MANAGING SECURITY OF GRID ARCHITECTURE WITH A GRID
SECURITY OPERATION CENTER
Julien Bourgeois and Raheel Hasan
Computer Science Laboratory (LIFC), University of Franche-Comte (UFC)
1 Cours Leprince-Ringuet, 25201 Montbeliard, France
Keywords:
Grid security, Grid security operation center, Specificities of grid networks.
Abstract:
Due to the nature of grid computing networks, security pitfalls are plethora and adversaries are sneaking to
launch attacks. Keeping this scope in mind, we will discuss our proposed solution for securing grid computing
networks that we have called gSOC (Grid Security Operation Center). The main advantage of gSOC is that
it can give a global view of security of the entire grid infrastructure. The main difficulty is to deal with
the specificities of grid infrastructure, that are: multi-sites networks, multi-administrative domains, dynamic
collaboration between nodes and sites, high number of nodes to manage, no clear view of the foreign networks
and exchange of security information among different domains.
1 INTRODUCTION
The industrial and scientific communities are always
looking for more computational power to achieve this
goal researchers have studied multiple solutions for
interconnecting organizations in order to share com-
putational resources, which has given birth to grid
networks.
There exist multiple definitions of a grid network
but in the rest of this article we will only consider the
following one. We define a grid network as a net-
work of multiple administrative domains where each
administrative domain consists of multiple network
sites which put a service (storage space, computa-
tional resource, etc.) at disposal of users. An ad-
ministrative domain is an entity which follows ho-
mogeneous security policy through all its sites. Ac-
cording to the above definition a grid network may
be composed of a high number of computational de-
vices across the globe. A grid system is interfaced
with its users through a middleware. There are dif-
ferent grid middlewares such as gLite (gli), Unicore
(Uni), ARC (kno) or Unibus (Dawid Kurzyniec and
Sunderam., 2007). The largest grid service in Europe
is EGEE (enabling grid for e-sciences) (ege) which is
based on gLite software. EGEE connects many local
grids from different countries for example, Germany
DE-Grid, Netherlands BIG-Grid or Belgium BE-grid.
It process around 300,000 jobs per day from scientific
domains ranging from biomedicine to fusion science.
Grid networks are therefore composed of different ad-
ministrative domains that are often located in differ-
ent countries. They have different security policies
and they must respect possibly different laws in each
country. This heterogeneity arise issues in the secu-
rity management of grid networks which are not tak-
ing into account by existing middleware or security
management softwares. Therefore, these properties
that makes grid networks special in regard to multi-
purpose networks which have to be taken into account
by security environment, like:
1. Grid network structure with different administra-
tive domains, each of them composed of multi-
sites networks.
2. High number of nodes which collaborate with
each other dynamically.
3. View of security events of foreign networks un-
available.
To cope with these issues, we have designed
gSOC (Grid Security Operation Center) whose aim
is to give a global view of security of a grid net-
work by taking into account its specificities. Section 2
presents our proposed architecture named gSOC. Sec-
tion 3 concludes this article and announces some fu-
ture works.
403
Bourgeois J. and Hassan R. (2009).
MANAGING SECURITY OF GRID ARCHITECTURE WITH A GRID SECURITY OPERATION CENTER.
In Proceedings of the International Conference on Security and Cr yptography, pages 403-408
DOI: 10.5220/0002188304030408
Copyright
c
SciTePress
2 PROPOSED ARCHITECTURE:
THE gSOC ENVIRONMENT
2.1 Introduction
gSOC is a security operation center dedicated to
grid computing networks. In grid networks, there is
no mechanism for sharing security information with
multiple administrative domains which consists of
multiple sites. Therefore in grid networks handling
of cross domain attacks and their expansion is very
difficult. Attacks launched by grid users with mas-
sive grid computational power and memory could be
more powerful than normal attacks. By considering
the above parameters, the need of gSOC is manda-
tory. gSoc also provides a better mechanism for trust
management within multiple administrative domains
connected to the grid.
At present gSOC design is under the development
phase, we started our work by keeping the target to
provide better trust management and sharing of secu-
rity information between multiple administrative do-
mains.
