Int. J. Communications, Network and System Sciences, 2011, 4, 456-463
doi:10.4236/ijcns.2011.47055 Published Online July 2011 (http://www.SciRP.org/journal/ijcns)
Copyright © 2011 SciRes. IJCNS
A Framework for Security-Enhanced Peer-to-Peer
Applications in Mobile Cellular Networks
Shuping Liu1, Shushan Zhao2, Weirong Jiang3
1Department of Electrical Engineering, University of Southern California, Los Angeles, USA
2Department of Computer Science, University of Windsor, Windsor, Canada
3Juniper Networks Inc, Sunnyvale, USA
E-mail: lius@usc.edu, zhao114@uwindsor.ca, weirongj@acm.org
Received May 24, 2011; revised June 21, 2011; accepted July 1, 2011
Abstract
Due to the dual trends of increasing cellular network transmission capacity and coverage as well as improv-
ing computational capacity, storage and intelligence of mobile handsets, mobile peer-to-peer (MP2P) net-
working is emerging an attractive research field in recent years. However, these trends have not been clearly
articulated in perspective of both technology and business. In this paper, we propose a novel MP2P frame-
work that is based on existing cellular network architecture to provide secure and efficient P2P file sharing
for 3G and future 4G systems. Our framework, which is built on P2P over Session Initiation Protocol (SIP)
mechanism, provides to network operators and P2P service providers efficient data transmission in cellular
networks. With a secure enhancement using identity-based cryptography, the framework also provides de-
sirable support for security, group management, mobility, and chargeability to meet business requirements.
Keywords: Mobile Cellular, 3G, P2P, SIP, IMS, Identity-Based Cryptography
1. Introduction
Mobile cellular networks and handsets are developing ra-
pidly in recent years. While GSM systems have proven
successful, 3G systems (including W-CDMA, CDMA-
2000, and TD-CDMA/TD-SCDMA) are burgeoning and
4G is on the way. These new technologies have brought
many conveniences to our life and in some way changed
the life style of modern people, although there are also
some criticisms against this change.
The P2P network has emerged as an efficient system,
being typically used for sharing content files containing
audio, video, data or any digital-format files and distrib-
uting services over fixed networks. Currently, P2P appli-
cations are considered to be generating most of the
internet traffic [1,2].
Meanwhile, the technology of P2P has recently begun
to extend its scope to address relevant problems of mo-
bile systems in the cellular networks. P2P data exchange
is a very promising business in cellular networks, but it
has attracted very little academic interest due to both busi-
ness and technical reasons. On the one hand, the business
model regarding how this business can operate and make
profit is not yet very clear. On the other hand, although
there is no technique impediment for applying P2P in
cellular networks theoretically, there is no complete so-
lution taking practical issues into consideration, e.g. se-
curity, group management, mobility, and chargeability.
There are many complications considering the complex-
ity of cellular networks. A P2P application makes a cellu-
lar network more prone to security issues such as trust
(privacy: how much information does the un-trusted peer
need to know about me? and confidentiality: what if the
peer who knows my information misuses it?) and DoS
attacks. Techniques widely applied in Internet, such as
PKI, are not suitable for cellular networks, due to spe-
cific characteristics of the latter, such as lack of infra-
structure, constrained communication and computational
resources etc. A reliable framework for authentication
without centralized elements is a challenge [3].
To inspire more interest and promote research on this
topic, we would like to propose a MP2P solution frame-
work in this paper. The framework considers P2P appli-
cations in cellular networks in particular 3G networks,
based on IMS and SIP signaling. The framework has a
security add-on layer to meet business requirements,
such as user authentication, non-repudiation, data and
identity integrity, confidentiality.
S. P. LIU ET AL.
Copyright © 2011 SciRes. IJCNS
457
The rest of the paper is organized as follows. In Sec-
tion 2, we will present types of architectures for MP2P
and propose a hybrid P2P architecture which can be ap-
plied to IMS-based P2P networks, and introduces the
application of SIP in P2P network. In Section 3, we first
state the security requirements for the MP2P system, and
then propose a solution. To evaluate the performance of
the proposed framework we build a mathematical model
that takes as input search and download queries, and re-
turns as output hit-rate probabilities and time delay. In
Section 4, we analyze the system performance and secu-
rity features mathematically. The last section concludes
the paper.
2. A MP2P Framework in IMS
2.1. Types of Traditional P2P Architectures and
Proposed Hybrid P2P Architecture
