprising frequent

disconnections and long delays. In this research, we aim to enhance the performance of OMSNs including

delivery ratio and delay. We build upon an interest-aware social forwarding algorithm, namely Interest Aware

PeopleRank (IPeR) in two ways 1) By embracing directional forwarding (Directional-IPeR), and (2) By

utilizing a combination of Directional forwarding and multi-hop forwarding (DMIPeR). Different interest

distributions and users’ densities are simulated using the Social-Aware Opportunistic Forwarding Simulator

(SAROS). The results show that Directional-IPeR with a tolerance factor of 75% performed the best in terms

of delay and delivery ratio compared to IPeR, and two other algorithms, namely MIPeR and DMIPeR.

1 INTRODUCTION

Network infrastructures are experiencing notable

challenges (Vahdat & Becker, 2000), especially with

increased demand, and at times of worldwide crises

when people work from home, and infrastructures

are stretched to their limits. Opportunistic Mobile

Social Networks (OMSNs), can provide excellent

complimentary support to existing network

infrastructures. OMSN is the combination of

Opportunistic Network (ON) and MSN. ON is

Mobile Ad hoc Network (MANET) with frequent

disconnections where there is no information about

the network connection or the nodes’ mobility

patterns. To deliver the message to its destination,

ON uses some nodes as intermediate carriers to host

the message and forward it to other nodes until it

reaches the destination. In ONs, the node forward or

store-and-carry the message occurs at or takes place

at every hop. Consequently, the delay between being

out of range and back must be considered.

Consequently, ON is also called Delay Tolerant

Network or Disruption Tolerant Networks (DTN).

Mobile Social Network (MSN) constitutes a Social

Network (SN) that is based on the interaction of a

group of people using their mobile devices. The

interactions in MSNs establish relationships or links

which can be physical contact, a shared interest, age,

language, place, or any other relationships. It

employs social advantages as well as the capabilities

of smartphones like GPS, sensing features, and

communication links. Thus, OMSNs are Ad hoc

networks in which the nodes are in motion with

recurring disconnections using mobile devices.

(Rajpoot & Rajendra, 2015) (Sobin, Raychoudhury,

Marfia, & Singla, 2016) (Pal, Saha, & S.Misra, 2017)

(Hom, Good, & Yang, 2017) (Zhu, Xu, Shi, & Wang,

2013) (Vasilakos, Spyropoulos, & Zhang, 2016) (Wu

& Wang, 2014) (Liu & Jing, 2012) (Sui, 2015).

There are many challenges however that face

OMSNs including long delays and frequent

disconnections. Consequently, the certainty of

delivering a message to its destination is

compromised (Li, Joshi, & Finin, 2010) (Moati,

Otrok, Mourad, & Robert, 2013).

In this research, we contribute to enhancing the

performance of some of the well-established

algorithms used for OMSNs including, but

notlimited to, enhancing delivery ratio and reducing

delay. Our work builds upon Interest Aware

Shahin, Y., Al Ayyat, S. and Aly, S.

Directional-IPeR: Enhanced Direction and Interest Aware Peoplerank for Opportunistic Mobile Social Networks.

DOI: 10.5220/0010105900190029

In Proceedings of the 10th International Conference on Pervasive and Parallel Computing, Communication and Sensors (PECCS 2020), pages 19-29

ISBN: 978-989-758-477-0

PeopleRank (IPeR), which was developed an interest

and social aware algorithm that outperformed

comparable algorithms (Al Ayyat, Harras, & Aly,

2013) and its multi-hop variant, Multiple Hops

Interest Aware PeopleRank (MIPeR) (Shahin, Al

Ayyat, & Aly, 2020). We worked on two major

fronts (1) embracing direction as a guiding criterion

in ranking nodes in support for content forwarding

decision making based on IPeR, Direction and

Interest Aware PeopleRank (Directional-IPeR), and

(2) utilizing the combination of MIPeR and

Directional-IPeR in ranking nodes in support for

content forwarding decision making Directional

Multiple Hops Interest Aware PeopleRank

(DMIPeR). For Directional-IPeR, we utilize

different values of what we call the tolerance factor

to experiment with different ways of selecting

forwarder nodes while keeping direction into

consideration. The tolerance factor is a percentage,

namely, 25%, 50%, and 75%. We multiply the IPeR

value by one of these tolerance factors to come up

with a new value which is less than IPeR value to

constitute a threshold below which we cannot send

the data to nodes whose IPeR value is less than this

threshold. For instance, the algorithm Directional-

IPeR-75 selects the next forwarders with a tolerance

factor of 75% of the IPeR value of the current

message holder. For each experiment, different

interest distributions as well as different user

densities are employed. Based on the results of the

simulation runs, adding direction guidance to IPeR

with a 75% tolerance factor performs the best in

terms of delay and delivery ratio compared to IPeR,

MIPeR, and DMIPeR.

Our contribution consists of (1) the addition of

direction awareness to the IPeR algorithm with some

preset tolerance factor improved both the delivery

ratio and the number of reached interested

forwarders. (2) Including 2 and 3 hops to

Directional-IPeR-75 do not gain any improvement.

(3) In high- density areas, Directional-IPeR performs

better in all metrics compared to IPeR except in

terms of delay. However, in less crowded

environments, it reduces delays. Therefore, it can be

employed in rural or disastrous areas, in which few

people have internet access with low connectivity

and spread over a big area. Furthermore, Directional-

IPeR-75 outperforms the SocialCast algorithm in all

metrics except for cost. For instance, Directional-

IPeR-75 reduced delay to 200% of that incurred by

SocialCast.