2.2 The Components of the gSOC
Environment
gSOC is composed of five components partly based
on the CIDF specifications (Staniford-Chen et al.,
1998), from bottom to top: data collectors (CBoxes),
remote data collectors (R-CBoxes), Local Analyzers
(LAs) and a Global Analyzer (GA) and Secure Virtual
Organization (SVOBox).
2.2.1 Data Collection Box
A CBox collects data from sensors located on the
same segment of a network. A sensor can be a host, a
server, a firewall, an IDS or any system that generates
logs. The advantage of our log collection approach
is that no software has to be installed on the sensors.
Moreover, our system is compatible with a wide range
of hardware and software. A CBox formats logs and
sends them to a local intrusion database (lidb). In each
site we have one or several CBoxes and one of them
acts as a Master CBox (M-CBox). The M-CBox is
responsible for the management of all the CBoxes lo-
cated on the same site. It polls regularly the other
CBoxes and when a CBox is down, the M-CBox will
collect data on the segment of the failed CBox. Each
Master CBox also has a backup which polls it regu-
larly and will become Master if it needs to be.
2.2.2 Remote Data Collector
An R-CBox is a special CBox which collects data
coming from some critical sensors and from sensors
hosting security tools in any site. Afterward, data is
forwarded to the local intrusion database of another
site and is analyzed to give in real time the approxi-
mate security level of the concerned site. This helps to
anticipate a reaction when a critical intrusion occurs
or to investigate and troubleshoot a site that could be
compromised, even if a hacker erases the logs on the
compromised sensors (including the security tools).
2.2.3 The Local Analyzer
A Local Analyzer (LA) is responsible for intrusion
detection at any site of a network. It analyzes for-
matted logs located in a local intrusion database (lidb)
and generates alerts. Afterward, it correlates the alerts
to find more complex intrusions (intrusions composed
of several events, distributed intrusions, intrusions di-
rected to many sensors, etc.). The LA also compress
alerts by merging similar ones. All the alerts gener-
ated by an LA are sent to the global intrusion database
(gidb). The gidb can have a mirror of itself for high
availability purpose.
2.2.4 The Global Analyzer
The Global Analyzer (GA) is a chosen LA respon-
sible for the global intrusion detection in a network.
It analyzes alerts from the Global Intrusion database
(gidb), correlates and merges them if possible to gen-
erate optimized outputs. It is also able to detect more
sophisticated intrusions that are directed to several
sites. The GA regularly polls the other LAs and when
one of them is down, the GA detects the occurring in-
trusion into the concerned site. Another LA acts as
the backup of the GA and polls it regularly. When the
GA is down, the backup becomes the GA and another
backup is elected.
The gSOC architecture is designed bearing in
mind that the data flow processed in the different sites
of the network is not always homogeneous. Indeed, in
some sites, a large amount of data is processed and in
this kind of situation, several CBoxes are needed for
data gathering.
Even though a single CBox has to be installed on each
segment, it is not excluded installing several CBoxes
on the same segment when the sensors located in this
part of the network are operating under high work-
loads. In quieter sites, only one CBox can be used to
collect data coming from all the sensors.
The gSOC also implements the different types of
boxes defined for network intrusion detection systems
SECRYPT 2009 - International Conference on Security and Cryptography
404
in (Northcutt and Novak, 2002). However, beside the
pure technical aspects involved in such implementa-
tions, it is necessary to consider the supervision of an
IT infrastructure as a full operational project.
Figure 1 displays the flow of security events
among multiple sites which are connected under one
administrative domain (AD). These security informa-
tion contains all the suspected threats which incurred
in an AD. The mechanism of security information
flow is similar in all the other ADs.
2.2.5 Secure Virtual Organization Box
Secure Virtual Organization Box (SVOBox) job is to
collect all the correlated security alerts (SA) gener-
ated in different administrative domains (AD) which
consists of multiple sites. SVOBox assigns certain
security level (SL) value using simple metric for real-
time security level evaluation which represents three
values indicated in colors (red, orange and green).