Traditionally, there are two types of P2P architectures:
centralized P2P and Decentralized P2P.
Centralized P2P: In wired network, centralized P2P
is the first generation architecture which uses several
central index servers to control the whole network
and provide metainformation service, such as index
of files for sharing by each peer. To participate in
these networks, the peer must connect as a client to
the centralized server and then locate specific contents.
When the peer is located, the requesting node starts
the transfer directly from the located peer.
Decentralized P2P: The decentralized P2P architec-
ture lacks central entities which could control the
network, and every node in the network acts as a peer
which is unaware of other peers. All the information,
including metainformation, is maintained by peers.
The decentralized architecture can be divided into
structured and unstructured. Though search and down-
load queries are flushed in both structured and un-
structured P2P network, queries in structured P2P
network are easier to be hit than unstructured P2P
network. That is because the topology of structured
P2P network is tightly controlled and files for sharing
are placed at specific locations. Such kinds of sys-
tems usually deploy Distributed Hash Table (DHT).
The unstructured P2P networks have no precise con-
trol over the network topology or the file placement.
Queries in such systems are usually forwarded to
random neighbors [4]. In decentralized P2P network,
flushing queries consumes a lot of time and band-
width.
Semi-centralized P2P: The semi-centralized P2P
architecture combines the efficiency and resilience of
both centralized and decentralized architectures. It
structures the network in hierarchies by establishing a
backbone network of super peers which function as
the central index servers in centralized P2P. As a re-
sult the whole P2P network is partitioned by these
super peers into several sub-networks logically. Each
logical sub-network is characterized by centralized
P2P architecture. When a client peer logs on to the
network, it makes a direct connection to a single su-
per peer which collects and maintains information
about client peers and content available for sharing.
Comparison among different P2P architecture is given
in Table 1.
In the context of cellular networks, a cellular network
covers a certain area that is divided into possibly over-
lapping cells. Each cell has a fixed base station (BS). The
base stations are connected to each other by a wired
network. In cellular network, radio link is costly and in-
adequate. For this reason, in normal structure of cellular
communication, there is no direct correspondence be-
tween cellular phones. Several architectures have been
proposed, such as mobile hash-based structured P2P ar-
chitecture (HS-P2P) [5] and mobile agent P2P architec-
ture [6].
In this paper we proposed a hybrid P2P architecture
that combines the efficiency and resilience of both cen-
tralized and decentralized architectures [7]. By using
semicentralized architecture, each mobile terminal con-
nects to a single super peer directly, which is basically
Base Station (BS) of cellular networks. The super peer
collects and maintains information about peers and me-
tainformation of contents available for sharing. Then all
these super peers form a decentralized P2P backbone
Table 1. Comparison of different P2P architectures.
Architecture CentralizedDecentralized Semi-Centralized
Network
Scalability Medium Low Very High
Resiliency Low Very High Medium
Search
Efficiency Very HighMedium High
Search
Coverage Very HighMedium High
Operator
Control Very HighLow High
Signaling
Overhead in
SP
Very High-- High
Signaling
Overhead in
OP
Low High Low
S. P. LIU ET AL.
Copyright © 2011 SciRes. IJCNS
458
network. In this way, the traffic loads are moved to the
network (super peers) side, which minimizes the usage of
the radio link. Additionally, due to the two-layer hierar-
chies, semi-centralized architecture is more scalable.
By using SP, the performance of P2P applications is-
dramatically improved. At the same time, decentralized
P2P applications are also supported for peers with spe-
cial requirements, for example high privacy in their com-
munication.
2.2. Database Tables in Super Peers
Each super peer maintains four tables: content table, cli-
ent table, caching table, membership table.
The content table stores the information about content
for sharing in its network and facilitates SP to locate the
client peers (in following sections client is used to refer
to client peer). To improve searching efficiency, SP
builds three levels index for the lookup table. The first
level index consists of different classes of content. For
instance, the first level index can be categorized into: mu-
sic, picture, literature and so on. The second level com-
prises different clusters in each class in the first level,
which divides the first level classes into more detailed