The remainder of this paper is organized as

follows. We discuss the related work in Section 2.

Section 3 illustrates the concept of integrating

direction awareness with interest-aware PeopleRank

(Directional-IPeR), Section 4 presents simulation

settings. Section 5 elucidates the evaluation results of

the new algorithms, followed by a conclusion in

Section 6.

2 RELATED WORK

Many contributions were made to mitigate some of

the challenges associated with OMSNs. Beyond

Epidemic routing (Vahdat & Becker, 2000), other

protocols use contact history (Jain, Chawla, Soares,

& Rodrigues, 2016) (Spyropoulos, Psounis, &

Raghavendra, 2005) (Abdelkader, Naik, Nayak,

Goel, & Srivastava, 2016) (Spaho, Bylykbashi,

Barolli, Kolici, & Lala, 2016). The Probabilistic

Routing Protocol using the History of Encounters

and Transitivity (PRoPHET) protocol (Pathak,

Gondaliya, & Raja, 2017) (Denko, 2016) (Vasilakos,

Spyropoulos, & Zhang, 2016), for instance, is

grounded on using a set of probabilities, which are

established on the history of past contact, to outline

the successful delivery of the message to its

destination. The Spray and Wait Protocol is

considered the most appropriate store-carry-forward

routing protocol. The goal of Spray and Wait is to

reduce transmissions by reducing the total number of

copies per message (Spyropoulos, Psounis, &

Raghavendra, 2005) (Jain, Chawla, Soares, &

Rodrigues, 2016). MaxProp gives more priority to

the packets with the minimum number of hops by

storing a vector that represents the likelihood to meet

other nodes in the network (Spaho, Bylykbashi,

Barolli, Kolici, & Lala, 2016).

Other protocols use centrality to represent the

importance of nodes within the social network.

Consequently, central nodes are better candidates to

send messages to other nodes in the network (Zhu,

Xu, Shi, & Wang, 2013) (Daly & Haahr, 2007).

SimBet employs betweenness centrality using 1-hop

and 2-hop neighbors and the local social similarity to

choose the intermediary nodes for efficient message

routing (Daly & Haahr, 2007).

Many protocols rely on constituted communities

that can accelerate message delivery. In social

networks, the members of one community tend to

meet with a higher probability compared to other

members who are not in the same community.

(Cherif, Khan, Filali, Sharafeddine, & Dawy, 2017)

(Palla, Derényi, Farkas, & Vicsek) (Hui, Crowcroft,

& Yoneki, 2011) (Meng, et al., 2019) (Chang &

Chen, 2014). Bubble RAP is a social network

protocol, which is based on community and

PECCS 2020 - 10th International Conference on Pervasive and Parallel Computing, Communication and Sensors

centrality metrics (Hui, Crowcroft, & Yoneki, 2011).

First, a contact graph is employed to represent

mobility traces. It relies on the number of contacts

and the contact duration, where physical nodes are

represented by nodes in the graph, the edges

represent the contacts, and the weights on the edges

represent the contact duration and frequency. To

detect the communities of nodes in a social network,

K-CLIQUE and Weighted Network Analysis (WNA)

are applied. Other protocols rely on friendship to

constituent communities. Friendship in OMSN is

based on having a regular and longtime duration of

contacts or have a shared interest (Wu & Chen, 2016)

(Chen & Wu, 2016) (Perrig, Stankovic, & Wagner,

2004) (Dubois-Ferriere, Grossglauser, & Vetterli,

2003) (Bulut & Szymanski, 2012). For instance,

Friendship-Based Routing Protocol (FBR) is based

on community and friendship metrics. It can detect

the direct and the indirect friendship between nodes

based on three features, namely, regularity,

frequency, and longevity of the contacts. Regularity

indicates the variance of the inter-meeting time.

Frequency indicates the average inter-meeting time.

Longevity is the average duration time of the meeting

sessions (Bulut & Szymanski, 2012).

Other protocols rely on location guidance

(Barkhuus & Dey, 2003) (LeBrun, Chuah, Ghosa, &

Zhang, 2005) (Kim, Choi, & Yang, 2015). Hotspots

or Stop points are locations, in which people tend to

gather. Forwarding a message in hotspots can

guarantee it reaches a big number of nodes. Some

other protocols rely on the direction (Dhurandher,

Borah, Woungang, Bansal, & Gupta, 2018) (Jeon,

Kim, Yoon, Lee, & Yang, 2014). People are moving

around in different directions, speeds, and visit

different places. Based on this fact, if a message is

forwarded to nodes, which are traveling to different

places where the source node cannot go, then the

message has a good chance to reach its destination.

An example is Direction Entropy Based Forwarding

Scheme (DEFS), which utilizes the main direction

and the direction entropy to predict the nodes’

direction and to identify the nodes that have high

mobility and consequently high probability of

meeting the destination of a message (Jeon, Kim,

Yoon, Lee, & Yang, 2014).

Using social aspects such as the user interest,

gender, age, or language is used to define a social

vector that defines a rank, which is employed to

forward the messages in OMSNs such as

PeopleRank (Mtibaa, May, & Ammar, 2012), and

Interest Aware PeopleRank (IPeR) (Al Ayyat,

Harras, & Aly, 2013). SocialCast is a publish-

subscribe routing framework that utilizes metrics of

social interaction such as, patterns of movements to

predict the best nodes to forward a message. It used

a mobility model based on a social network (Costa,

Mascolo, Musolesi, & Picco, 2008).