Green indicates no threat occurring in the network,
orange indicates that threats are occurring but not
critical at this time and red indicates intrusions are
in progress which can lead to critical security prob-
lems(Ganame, 2008). In addition to this method we
have added another step which will be performed at
Global Intrusion Data Base (gidb) of every Adminis-
trative Domain (AD). According to this step the gidb
deployed in an AD will forward this security informa-
tion to SVOBox where the SVOBox will assign secu-
rity level value to each AD (see figure 2) as follows,
If all the sites having green status in an AD that
will be placed in security level 1 which is most
secure than SL 2 and SL 3.
If any one site in an AD having status indicating
orange that AD will be placed in security level 2
which is more secure than SL 3.
If any one site in an AD having status indicat-
ing red that AD will be placed in security level
3 which is the least secure level.
After the security level (SL) assignment to all the
ADs which are now the part of this grid. Before start
sharing the resources every AD of this grid would like
to have a global view of the security, for global view
they must need to share the security information with
other ADs in the grid. To share the security alerts
among ADs two methods are employed.
First method is to send security alerts to the ad-
ministrative domains (ADs) of the different security
levels (1,2 and 3). In each AD there exists a gidb
which contains three different kinds of security alerts
(SA) (see figure 3). From top to bottom, the first sec-
tion collects security alerts from the lesser secure lev-
els. The middle section is for holding security alerts
from its local sites which are residing in the same SL.
The third section collects security alerts from the most
secure levels. Finally, the last section is used to store
alerts which an AD do not want to share with others
(private alerts). For example, in figure 3, gidb of AD
3 and AD 4 which are in SL 2 can share their security
alerts with AD 1, AD 2 at SL 1, and AD 5, AD 6 at
SL 3, whereas AD 3 and AD 4 can share their security
alerts using the middle section which is reserved for
their local sites security alert information as they lies
in the same security level that is SL 2. When security
alerts are being shared among ADs the two following
fields will be appended with each security alert mes-
sage:
1. Direction of security alert field which directs the
security alert to move in the upper or lower direc-
tion depending upon the security level value.
2. Time to Live (TTL) field which is used to define
the authorized propagation of the security alert.
For example, if AD 1 which is in SL 1, wants to
share its security alerts only with ADs of SL 2,
it will put an upper direction with a TTL set to 1.
The TTL will be decreased when the security alert
arrives at SL2 so it will not be propagated till SL
3.
Similarly AD 1 and AD 2 can share security alerts
with AD 3, AD 4 and between themselves also. AD
5 and AD 6 can share their security alerts with AD
3, AD 4 respectively. This sharing of security alerts
will present an updated global view of security of the
entire grid. Now each AD can decide according to
its organization policies that sharing of resources can
be done with the most or least secured AD. This is
the objective which we will going to be achieved by
using gSOC, as it gives the global view of the entire
grid without sharing of all the data (see figure 3).
The second method, is to allow a remote network
administrator to query gidb of other administrative
domains. The gidb of an AD will connect with gidb
of other AD and the remote network administrators
of both the ADs will be allowed to query each others
gidb to see the security alerts. After the inspection
of security alerts by each network administrator they
can decide for the sharing of resources. This solution
is also possible but the drawback is that it requires a
constant monitoring and constant execution of queries
on gidb which is not an efficient solution in case of
major attacks like DoS/DDoS attack.
These two kinds of sharing of security information
present some advantages and drawbacks, but they can
be used alternatively depending on the type of net-
work and number of generated security alerts.
MANAGING SECURITY OF GRID ARCHITECTURE WITH A GRID SECURITY OPERATION CENTER
405
M-Cbox
Cbox
Cbox
lidb
R-CBox
SE
SE
M-Cbox
R-Cbox
Cbox
lidb
CBox
M-Cbox
Cbox
Cbox
R-CBox
LA (Global Analyzer)
LA
(B-GA)
LA
Site 1
Site 2
SE:Security Events
F.log: Formatted log
m/c alertes: Merged/correlated
alerts
M-Cbox: Master Cbox
R-Cbox: Remote Cbox
B-GA: Backup of Global
Analyzer
SE
SE
SE
gidb
gidb
SE
SE
m
/
c alerts
Analyzing
Analyzing
SE
SE
Analyzing
F.log
SE
F.log
m/c
alerts
m/c
alerts
Analyzing
Internet
Secure Virtual
Organization Box
(SVOBox)
Internal view of an
Administrative Domain
Site 3
lidb
Figure 1: Overview of the internal structure of an Administrative Domain (AD).
