Mobile Distributed Authentication Protocol: II. Authentication in GSM Networks
Mobile Distributed Authentication Protocol: II. Authentication in GSM Networks
Mobile Distributed Authentication Protocol: II. Authentication in GSM Networks
I. Introduction
A. Drawbacks
III. G-SNAP
AuC
k
S2
Sk
AuC
2
Core
Network
(CN)
RAN
S1
AC
ms
Local
AuC
AuC
1
User
Local AuC
Authentication centres
Access Request
Shares Request ?
Share Request 1
AuC 1
AuC 2
Share Request k
AuC k
Share reply 1
Share reply k
1.
2.
3.
4.
5.
6.
Top
Shares message
AC Reconstruction
Access-reply
Network access
Fig. 2: G-SNAP signalling
6
3
2
MSC
TABLE I
Quorum number
Q1
Q2
Q3
Q4
Mobile
Mobile
Core Network
Mobile
Mobile
Mobile
Top
Quorum members
1, 2, 5
2, 3, 4
0, 4, 5
0, 1, 3
User allocation
1
2,3,4
5,6
0
A. Assumptions
Mobile
Quorum system
Mobile
Q2
Mobile
Q1
Mobile
0
Q3
Q4
B. Simulation results
To improve the statistical significance of the results, the
simulation was run for a large number of cycles. It is
assumed that each node represents an authentication centre
The user will send a packet that contains the required
information for the user to identify themselves to the other
nodes. In this case we assume that the authentication packet
is large enough to handle all the needed information. The
packet size is 256 bytes. Using G-SNAP, the user will
simultaneously send k packets to the authentication nodes.
The selection of the routing path is based on the built-in
routing protocol provided by ns-2. Since ping agents are
used, packets will go back directly after reaching the
destination towards the source. In G-SNAP the user needs to
average delay ms
90
80
G-SNAP/Q2
CAS/2
70
60
50
40
30
20
10
0
0
Users
Top
60
50
40
30
20
10
0
G-SNAP/local quorums
CAS/2
80
70
Users
90
80
average dealy ms
90
average delay ms
wait for the last packet to arrive such that the waiting time
does not exceed the predefined time limit. In order to
compare G-SNAP with centralised approaches, three
scenarios have been considered.
In the first one, centralised users contact a centralised
authentication server (CAS) in node 2. G-SNAP users will
contact one of the quorums. Q2 is selected. This means that
each user should contact the quorums members which are
node 2, 3 and node 4. Fig. 5 shows the average delay
experienced by users in both cases. It can be observed from
that figure the following:
70
60
50
40
30
20
10
0
0
Users
V. Discussion
In this discussion a comparison between G-SNAP and
centralised approaches is investigated. In order to determine
the signalling overhead caused by G-SNAP, the number of
messages sent to the authentication nodes is required. In GSNAP the user must contact k nodes to collect the required
shares. Hence, signalling overhead is increased by k
compared to centralised approaches. Contacting the local
AuC is not considered in the signalling overhead since it is
not significant. In centralised approaches all users contact a
single node increasing the traffic destined for that node. At
the same time the load on the single entity will be high. In
contrast G-SNAP distributes the load to the quorums. If there
are l quorums then the load to each quorum is equal the total
load divided by l. G-SNAP can achieve load balancing as
well. G-SNAP overcomes the drawbacks of centralised
approaches as mentioned in section II. G-SNAP increases
VI. Conclusion
There is a need to have multiparty authentication protocols
such that more than one server controls the authentication
process. This results in increased security and availability,
and also distributed trust, which overcomes the centralised
authentication approaches deficiencies. The importance of
this will increase with 3G networks, which are more
distributed in nature and have more complex business
arrangements. G-SNAP is a new protocol, which achieves
these objectives. The performance evaluation of G-SNAP
compared to centralised approaches is discussed in this paper.
Although G-SNAP does introduce more signalling, the delay
impact on users is marginal. G-SNAP can be more efficient
than the centralised approach if it is used to provide more
localised distributed authentication.
References
[1] A. F. Al Shahri, D. G. Smith and J. M. Irvine, Implementation
of quorum systems to increase network security, PGNET 2002,
June 2002, Liverpol University, UK.
[2] A. F. Al Shahri, D. G. Smith and J. M. Irvine, Implementation
of Secret Sharing to Increase Network Security and Reliability,
ESPRC
Postgraduate Research in Electronics and Photonics
(PREP), April 2002, Nottingham University, UK.
[3] G. J. Simmons. An introduction to shared secret and/or shared
control schemes and their application, In Contemporary
Cryptology, The Science of Information Integrity, pages 441-497.
IEEE Press, 1992.
[4] Malkhi, Quorum systems, chapter in the encyclopaedia of
distributed computing march 1999
[5] J. Dunlop, D. Girma and J. Irvine, Digital Mobile
Ccommunications and the TETRA System, John Wiley & Sons.
1999.
[6] A. F. Al Shahri, D. G. Smith and J. M. Irvine, A Secure
network access protocol (SNAP), unpublished.
[7] L. Gong. Increasing availability and security of an
authentication service, IEEE J. Selected Areas Comm., 11(5):657662, 1993.
[8] M. Naor and A. Wool. Access control and Signatures via
quorum secret sharing, In Proc. 3rd ACM Conf. Comp. And comm.
Security, pages 157-168, New Delhi, India, Mar. 1996.
[9] Network Simulator version 2 (NS-2) UC Berkeley, USA, http://
www.isi.edu/nsnam/na/ns-documentation.
Top