One of the protocols that are based on contacts is

the Two Hops Prediction Protocol. To consider two

hop communication, initially, the source node must

have the information about the probability of contact

of each of its neighbors with the destination. When

two nodes meet, they exchange their unordered list

of node IDs which have a contact probability above

a certain threshold. Then, each node checks whether

it has any message that is destined to any of the

neighbors of the nodes in the list of the encountered

nodes. If a message is found, a copy of the message

is sent to the encountered node. This is repeated until

the message reaches the destination. This protocol

performed efficiently compared to Epidemic,

Random, and PRoPHET Protocols (Song & Kotz,

2016)

In this paper, we explored the combination of

direction, interest, and social awareness (Directional-

IPeR) with one hop and multiple hops prediction.

Directional-IPeR with one hop improved the delivery

ratio and the number of reached interested

forwarders. Employing multiple hops to Directional-

IPeR does not gain any enhancement and instead

increased the cost. In crowded areas, Directional-

IPeR performs better in all metrics compared to IPeR

except for delay. However, it reduces delays in less

crowded areas. Consequently, it fits in all

environments.

3 INTEGRATING DIRECTION

AWARENESS WITH

INTEREST-AWARE PeopleRank

In this section, we introduce direction awareness to

interest-aware social-based forwarding algorithms

that recognize destination nodes by their interest

proﬁle. First, we use IPeR and its variation MIPeR as

interest-aware social-based forwarding algorithms in

OMSNs. We then introduce the Directional-IPeR

algorithm which integrates direction awareness in

IPeR.

IPeR is an interest-aware social forwarding

algorithm (Al Ayyat, Harras, & Aly, 2013) that

introduces interest awareness in ranking the mobile

nodes besides the typical social ranking and

activeness used in the social-based ranking

PeopleRank algorithm (PeR) (Mtibaa, May, &

Ammar, 2012). Each node carries a PeopleRank

Directional-IPeR: Enhanced Direction and Interest Aware Peoplerank for Opportunistic Mobile Social Networks

value, which ranks nodes as per their social

popularity, the node has its interest vector in terms of

the topivcs of interest not concerning the exchanged

messages. When a message is going to be

exchanged/forwarded to a node, the interest vector of

the message is sent to the nearby nodes. Each node

compares its interest vector with that of the message

to come up with a Jaccard similarity (Bank & Cole,

2008) index value. To elaborate, when a group of

nodes has a higher or equal interest and PeopleRank

values than the current node, each member of this

group receive a copy of the message. On the other

hand, to make nodes other than the source and

destination nodes participate in delivering a

message; a copy of the message is directed to the

nodes, which have an interest in its content. It is a

sort of an incentive for these nodes to sacrifice part

of their storage and power when participating in

delivering a message to other nodes. IPeR reduces

the delivery cost, and delay in comparison to that of

Epidemic and PeopleRank algorithms (Al Ayyat,

Harras, & Aly, 2013)

Routing tables can be evolved to enhance the

forwarding mechanism in OMSNs to reduce network

flooding (Takasuka, Hirai, & Takami, 2018).

Intending to explore the effect of considering the

number of contacts with the encountered nodes,

MIPeR uses the IPeR value of the nodes and

accumulates them by using such routing tables. Thus,

the routing information includes the node’s

PeopleRank and degree of interest in the message

content (IPeR). However, these values are updated

based on the number of contacts with the

encountered nodes, to compute the 2 or 3 hop routing

(MIPeR) values. As per the simulation results

published in earlier research (Shahin, Al Ayyat, &

Aly, 2020), the 2Har-MIPeR algorithm performs

better in terms of the number of reached interested

forwarders and delay compared to IPeR. It even

performed better than its 3 hop version. The denser

the environment is, the more delivery ratio, the more

reached interested forwarders, the less cost and less

delay exerted by the algorithm.

3.1 Directional-IPeR

Directional-IPeR introduces direction awareness into

the forwarding decision of the IPeR algorithm. The

aim is to increase the chance that a copy of the

message is sent to all four directions with certain

ratios to increase the probability of reaching the

destinations. With the inspiration of Direction

Entropy Based Forwarding Scheme (DEFS) (Jeon,

Kim, Yoon, Lee, & Yang, 2014) each node has its

transmission range divided into 4 quarters namely

R1, R2, R3, R4 with an angle of 90 degrees as

illustrated in Figure 1. Each node stores its locations

at the latest two-time slots. Then, it compares them

to find the difference, which will map its direction of

motion to one of the 5 states (S0, S1, …S4). Note that

state S0 indicates that the node did not move.

In IPeR, when a group of nodes has IPeR values

higher than or equal to that of the message holder

node, they receive a copy of the message. They are

illustrated in Figure 1 as rectangles. The Directional-

IPeR algorithm examines the four main directions of

the nodes in this group. If MH finds a direction to

which no nodes are heading, it sends a copy of the

message to other nodes that are heading in this

direction but has an IPeR value greater than the

tolerance factor. Such a case is illustrated in Figure 1

as the circle in R3. It sacrifices a percentage of The

IPeR value threshold for selecting the forwarder

nodes to be sure that all of the four main directions

are covered. The goal is to increase the delivery ratio

but with the consideration of the cost. The

Directional-IPeR algorithm is illustrated in

Algorithm 1.

Figure 1: Forwarder node selection in the Directional-IPeR

algorithm.

To simulate real-life scenarios, we should expect

that some uninterested nodes may refuse to

participate in the message delivery process as they

have no interest in This message content. We tried

this attitude in Directional-IPeR. Consequently, no

message delivery that is taking place from those

nodes.

