GPS: A Method for Data Sharing in Mobile Social
Networks
Bo Fan
∗
, Supeng Leng
∗‡
, Kun Yang
†
and Qiang Liu
∗
∗
School of Communication and Information Engineering,
University of Electronic Science and Technology of China, Chengdu, China
†
School of Computer Science and Electronics Engineering,
University of Essex, Colchester, United Kingdom
‡
Corresponding author, Email: spleng@uestc.edu.cn
Abstract—In Mobile Social Networks (MSNs), users with
specific relationships are usually treated as a community for data
sharing. However, the demand of data sharing among distributed
strangers also exists. Those users that have the same interest but
do not necessarily know or usually encounter each other can form
a gossip community and share information. This paper proposes
a data dissemination approach, i.e., the Gathering Point-aided
Spreading (GPS) algorithm, which explores the encounter pattern
of users and the aid of gathering points to facilitate data sharing
in the gossip community. Based on the past encounter pattern, the
GPS algorithm predicts the encounter probability among users
and assigns the best users to carry the data for a wide spreading.
Moreover, by storing a copy of data at the gathering points, GPS
enables a further sharing of the data even the carriers leave the
gathering points. With different utility functions, GPS can be
modified into three versions (GPS-DR, GPS-DE and GPS-TR).
Simulation experiments show that GPS outperforms SocialCast
in both delivery ratio and delay in data sharing. In addition,
among the three versions, GPS-DR and GPS-DE perform the
best in terms of delivery ratio and delay respectively, while GPS-
TR makes a tradeoff between them.
Index Terms—Mobile Social Networks, gossip community, data
sharing, gathering point.
I. INTRODUCTION
Mobile Social Network (MSN) is a special type of Delay
Tolerant Network (DTN), which explores the social attributes
(interest, social relationship, daily schedule, etc.) of users
to improve the performance of the services. In MSN, users
with strong relationships (such as friends, one-hop neighbors,
frequently encountered strangers, etc.) are usually regarded as
a community that has a strong inner connection to share or
help delivery data. For example, users with common interest
keywords, similar location histories and nearby current loca-
tion are recommended as friends for data sharing [1]. This
idea is reasonable but confined. In fact, distributed strangers
without the above relationships also have the demand for data
sharing. For instance, the housewives in a town are usually
interested in the discount in the supermarket and are willing
to share the discount information if possible despite the weak
relationship among them.
We define such a group of people that have common interest
but are usually weak in relationship as a gossip community.
They are located usually within the same piece of area but
do not necessarily know or encounter each other, such as
the housewives mentioned above. Different from traditional
social communities [2][14] that have strong inner connection
(social relationship, co-location, etc.), the gossip community
has a loose connection among the community members. The
only relationship among them is the common interest. Gossip
communities widely exist in our social life, however, few
proposal in MSN has focused on this type of community for
efficient data sharing. In literature [3], a data sharing method
is proposed. However, when the users are multiple hops away
from each other, the sharing of data is based on the Ad hoc On-
demand Distance Vector (AODV) [4] routing protocol, which
is infeasible in MSN for the intermittent connection among
users.
This paper proposes Gathering Point-aided Spreading (G-
PS) algorithm, a data dissemination approach that adopts the
idea of gossip spreading, namely first letting some gossipers
get the information, and then the gossipers spreading it out
quickly. The gossiper is assumed to be selfless to carry the
data. In order to facilitate the sharing of data, GPS explores
the past encounter pattern of users and the aid of Gathering
Point (GP) for help. Based on the past encounter pattern
among users, GPS predicts their future encounter probability
and assigns the best users as data carriers (gossipers) that take
the data for a wide spreading. Data carriers are helpful for
the sharing of data, for they can take data to the users that
are not going to encounter the data source and overcome the
shortcoming of loose connection among members of the gossip
community. The assignment of data carriers is based on the
utility function of users. In this paper we establish three utility
functions based on the predicted probability to select the best
data carriers respectively aiming at three performance objects
in data sharing: maximizing delivery ratio, minimizing delay
and balancing the tradeoff between them.
The GP is the popular place that usually attracts a large
number of users, such as a bar. It is very helpful in facilitating
the spreading of data as it brings an opportunity for some
interested users to share data directly and for the data source
to select the excellent data carriers due to the large number
of users gathering. In GPS, it is assumed that each GP hasISBN 978-3-901882-58-6
c
2014 IFIP