sub-classes. The third level is made up of leaf nodes
which are the specific metainformation of content file for
sharing. Each file has a unique id.
To illustrate the use of index table, let us look at an
example. One client has a piece of song, called “Country
Road” for sharing. According to the classification of in-
dex, this song belongs to class of music in the first level
index and country music in the second level index. SP
obtains this information and adds the new item along
with the file profiles, such as ownership and file size into
its content table. Because the content of shared file with
the same name may be different, each content table also
contains the basic information of shared file, such as file
size, modified time and so on.
In client table, each entry is a client which is identified
by the client’s ID (such as the telephone number). Each
entry records the specific bandwidth and membership
information of one client. The client table provides the
parameters which SP can choose so that QoS is guaran-
teed when downloading the files.
The caching table, by nature, is similar with the con-
tent table. The only difference is that caching table stores
the sharing content owned by clients in other SP net-
works. For example, initially the caching table is empty.
When a SP A can not find the requested information in
its content table, it floods the requests to all the other SPs.
If a SP B responses to SP A, which means B has the re-
quested information, A locates the client (who has the
requested resource) and at the same time stores the re-
sponse information and corresponding SP in its caching
table. The caching table is updated in the way of Least
Recently Used (LRU) cache, in which the items are
saved in decreasing order by the times of most recent
request. More popular the file, closer it is to the top of
the lookup table. To fit the fixed length of caching table,
once response information reaches the end of the list, it is
removed from the caching table, when a new item is
added. Once the caching table has been built, the SP may
obtain its next hop without flushing the request message,
which, to some extends, reduces traffic flooding.
The membership table contains information about
groups. Groups are categorized by interest. To participate
or leave a group, the client has to apply to SP. SP exam-
ines whether the group ID in the request message matches
the group ID in the response message.
2.3. Backup Super-Peer
In the introduced MP2P hybrid architecture, the super-
peer becomes a single-point failure for its network, when
super-peer fails or be offline the all shared information in
the network is lost. To increase the reliability of the ar-
chitecture, the backup super-peer is introduced. They
copy all the data information from the super-peer peri-
odically. When the super-peer fails or be offline net-
work the backup peer replaces it and operates as a super-
peer. The possibility of both a super-peer and its backup
peer failing at the same time is much smaller than failure
of super-peer alone. Therefore, the introduction of the
backup super peer improves the architecture's robustness
greatly.
2.4. P2P Based on SIP in the IMS
To apply the hybrid architecture in the IMS, it is natural
to implement the SP as an SIP application server (SP-AS)
which hosts and executes services. SP-AS interfaces with
the S-CSCF (Serving Call/Session Control Function)
using SIP in the ISC (IMS Service Control) interface, as
shown in Figure 1. The ISC interface is between the
Serving CSCF and the service platforms. The SP-AS
uses the ISC interface to communicate with the S-CSCF.
The S-CSCF is connected to a P-CSCF (Proxy Call/Ses-
sion Control Function) through the Mw interface. The
Mw reference point allows the communication and for-
warding of signaling messaging between CSCFs, e.g.
during registration and session control [8]. The S-CSCF
and PCSCF are SIP proxies with IMS functionality. The
SCSCF is the central node in the signaling plane and the
P-CSCF is the inbound/outbound SIP proxy between the
terminal and the IMS network [9].
S. P. LIU ET AL.
Copyright © 2011 SciRes. IJCNS
459
Figure 1. Interfaces and elements of P2P in IMS.
The SP-AS server has three main functions: content
management, cache management and membership ser-
vices. The SP-AS server maintains the directory of meta-
information. When he/she registers to the network, the
client sends the initial list of content for sharing as well
as further updates as the content changes to the SPAS.
When a client requests resources, the SP-AS consults its
lookup tables and replies back with matching contents. It
as well forwards the requests to other SP-AS.
SIP provides an established method for routing the
P2P connection to the end peer utilizing a set of proxy
servers. The INVITE message performing this function
carries the P2P control message in SDP (Session Descri-
ption Protocol) format. The SDP describes the streams
used for communication and allows the end peers to
agree on the session parameter.