PECCS 2020 - 10th International Conference on Pervasive and Parallel Computing, Communication and Sensors

Algorithm 1: Directional-IPeR.

Function Directional-IPeR (runs on message holder

node)

Input:

IPeRList



all the nodes in contact with the message

holder node and their IPeR value is >= to the message

holder node’s IPeR value

ContactList



all the nodes in contact with the current

node and their IPeR value >= X% of message holder’s

IPeR value, and their IPeR value < the message holder

node’s IPeR value

Output:

Directional_IPeRList

Declare lists s1, s2, s3, s4, Directional_IPeRList as lists of

type node

Directional_IPeRList  IPeRList

For each node N in Directional_IPeRList do

nodeState  requestState(N)

map N to the corresponding state list

For each empty state SX

While SX is empty

For each node CN in ContactList do

nodeState  requestState(CN)

if nodeState == SX

map CN to SX

add CN to Directional_IPeRList

3.2 DMIPeR: Hybrid Directional

Multi Hops IPeR

Based on the best results of Directional-IPeR, MIPeR

is merged with the best variation of Directional-IPeR.

Therefore, two extra pieces of information are added

to each node, namely, its contacts and its direction of

motion. When two nodes meet, then specify that its

value will be the MIPeR value, not the IPeR value.

Then, a group of nodes is formulated. Each node in

this group has their MIPeR value greater than the

IPeR value of the message holder node. Based on that,

if any direction is not covered by this group, other

nodes which are expected to head to this direction get

a copy of the message based on a predefined IPeR

tolerance factor. DMIPeR algorithm is illustrated in

Algorithm 2.

Algorithm 2: DMIPeR.

Function DMIeR (runs on message holder node)

Input

IPeRList



all the nodes in contact with the message

holder node and their IPeR value is >= to the message

holder node’s IPeR value

ContactList



all the nodes in contact with the current

node and their IPeR >= X% of message holder IPeR

value, and their IPeR value< the message holder node’s

IPeR value

Output:

DMIPeRList

2HarIPeRList  2HopHarmonicMean(IPeRList)

(Shahin, Al Ayyat, & Aly, 2020)

DMIPeRList  Directional-IPeR(2HarIPeRList,

ContactList)

4 SIMULATION

In this section, we evaluate our proposed algorithms

via simulation and validate our results using Self-

similar Least Action Walk (SLAW) mobility models

(Lee, Hong, K, Rhee, & Chong, 2012). We brieﬂy

describe our setup, and present a subset of our results.

4.1 Simulation Setup and Parameters

We used the SAROS simulator (Al Ayyat, Aly, &

Harras, 2016) because it provides a wide variety of

opportunistic forwarding algorithms and their related

evaluation metrics. Besides, it correlates a diversity

of interest distributions and social network

integration associated with imported real traces.

Besides, it generates random social proﬁles including

interest for each user. To gain authentic results we

used SAROS as it is the same simulator used to

evaluate IPeR, which is the algorithm we are

enhancing.

In our experiments, SAROS was adjusted to work

over an area of 1000m x 1000m on extracted user

traces from (SLAW) mobility model (Lee, Hong, K,

Rhee, & Chong, 2012). SAROS incorporates social

contexts and interests among people in small scale

communities such as malls. Furthermore, the

constructed friendship graph includes up to 20% of

the available users in the friend list per user. To get

authentic results, each experiment is run 20 times and

the average is calculated. Each run delivers 2

messages in an hour. Every 20 runs are applied with

different user densities and different interest

distribution, namely; discrete uniform, normal and,

two disjoint subgraphs. In the discrete uniform

interest distribution, the users are spread equally

between 11 categories with varying interest rates

ranging from 0 to 1. Accordingly, the destination set

establishes 18% of the nodes while the interested

forwarders cover 36%. In the normal interest

distribution, the destination set embraces 2% of the

community, the interested forwarders set comprises

48%, while the remaining 50% are uninterested

nodes. In the two disjoint subgraphs distribution of

interest, which is a challenging environment, the

destination set embraces 2% of the community and

Directional-IPeR: Enhanced Direction and Interest Aware Peoplerank for Opportunistic Mobile Social Networks

the remaining 98% are uninterested nodes. The most

important simulation environment parameters are

listed in Table 1.

Table 1: Simulation Environment Parameters.

Parameter Value

No. of users 50, 100, and 200

No. of messa

es: 2

Set of Interests 10

Similarity interest

distribution

Discrete uniform,

discrete normal, and two

disjoint subgraphs

Initial Batter

Distribution Full Batter

Distribution

max user move 1.42m / 1 sec.

Simulation Duration 1 hou

Tolerance Factor

of Directional-IPeR

Zero, 25%, 50% and 75%

4.2 Simulation Metrics

Since our goal is to evaluate the efficiency and

effectiveness with which we can opportunistically

reach users in OMSNs. We particularly use the

following metrics: delivery ratio, the ratio of

contacted interested forwarders, delay, F-

measure, and delivery cost. The delivery ratio is the

number of reached destination nodes to the total

number of destination nodes that should receive the

message. Delay is the average time consumed for a

message to reach the destination node since it was

sent from the source node. It is presented in the

figures in a normalized form where the delay in

minutes is divided by the simulation time (60

minutes). Delivery cost is the number of copies of the

message. it is presented in the figures in a normalized

form where the cost is divided by the max. the number

of message replica that can be generated among the X

number of users (e.g. 200 message replica by 200

users). F-measure is the harmonic mean of precision

and recall. It is utilized to implement a type of penalty

for reaching uninterested forwarders. Note that the

targeted true set consists of the interested forwarder

nodes in addition to the destination nodes, while the

false set contains the uninterested nodes.

Each experiment is implemented using 3 different

densities, which are 50, 100, and 200 users. Each user

density is implemented with three different

distributions of interests, which are uniform, normal,

and two disjoint subgraphs.

5 RESULTS

In this section, we present the simulation results.

First, we present Directional-IPeR. Then, the special

case of Directional-IPeR, which is Directional-IPeR

with random discard is demonstrated. It happens

when some uninterested forwarders decide to discard

the message and not to forward it to other nodes.

Finally, Hybrid Directional Multi Hops IPeR

(DMIPeR) is presented. It utilized 2 and 3 hops to

Directional-IPeR.

5.1 Directional-IPeR

For all distributions of interest and users’ densities,

Directional-IPeR-75, which is based on including

nodes with IPeR value not less than 75% of the

message holder’s IPeR value, is the best algorithm

proposed for direction guidance in terms of F-

measure and cost. For uniform distribution, it

performs better than IPeR in terms of delivery ratio

(up to 1% for 200-user experiments), reached

interested forwarder nodes (up to 19% for 50-user

experiments), and F-measure (up to 1% for 50 and

100 user experiments). Also, it reduces delays (up to

11% for 50 users). For the different densities of

users, the denser the environment is the more

delivery ratio, the more reached interested

forwarders, and the less delay exerted by the

algorithm as illustrated in Figure 2a. For example, if

we have two environments. one environment

encompasses 100 users while the other environment

has 1000 users. Within the latter environment, more

destination nodes, more interested forwarders, and

fewer delays are achieved.

For the normal distribution, it performs better than

IPeR in terms of reached interested forwarder nodes

(up to 13% for 50 users), and delay (up to 27% for 50

users). For F-measure, it performs equally or better

compared to IPeR. For the different user densities, the

denser the environment is, the more reached

interested forwarders, and the less delay exerted by

the algorithm as illustrated in Figure 2b.

For the two disjoint subgraphs distribution, it

performs better than IPeR terms of delay by 33% in

low users’ density and in terms of the delivery ratio

by 1% in high users’ density. Figure 2c indicates the

performance in dense environments. In this such

challenging environment, where there are no

interested forwarders, Directional-IPeR-75

performed better than IPeR as it can approach more

delivery ratio with a slight increase of delay in dense

environments and perform equally in terms of

delivery ratio with decreased delay in sparse

environments. That is why it is not tested among the

other two algorithms, Dir-IPeR_75Int_Random, and

DMIPeR_75.

PECCS 2020 - 10th International Conference on Pervasive and Parallel Computing, Communication and Sensors

5.2 Directional-IPeR with Random

Discard

Generally speaking, Directional-IPeR-75-random

performed better in all metrics compared to IPeR. For

uniform distribution, the enhanced F-measure (up to

8% for 50 users) and cost (up to 7% for 50 users). For

the normal distribution, it enhanced F-measure (up to

18% for 100 users) and cost (up to 21% for 200 users).

The denser the environment is, the more delivery

ratio, the more number of reached interested

forwarders, and the less delay as illustrated in Figures

2a, 2b, 2c, and 3.

Figure 2: Performance of all algorithms within 200 users’

density.

Figure 3: Delay for 50 users, *setting 1= Uniform Interest

Distribution, setting 2 = Normal Interest Distribution and

setting 3= 2 Disjoint Subgraphs Interest Distribution.

5.3 DMIPeR: Hybrid Directional

Multi Hops IPeR

Directional-IPeR-75, which is the best algorithm for

Directional-IPeR, performs equal to its corresponding

DMIPeR in all metrics. However, it increases cost (up

to 36% for 100 users’ density in Uniform Interest

Distribution, up to 13% for 50 users’ density in

Normal Interest Distribution, up to 61% for 200

users’ density in 2 Disjoint Subgraphs Interest

Distribution). Adding 2 hops did not gain any

0,00

0,10

0,20

0,30

0,40

0,50

0,60

0,70

0,80

0,90

1,00

2a. Uniform Interest Distribution 200

users

IPeR 2Har MIPeR Dir-IPeR_75 Dir-IPeR_75Int_Random DMIPeR_75

0,00

0,10

0,20

0,30

0,40

0,50

0,60

0,70

0,80

0,90

1,00

Forwarders F-measure Cost Delay

2.b Normal Interest Distribution 200

users

IPeR 2Har MIPeR

Dir-IPeR_75 Dir-IPeR_75Int_Random

DMIPeR_75

0,00

0,10

0,20

0,30

0,40

0,50

0,60

0,70

0,80

0,90

1,00

Destinations F-measure Cost Delay

2c. 2 Disjoint Subgraphs Interest

Distribution 200 users

IPeR 2Har MIPeR Dir-IPeR_75

0,03

0,04

0,05

0,06

0,07

0,08

0,09

0,10

0,11

0,12

0,13

Delay for 50 Users Settings

Directional-IPeR: Enhanced Direction and Interest Aware Peoplerank for Opportunistic Mobile Social Networks

enhancement in all metrics as illustrated in Figure 2a

and Figure 2b, 2c, and 3.

6 DISCUSSION AND

COMPARISON

To authenticate the performance of Directional-IPeR-

75, direction awareness is integrated into other

algorithms that IPeR proved to be their superior.

These algorithms include the Interest aware

forwarding algorithm and PeopleRank algorithm. The

resulted algorithms are Directional-Interest and

Directional-PeopleRank, respectively.