3. Security Enhancement to the Framework
3.1. Security Requirements for MP2P Business
The MP2P framework proposed above provides a high-
performance content sharing model, and is the carrier of
P2P traffic, but it will not work without meeting security
requirements for MP2P business. From business perspec-
tive, security requirements come from two aspects: the
network operator and terminal users, although there is
some overlap between them. These requirements include:
Identity authentication and non-repudiation: verifies
that the data or request came from and goes to a spe-
cific, valid user. From network operator perspective,
this is to ensure that only subscribed users can get the
service, and bills are charged to correct users. From
end user perspective, this is to ensure communicating
with the correct peer.
Data confidentiality: keeps data secret to outsiders.
Only the destination user can get the data, and not
anybody else; even an outsider gets it, he/she cannot
get the meaning of the data. From the network opera-
tor perspective, this is to prevent free access to con-
tent, and encourage more traffic. From the end user
perspective, this is to prevent eavesdropping and pro-
tect their privacy.
Data integrity: prevents data from being altered.
Data freshness: keeps data in correct order and up-to-
date.
Data availability: data should be available on request.
With P2P-over-SIP architecture, the operator can
identify and control the transferred contents. As a SIP
signaling message passes through the SP, the content
type, content name and potentially some other informa-
tion can be extracted. This allows the operator to deny
access to illegitimate contents and have group manage-
ment.
Cellular networks have already provided security from
Mobile Switching Centre (MSC) to base stations, and
base station to handsets. GSM network access security
uses A3/A8 (COMP128 actually used in GPRS) authen-
tication algorithm and A5 encryption algorithm, which
have already been broken [10,11]. 3GPP has developed
proprietary cryptographic algorithms—the MILENAGE
algorithm set and KASUMI cryptographic core to re-
place the broken ones in GSM. However, much of the
work with the UMTS access architecture has been fo-
cused on backward compatibility with GSM/GPRS [12].
From a security point of view, backward compatibility
with a system with weaker security is undesirable but
dictated by commercial reality.
On the other hand, as for end-to-end security between
two cellular terminals, current 3G networks do not pro-
vide any mechanism. In Internet world, the traffic can be
protected by using Public Key Infrastructure (PKI). It is
not feasible in cellular networks, especially for handsets.
First, there is no such infrastructure in cellular networks.
Second, PKI is too heavy weight for cellular handsets.
In light of the above reasons, we propose to employ
identity-base cryptography (IBC) in this framework.
Such a scheme has the property that a user’s public key
is an easily calculated function of his identity, while a
user’s private key can be calculated for him by a trusted
authority, called Private Key Generator (PKG). The
identity-based public key cryptosystem can be an alter-
native for certificate-based PKI, especially when effi-
cient key management and moderate security are re-
quired.
Compared to PKI, it has the following advantages in
cellular networks:
Lightweight: PKI is implemented in RSA, and IBC is
implemented in Elliptic Curve Cryptography (ECC).
S. P. LIU ET AL.
Copyright © 2011 SciRes. IJCNS
460
For a security level of 2048-bit RSA, ECC outper-
forms in every aspect; for a security level of 1024-bit
RSA, ECC outperforms in scenarios such as in mo-
bile cellular networks [13]. This saves storage and
computational resources of the handsets.
Easy to deploy: In PKI, the Public Key Certificate
(PKC) of a user need be signed by the Certificate
Authority (CA) before being distributed to other users,
and a user need to store all PKC’s of all other users
he/she wants to communicate. In IBC, the public key
is implicit in the communication traffic, and there is
no need to have it signed, distributed and stored. This
saves communication and storage resources of the
network and handsets.
3.2. Basic System Setup
Our security scheme is based on Boneh’s implementation
of IBC [14]. The scheme needs a setup phase in which
system parameters are distributed to its users. These pa-
rameters include system public key, master key, private
key of each user, and algorithms to be used for hash-
ing/encryption/decryption.
The operator chooses a Bilinear Map denoted as ê:
11 2
GG G between two cyclic groups G1, G2 of
order q for some large prime q, where G1 is the group of
points of an elliptic curve over Fp and G2 is a subgroup
of 2
*
p
F
. At system setup phase, the operator sets the
system public key Ppub as sP where s is a random num-
ber in *
q
Z
, and P is an arbitrary point in E/Fp of order q.
It also chooses a cryptographic hash function G:

0, 1
F
p
to map variable identity strings to points
in E/Fp, and chooses hash functions H1:

2
*0, 1
p
F
n
,
H2: 2
*
p
F
Zq, H3:
 
0,1 0,1nnFp

, and H4:
 
0,1 0,1nm

for keys of specified length.
The initial master key
Zq
and the system pa-
rameters <p, n, P, Ppub, G, H1, H2, H3, H4 >are deter-
mined and calculated by the operator. The network op-
erator assigns a short code to each super peer in the net-
work and publishes the numbers. For a MP2P subscriber,
the operator uses the MSISDN as the unique identity and
hence public key. For each

0,1ID
, the crypto-
graphic scheme builds an initial private key dID as dID
= sQID where QID is a point in E/Fp mapped from ID.
Every user gets the system parameters and its private key
from the operator when he/she subscribes the MP2P ser-
vice.
3.3. Secured P2P Communication
The P2P applications installed on cellular terminals take
care of secure communication between two terminals and
between a terminals and a SP. When node A wants to
send a message to node B, either a SP or a regular peer
node, the application encrypts and authenticates the
message as follows:
1) A first generates an implicit shared key with B,
without any interaction with B:

1r
kHg, where
ˆ,
g
edAQB,
ˆ
2,rHeQAQB
,
QAG IDA,
QBGIDB, dA sQA
. IDB is the MSISDN or short
code of the receiver to whom the sender intends to send
the message.
2) A encrypts the message M, and outputs the cipher
text
4CEHkM, where Ek is a secure symmetric
cryptosystem encryption function.
3) A signs the message with its own private key and
the receiver’s public key using a message authentication
code (MAC) function, named H3 here, the authentication
code
3,
H
Ck
is determined by the message to be
sent and the private key of the originator, and serves as a
signature to the message signed by the node. The en-
crypted message C and signature σ are put into the pay-
load field of the packet, ,MC
.
At the receiver B side, the message is verified and de-
crypted as follows:
1) B first generates the implicit shared key with A,
without any interaction with A:

1r
kHg, where
ˆ,
g
eQAdB. Note that IDA is the MSISDN of sender
derived from the packet header and
ˆˆ
,,eQAdB edAQB.
2) For a received message ,MC
in the de-
fined format with signature σ and cipher text C, the sig-
nature is re-calculated:

3,
H
Ck
and verified
against the one in the received message. If they match,
the message is processed further. Otherwise, the message
is discarded.
3) For the received message, B decrypts it with the
shared key:
 