Figure 4: Comparing Directional-IPeR to Dir-Interest and

Dir-PeopleRank.

Comparing Directional-IPeR-75 to Directional-

Interest, they performed equally in all metrics,

however, Directional-IPeR-75 performed better in

terms of cost up to 17% in sparse density for uniform

distribution as illustrated in Figures 4a and 4b. For the

normal distribution, the enhancement in cost is up to

11%. In high users’ density, they performed equally

for uniform distribution but Directional-IPeR-75

0,00

0,20

0,40

0,60

0,80

1,00

4a. Uniform Interest Distribution 50

users

Dir-IPeR_75 Dir_Interest_75 Dir_PeopleRank

0,00

0,20

0,40

0,60

0,80

1,00

4b. Uniform Interest Distribution 200

users

Dir-IPeR_75 Dir_Interest_75 Dir_PeopleRank

0,00

0,10

0,20

0,30

0,40

0,50

0,60

0,70

0,80

0,90

1,00

Forwarders F-measure Cost Delay

4.c Normal Interest Distribution 50

users

Dir-IPeR_75 Dir_Interest_75 Dir_PeopleRank

0,00

0,10

0,20

0,30

0,40

0,50

0,60

0,70

0,80

0,90

1,00

Forwarders F-measure Cost Delay

4.d Normal Interest Distribution 200

users

Dir-IPeR_75 Dir_Interest_75 Dir_PeopleRank

PECCS 2020 - 10th International Conference on Pervasive and Parallel Computing, Communication and Sensors

enhanced cost by 12% for normal distribution as

illustrated in Figures 4c and 4d.

Comparing Directional-IPeR-75 to Directional-

PeopleRank, Directional-IPeR-75 performed better in

terms of F-measure and cost up to 19% in low users’

density for uniform distribution. For normal

distribution, the enhancement in F-measure is up to

7% and the cost is up to 14%. In high users’ density,

Directional-IPeR enhanced F-measure by 17% and

the cost is up to 39% for uniform distribution

illustrated in Figures 4a and 4b. For the normal

distribution, Directional-IPeR enhanced F-measure

up to 8% and cost up to 13% illustrated in Figures 4c

and 4d.

Figure 5: Comparing Directional-IPeR to SocialCast.

Further, Directional-IPeR-75 is compared to the

SocialCast algorithm (Costa, Mascolo, Musolesi, &

Picco, 2008) in a uniform distribution setting.

Directional-IPeR-75 performed better in all metrics

except cost in high and low density of users as

illustrated in Figures 5a and 5b. Directional-IPeR-75

performs better than SocialCast in terms of delivery

ratio (up to 39% for 50-user experiments), reached

interested forwarder nodes (up to 87% for 200-user

experiments), and F-measure (up to 51% for 50 user

experiments). Also, it reduces delays (up to 1200%

for 200 users).

7 CONCLUSION

In this paper, we have taken the ﬁrst steps towards

showing the impact of incorporating direction

awareness with opportunistic forwarding algorithms

such as IPeR, PeopleRank, and Interest aware

forwarding algorithm. The proposed algorithms are

Directional-IPeR, Directional-PeopleRank, and

Directional-Interest, respectively. Furthermore, it

outperforms the state of the art social-based

opportunistic algorithm, SocialCast. Our simulation-

based evaluation demonstrates the promising gain in

delivery ratio, the number of reached interested

forwarders, delay, and F-measure. Our contribution is

defined as (1) Including direction awareness with

tolerance up to 75% less than the IPeR value of the

message holder (exemplified in the Directional-IPeR-

75 version) improves delivery ratio and the number

of reached interested forwarders. However, when

some of the uninterested forwarders did not

participate in messages delivery, which is a realistic

behavior, the performance is enhanced and generally

performed better in all metrics compared to IPeR. (2)

Adding multiple hops to directional guided IPeR does

not gain any enhancement. (3) Directional-IPeR-75

performs better in high densities in all metrics except

delay. Even though, it enhances delay in sparse

environments. Consequently, it can be utilized in

rural or disastrous areas, in which few people have

internet access with low connectivity and spread over

a big area.

ACKNOWLEDGEMENTS

I would first like to thank my family, especially Mom

and late Dad, for the continuous support they have

given me throughout my entire life; I could not have

done this research without them. My late Dad, I wish

0,00

0,20

0,40

0,60

0,80

1,00

5a. Uniform Interest Distribution 50

users

Dir‐IPeR_75 SocialCast

0,00

0,20

0,40

0,60

0,80

1,00

5b. Uniform Interest Distribution 200

users

Dir‐IPeR_75 SocialCast

Directional-IPeR: Enhanced Direction and Interest Aware Peoplerank for Opportunistic Mobile Social Networks

you were here. May God bless your soul in heaven.

Second, my daughters who brought happiness and

inspiration to every day. Thanks go out to my

advisors, Professor Sherif Aly and Professor Soumaia

Al Ayyat, for their incredible guidance (academic,

scientific, and otherwise) through the course of this

research. My fellow graduate students need to be

acknowledged for making my experience in graduate

school truly pleasurable.

REFRENCES

Abdelkader, T., Naik, K., Nayak, A., Goel, N., &

Srivastava, V. (2016). A performance comparison of

delay-tolerant network routing protocols. IEEE

Network, 30(2), 46-53.

Al Ayyat, A., Aly, S., & Harras, K. (2019). On Integrating

Space Syntax Metrics with Social-aware Opportunistic

Forwarding. IEEE Wireless Communications and

Networking Conference (WCNC). I, 1-7.

Al Ayyat, S., Aly, S., & Harras, K. (2016). SAROS: A

social-aware opportunistic forwarding simulator. IEEE

Wireless Communications and Networking Conference,

1-7.

Al Ayyat, S., Harras, K., & Aly, S. (2013). Interest aware

peoplerank Towards effective social-based

opportunistic advertising. IEEE Wireless

Communications and Networking Conference

(WCNC), Shanghai, 4428-4433.

Bank, J., & Cole, B. (2008). Calculating the jaccard

similarity coefficient with map reduce for entity pairs

in wikipedia. Wikipedia Similarity Team, 1-18.

Barkhuus, L., & Dey, A. (2003). Location-Based Services

for Mobile Telephony: a Study of Users' Privacy

Concerns. Interact, 702-712.

Bulut, E., & Szymanski, B. (2012). Exploiting friendship

relations for efficient routing in mobile social networks.

IEEE Transactions on Parallel and Distributed

Systems, 23(12), 2254-2265.

Burgess, J., Gallagher, B., Jensen, D., & Levine, N. (2006).

Max Prop: Routing for Vehicle-Based Disruption-

tolerant Networks. Proceedings IEEE INFOCOM,

25TH IEEE International Conference on Computer

Communications, (pp. 1-11).

Chang, J., & Chen, C. (2014). CROP: Community

relevance-based opportunistic routing in delay tolerant

networks. EICE TRANSACTIONS on Communications,

E97-B(9), pp. 1875-1888. Retrieved from

https://ir.nctu.edu.tw/bitstream/11536/25183/1/000342

729800015.pdf

Chen, Z., & Wu, J. (2016). Applying a sensor energy supply

communication scheme to big data opportunistic

networks. TIIS, 10(5), 2029-2046.

Cherif, W., Khan, M., Filali, F., Sharafeddine, S., & Dawy,

Z. (2017). P2p group formation enhancement for

opportunistic networks with wi-fi direct. IEEE Wireless

Communications and Networking Conference

(WCNC), 1-6.

Costa, P., Mascolo, C., Musolesi, M., & Picco, G. (2008).

Socially-Aware Routing for Publish-Subscribe in.

IEEE Journal on selected areas in communications,,

26(5), 748-760.

Daly, E., & Haahr, M. (2007). Social network analysis for

routing in disconnected delay-tolerant manets.

Proceedings of the 8th ACM international symposium

on Mobile ad hoc networking and computing, 32-40.

Deng, X., Chang, L., Tao, J., Pan, J., & Wang, J. (2013).

Social profile-based multicast routing scheme for

delay-tolerant networks. IEEE International

Conference on Communications (ICC), 1857-1861.

Denko, M. K. (Ed.). (2016). Mobile Opportunistic

Networks: Architectures, Protocols and Applications.

CRC Press.

Dhurandher, S., Borah, S., Woungang, I., Bansal, A., &

Gupta, A. (2018). A location Prediction-based routing

scheme for opportunistic networks in an IoT scenario.

Journal of Parallel and Distributed Computing, 118,

369-378.

Dubois-Ferriere, H., Grossglauser, M., & Vetterli, M.

(2003). Age matters: efficient route discovery in mobile

ad hoc networks using encounter ages. Proceedings of

the 4th ACM international symposium on Mobile ad

hoc networking & computing , 257-266.

Gondaliya, N., Kathiriya, D., & Shah, M. (2016). Contact

frequency and contact duration based relay selection

approach inside the local community in social delay

tolerant network. Proceedings of 3rd International

Conference on Advanced Computing, Networking and

Informatics (pp. ). Springer, ., 609-617.

Guo, B., Zhang, D., Yu, Z., Zhou, X., & Zhou, Z. (2012).

Enhancing spontaneous interaction in opportunistic

mobile social networks. Communications in Mobile

Computing, 1(1), 1-6.

Hom, J., Good, L., & Yang, S. (2017, January). A survey of

social-based routing protocols in delay tolerant

networks. Computing, Networking and

Communications (ICNC), (pp. 788-792).

Hossen, S., & Rahim, M. (2016). Impact of mobile nodes

for few mobility models on delay-tolerant network

routing protocols. International Conference on

Networking Systems and Security (NSysS), 1-6.

Hu, X., Chu, T., Leung, V. C., Ngai, E. C., Kruchten, P., &

C.Chan, H. (2015). A survey on mobile social

networks: Applications, platforms, system

architectures, and future research directions. IEEE

Communications Surveys & Tutorials, 17(3), 1557-

1558.

Hui, P., Crowcroft, J., & Yoneki, E. (2011). BUBBLE Rap:

Social-based Forwarding in Delay Tolerant Networks.

IEEE Transactions on Mobile Computing, 10(11),

1576-1589.

Jain, S., Chawla, M., Soares, V. N., & Rodrigues, J. J.

(2016). Enhanced fuzzy logic‐based spray and wait

routing protocol for delay tolerant networks. .

International Journal of Communication Systems,

29(12), 1820-1843.

PECCS 2020 - 10th International Conference on Pervasive and Parallel Computing, Communication and Sensors

Jeon, M., Kim, S. K., Yoon, J. H., Lee, J., & Yang, S. B.

(2014). A direction entropy-based forwarding scheme

in an opportunistic network. Journal of Computing

Science and Engineering, 8(9), 173-179.

Kawecki, M., & Schoeneich, R. O. (2016). Mobility-based

routing algorithm in delay tolerant networks. EURASIP

Journal on Wireless Communications and Networking,

81-90.