4
M
DHk C
 , where Dk is a secure
symmetric cryptosystem decryption function.
M
should
be the same as the original message M.
One weakness of this scheme is key escrow [15], i.e.
the network operator knows the encryption key between
A and B. To transfer content that A and B want to keep
secret from the operator, the following extra steps can be
executed for calculating a shared session key between A
and B after they get a pairwise secret key:
ˆˆ
,,
K
ABedA QBeQAdBKBA. KAB is then di-
vided into two parts Ke and Ka. Encryption under Ke
prevents all other networks nodes from reading the mes-
sages, whereas Ka is used in a message authentication
code to enable mutual authentication. Then, A
B: A,
EKe (K1), B
A: B, EKe(K2), MACKa (A,EKe(K1),
Eke (K2)), A
B: MACKa (B,EKe(K2), EKe (K1)). A
shared session key can be set up as
1, 2
K
sesf KK
[16]. This key provides forward security and prevents the
S. P. LIU ET AL.
Copyright © 2011 SciRes. IJCNS
461
operator from being a key escrow.
3.4. Group Generation and Communication
The MP2P architecture supports fixed group and ad-hoc
group for peers to share data by broadcasting to group
members. The security scheme generates group key for
each fixed group and ad-hoc group.
Each fixed group is formed by the network operator,
and includes one super peer and any number of regular
peers. The group broadcasting key is chosen by the super
peer, and can be generated from a random number. One
regular peer can join a super peer’s group by sending a
request message and get a reply message, including the
group key, from the super peer. These messages are en-
crypted and signed using the scheme introduced in Sec-
tion 3.3. After that, the super peer can announce infor-
mation by broadcasting to its subscribers. The informa-
tion can include the updates in its content table, client
table, caching table, and membership table.
Regular peers can generate ad-hoc groups by them-
selves, and exchange data that are protected to outsiders
using a group-wide secret key. The secret key is gener-
ated in this way:
1) Node 1 computes its broadcast parameter that is uni-
que for node 1:

ˆ
11,23
K
Ne sQidQidQidQidn

ˆˆ ˆ
1,21, 31,esQidQidesQid QidesQid Qidn and
distribute 11PbrdcstKNP to all candidate nodes
using respective pairwise encryption.
2) Node 1 use P1brdcst to encrypt a message sent to
the group, and sign a message as

1
11 1
:, ,
id Nid
MUVrQ KrhQ
 where r is a random
number and h is a hash of the original message.
Other group members decrypt the message with
1Pbrdcst, and verify if

ˆˆ
1, , 1ePbrdcst VeP UhQid
holds.
4. Discussion
4.1. Performance Analysis
From previous sections, we know that the search effi-
ciency depends on the search in different SP. The effi-
ciency of caching policy will be analyzed in the follow-
ing section. Suppose all of the SPs form a graph in which
all nodes have the same degree D. And there is up to Nf
different files for sharing being cached in snumber of N
SPs, with Ni files in each SP. Each SP reserves the cache
memory space for at most number of K files. Maximum
searching steps, which limits the searching steps, are


log11 1nDDNi. Let s be the possible
number of steps used to search the target.
And we assume searching steps and file location are
evenly distributed in the content table and caching table.
The analysis falls into the following three categories,
N × K = Nf and without search loops:
Assume that there are no duplicate contents in both the
content table and the caching table. When no repeat con-
tent and search loop exists, the probability for hitting the
target is,


1
0
0
00
11
12
1
11
1
hit
ffff
n
iff
n
iff
PPhits Ps
PhitsnPs n
KKKK
nN NNKNK
KK
NiKNnK
KK
nNjKNiK


 
 
 
 

 











N × K > Nf and without search loops:
At a certain time point, there are totally N × K files in
SPs and Nf is fixed. If Nf < N × K, the overlap of con-
tents occur between the content table and caching table.
The repeat probability is defined as follows Pr(n) the
probability for a certain cached file in the SP checked in
nth steps to be the same as some file in the SP checked
before, that is, in the SP checked in 0th , 1th, …, (n–1)th
steps.
00;
r
P

11
f
r
f
NK
K
PN
K



 




r
1P1
21 ;
f
r
f
NKK
K
PN
K

 



 





1
r
01P
1;
n
fi
r
f
NKi
K
Pn N
K





 


Therefore the hitting probability is,
S. P. LIU ET AL.
Copyright © 2011 SciRes. IJCNS
462

















r
1
11
0
0
0
1
0
0
11
11
1
11 12
11 11
11
1
1
1
1
1
1
1
1
hit
fff
rr
ff fr
nrr
jn
j
fr
fr
m
k
nnr
j
i
f
KP
KK
PnNN NK
KP KP
K
NNK
NKK P
KPj KPn
NKPm
NKPk
n
KPj
NK








 
 











 













11
00
1
1
r
jj
rf r
km
KPi
Pk NKPm










With search loops:
Now we consider the condition that loops are possible.
If the search process hits the target in n steps, then the
maximum number of the checked SPs is n, and the
maximum number of steps that can be wasted due to
loops is n - 2. The wasted step means that one SP that
has already been checked is checked for a second time.
Suppose the number of possible loop is evenly distrib-
uted.
Let

P
lab denotes the probability for hitting target
in a steps along with b steps wasted because of loops,
then,
 
11
P
labPlA aB bPlAaBb 

22PlAaBb Denote the probability for
hitting the target in nth steps by Sn, then,

000
l
SP;

110
l
SP;

220
l
SP;
   

3
111
303130 20
222
ll ll
SPPP P

2
10
1
n
nl
i
SPi
n
.
Then the hitting probability is as follows,
  
32
11
10 200
11
ni
hitl ll
ij
PPPPj
ni







where

0
l
Pj








1
11
0
00
11
1
11
jll
in
i
flfl
mi
KPi KPj
NKPmNKPi






 

 



.
From Figures 2 and 3, the advantages of the proposed
Figure 2. File number is 10000, n = 3, file caching room is in
the range of 100 to 2000, connectivity degree = 10. File re-
peat probability is 0.
Figure 3. File number is 10000, n = 2 to 3, file caching room
is 1000, connectivity degree = 5 to 40. File repeat probabil-
ity is 0.
architecture when evaluating hit probability are clear.
The hit probability is improved significantly as the cache
room and connectivity degree rises. Our results reasona-
bly show some trends for parameters changes.
4.2. Business Security Analysis
The security enhancement provides integrity, authentica-
tion, and confidentiality to MP2P messages by binding a
message with a private key possessed only by the cellular
terminal. It effectively prevents the following attacks
previously available in cellular networks:
Identity Impersonation: In cellular networks, the pos-
sibility exists that an attacker manages to impersonate
the service provider or a terminal user. With the pro-
posed security enhancement, the originating address
of the sender is bound to the private key of the sender.
The attacker, not knowing the private key, cannot
forge an arbitrary address. This ensures correctness of
delivered service and billing.
Message Forgery and Tampering: Similarly to ad-
dress forgery, an attacker can also forge or tamper the
payload field of a data packet, namely the content of
a message, in current cellular networks. With the pro-
posed security enhancement, every message is signed
by the sender. The attacker, not knowing the private
S. P. LIU ET AL.
Copyright © 2011 SciRes. IJCNS
463
key of the sender, cannot tamper the message and ge-
nerate a correct signature. It’s easy to verify the inte-
grity of the message.
Eavesdropping: As was stated earlier, there are many
possibilities an attacker can get access to messages
transmitted through current cellular networks. The
attacker can intercept the messages from the Internet
or over the air, and easily get interesting information,
since there is no strong protection to them. With the
proposed security enhancement, every message is
encrypted, and only the sender and receiver know the
decryption key. Any attacker will need a great deal of
effort if he wants to crack the encryption.
With these features, a MP2P service can be securely
deployed; security requirements from both the network
operator side and terminal user side can be well satisfied.
A business model based on this framework is thus feasi-
ble, profitable, and also promising.
5. Conclusions
In this paper, we propose a framework for implementing
P2P as a SIP-based service in cellular networks, in par-
ticular in IMS. The framework is based on MP2P hybrid
architecture. We show several benefits of this framework