Kim, S. -K., Choi, J. -H., & Yang, S. B. (2015). Hotspot:

Location-based Forwarding Scheme in an

Opportunistic Network. Adhoc & Sensor Wireless

Networks(26).

LeBrun, J., Chuah, C., Ghosa, l. D., & Zhang, M. (2005).

Knowledge-based opportunistic forwarding in

vehicular wireless ad hoc networks. IEEE 61st

Vehicular Technology Conference, 2289-2293.

Lee, K., Hong, S., K, S. J., Rhee, I., & Chong, S. (2012).

SLAW: Self-Similar Least-Action Human Walk.

IEEE/ACM Transactions on Networking, 20, 515-529.

Li, W., Joshi, A., & Finin, T. (2010). Coping with node

misbehaviors in ad hoc networks: A multi-dimensional

trust management approach. in 2010 Eleventh

International Conference on Mobile Data

Management, 85–94.

Liu, L., & Jing, Y. (2012). A Survey on Social-Based

Routing and Forwarding Protocol in Opportunistic

Networks. Computer and Information Technology

(CIT), IEEE 12th International Conference.

Meng, X., Xu, G., Guo, T., Yang, Y., Shen, W., & Zhao, K.

(2019). A novel routing method for social delay-

tolerant networks. Tsinghua Science and Technology,

24(1), 44-51.

Moati, N., Otrok, H., Mourad, A., & Robert, J.-M. (2013).

Reputation-based cooperative detection model of

selfish nodes in cluster-based qos-olsr protocol.

Wireless Personal Communications, 75(3), 1747–

1768.

Mtibaa, A., May, M., & Ammar, M. (2012). Social

forwarding in mobile opportunistic networks: A case of

peoplerank. In Handbook of Optimization in Complex

Networks (pp. 387-425). New York, NY: Springer.

Pal, S., Saha, B. K., & S.Misra. (2017). Game Theoretic

Analysis of Cooperative Message Forwarding in

Opportunistic Mobile Networks. IEEE transactions on

cybernetics, 47(12), 4463-4474.

Palla, G., Derényi, I., Farkas, I., & Vicsek, T. (n.d.).

Uncovering the overlapping community structure of

complex networks in nature and society. Nature,

435(7043), 814-824.

Pathak, S., Gondaliya, N., & Raja, N. (2017). A survey on

PRoPHET based routing protocol in delay tolerant

network. Emerging Trends & Innovation in ICT (ICEI),

2017 International Conference, (pp. 110-115).

Perrig, A., Stankovic, J., & Wagner, D. (2004). Security in

wireless sensor networks. 53-57.

Rajpoot, N., & Rajendra, S. K. (2015). An improved

PRoPHET routing protocol for underwater

communication. In Communication Networks (ICCN),

27-32.

Shahin, Y. S., Al Ayyat, S. A., & Aly, S. G. (2020). MIPeR:

Enhanced Multiple Hops Interest Aware PeopleRank

for Opportunistic Mobile Social Networks. American

University in Cairo, Computer Science and

Engineering, Cairo.

Sobin, C., Raychoudhury, V., Marfia, G., & Singla, A.

(2016). A survey of routing and data dissemination in

delay tolerant networks. Journal of Network and

Computer Applications(67), 128-146.

Song, L., & Kotz, D. F. (2016). Routing in Mobile

Opportunistic Networks. . Mobile Opportunistic

Networks: Architectures, Protocols and

Applications(1).

Spaho, E., Bylykbashi, K., Barolli, L., Kolici, V., & Lala,

A. (2016). Evaluation of different DTN routing

protocols in an opportunistic network considering

many-to-one communication scenario. 19th

International Conference on Network-Based

Information Systems (NBiS).

Spyropoulos, T., Psounis, K., & Raghavendra, C. S. (2005).

Spray and Wait: an efficient routing scheme for

intermittently connected mobile networks. Proceedings

of ACM SIGCOMM 2005 Workshop on Delay Tolerant

Networking and Related Networks (WDTN-2005), (pp.

252-259). Philadelphia.

Sui, Z. (2015). The Research of the Route Protocols in

Opportunistic Network. Computational Intelligence

and Communication Networks (CICN), 192-196.

Takasuka, H., Hirai, K., & Takami, K. (2018).

Development of a Social DTN for Message

Communication between SNS Group Members. Future

Internet, 10(4), 32.

Vahdat, A., & Becker, D. (2000). Epidemic routing for

partially connected ad hoc networks. Technical report

CS-200006, Duke University.

Vasilakos, A., Spyropoulos, T., & Zhang, Y. (2016). Delay

tolerant networks: Protocols and applications. CRC

press.

Vastardis, N., & Yang, K. (2013). Mobile social networks:

Architectures, social properties, and key research

challenges. IEEE Communications Surveys & Tutorials

15(3), 1355-1371.

Wu, J., & Chen, Z. (2016). Data decision and transmission

based on mobile data health records on sensor devices

in wireless networks. Wireless Personal

Communications, 90(4), 2073-2087.

Wu, J., & Wang, Y. (Eds.). (2014). Opportunistic Mobile

Social Networks. CRC Press.

Zhu, Y., Xu, B., Shi, X., & Wang, Y. (2013). A survey of

social-based routing in delay tolerant networks:

Positive and negative social effects. IEEE

Communications Surveys & Tutorials, 15(1), 387-401.

Directional-IPeR: Enhanced Direction and Interest Aware Peoplerank for Opportunistic Mobile Social Networks