by mathematical analysis and simulation. The framework
also includes a security enhancement, with which the
operators can have control on security and group man-
agement, and chargeability is possible. The enhancement
is lightweight and convenient to deploy in cellular net-
works environment by using identity-based cryptography.
Business model using this MP2P framework with secu-
rity enhancement can be easily and successfully setup,
which is one of our future works.
6. References
[1] N. B. Azzouna and F. Guillemin, “Experimental Analysis
of the Impact of Peer-to-Peer Applications on Traffic in
Commercial IP Networks,” European Transactionson
Telecommunications, Special Issue on P2PNetworking
and P2P Services, Vol. 15, No. 6, November-December
2004, pp. 511-522.
[2] T. Karagiannis, A. Broido, N. Brownlee, K. C. Claffy and
M. Faloutsos, “Is P2P Dying or Just Hiding?” Proceeding
s of IEEE Global Telecommunications Conference, Dal-
las, Vol. 3, 29 November-3 December 2004, pp. 1532-
1538.
[3] K. Singh and H. Schulzrinne, “Peer-to-Peer Internet Te-
lephony using SIP,” Columbia University Technical Re-
port, CUCS-044-04, New York, October 2004.
[4] J. Yang, Y. P. Zhong and S. Y. Zhang, “An Effcient In-
terest-Group Based Search Mechanism in Unstructured
Peer-to-Peer Networks,” Proceddings of the International
Conference on Computer Networks and Mobile Comput-
ing, Shanghai, 20-23 October 2003, pp. 247-252.
[5] H. C. Hsiao and C. T. King, “Bristle: A Mobile Struc-
tured Peer-to-Peer Architecture,” Proceeding of Interna-
tional Parallel and Distributed Processing Symposium,
Nice, 22-26 April 2003, pp. 33-40.
[6] H.-T. Hu, B. Thai and A. Seneviratne, “Supporting Mo-
bile Devices in Gnutella File Sharing Network with Mo-
bile Agents,” Proceedings of the 8th IEEE International
Symposium on Computers and Communications, Kemer-
Antalya, 28 June-1 July 2004, pp. 25-30.
[7] S. Liu, W. Jiang and J. Li, “Architecture and Performance
Evaluation for P2P Application in 3G Mobile Cellular
Systems,” Proceedings of International Conference on
Wireless Communications, Networking and Mobile Com-
puting, Shanghai, 21-25 September 2007, pp. 914-917.
doi:10.1109/WICOM.2007.235
[8] 3GPP TS 23.002 v5.12.0 3rd Generation Partnership
Project; Technical Speci_cation Group Services and Sys-
tems Aspects Network architecture (Release 5). Septem-
ber 2003.
[9] G. Camarillo and M. A. Garcia Martin, “The 3G IP Mul-
timedia Subsystem,” John Wiley & Sons, Hoboken, ISBN
0470 871563, 2004.
[10] D. Wagner, “GSM Cloning,” June 2010
http://www.isaac.cs.berkeley.edu/isaac/gsm.html
[11] A. Biryukov, A. Shamir and D. Wagner, “Real Time Cryp-
tanalysis of A5/1 on a PC,” Proceedings of the 7th Inter-
national Workshop on Fast Software Encryption, 2000,
pp. 1-18.
[12] G. Koien, “An Introduction to Access Security in
UMTS,” IEEE Wireless Communications, Vol. 11, No. 1,
2004, pp. 8-18. doi:10.1109/MWC.2004.1269712
[13] V. Gupta, S. Gupta, S. Chang and D. Stebila, “Perform-
ance Analysis of Elliptic Curve Cryptography for SSL,”
Proceedings of the ACM Workshop on Wireless Security,
Atlanta, September 2002, pp. 87-94.
[14] D. Boneh and M. Franklin, “Identity-Based Encryption
From The Weil Pairing,” Proceedings of Cryptology, Lec-
ture Notes in Computer Science, Vol. 2139, Springer,
2001, pp. 213-219.
[15] S. Zhao, A. Aggarwal and S. Liu, “Building Secure User-
to-User Messaging in Mobile Telecommunication Net-
works,” Proceedings of Wireless Telecomunications Sym-
posium, Pomona, 24-26 April 2008, pp. 151-157.
doi:10.1109/WTS.2008.4547559
[16] K. Hoeper and G. Gong “Preventing or Utilising Key
Escrow in Identity-Based Schemes Employed in Mobile
Ad Hoc Networks,” International Journal of Security and
Networks, Vol. 2, No. ¾, 2007, pp. 239-250.