Secure Virtual
Organization Box
(SVO-Box)
AD 1
AD 2
AD 3
AD 4
AD 5
AD 6
SL 1
SL 1
SL 2
SL 2
SL 3 SL 3
AD=Administrative Domain
SL=Security Level
SA=Security Alerts
SA
SA
SA
SA
SA
SA
Figure 2: gSOC overview.
SECRYPT 2009 - International Conference on Security and Cryptography
406
Alerts of AD 6 which is in least secure SL 3
Administrative Domain 2 in Security Level 1
Administrative Domain 4 in Security Level 2
SA of all the sites
connected under AD 6
passing SA to AD4
gidb of AD 4
gidb of AD 2
For Future Use
For Future Use
LA of all the sites connected under AD 4
LA of all the sites connected under AD 2
Alerts from AD 2 which is in most secure SL 1
Alerts of AD 4 which is in more secure SL 2
SA of AD 4 which is in
more secure SL 2
passing its SA to AD6
Security Alert
information
Alerts of AD 5 which is in least secure SL 3
Administrative Domain 1 in Security Level 1
Administrative Domain 3 in Security Level 2
Security Alerts of
all the sites
connected under AD 5
passing AS to AD3
gidb of AD 3
gidb of AD 1
For Future Use
For Future Use
LA of all the sites connected under AD 3
LA of all the sites connected under AD 1
Alerts from AD 1 which is in most secure SL 1
Alerts of AD 3 which is in more secure SL 2
SA of AD 3 which is
in more secure SL 2
passing its SA to AD3
Security Alert
information
Security Alert information
Security Alert information
LA=Local Analyzer
gidb=Global Intrusion Data Base
AD=Administrative Domain
SA=Security Alert
Figure 3: Sharing of security alerts using Global Intrusion Data Base (gidb).
2.3 Protecting the Communications
between the gSOC Components
One of the key points here is to make sure that no ille-
gitimate computer will act as an LA, a CBox or an R-
CBox in order to get privileged access to the system.
To ensure the security of our system, any LA or any
R-CBox that needs to exchange information with an-
other LA has to use a certificate issued by the network
administrator of that AD to prove its identity. More-
over, all the communications between the LAs (also
including the GA) and the communications between
the R-CBoxes and the LAs will have to pass through
an encrypted tunnel, available via the SSL protocol.
3 CONCLUSIONS
The purpose of gSOC is to give a better global view
of security without sharing all the data in order to
achieve maximum confidentiality. The idea behind
gSOC is to give the better tradeoff between confi-
dentiality of security data and having the best possi-
ble global view of the network which means sharing
all the data. gSOC gives correlated overview by dis-
carding similar information to pass every time, gSOC
matches sequence pattern, time pattern, system expo-
sure performance, critical analysis and in final step
security policy matching. gSOC does data analysis
by correlation, structural analysis and intrusion path
analysis. In future experiments we have to check that
what is the confidentiality level that can be achiev-
able, the TTL value which needs more study. As LA
can be elected to become the gidb in case of failure,
MANAGING SECURITY OF GRID ARCHITECTURE WITH A GRID SECURITY OPERATION CENTER
407
in each LA there should therefore have four sections
reserved for security alerts.
REFERENCES
http://glite.web.cern.ch/glite/. Lightweight Middleware for
Grid Computing.
http://www.eu-egee.org/. Enabling Grids for E-science
(EGEE).
http://www.knowarc.eu/. Grid-enabled Know-how Sharing
Technology Based on ARC Services and Open Stan-
dards (KnowARC).
UNICORE (Uniform Interface to Computing Resources).
http://www.unicore.eu/UNICORE.
Dawid Kurzyniec, Magdalena Slawinska, J. S. and Sun-
deram., V. (2007). Unibus: a contrarian approach to
grid computing. J. Supercomput., 42(1):125–144.
Ganame, A.K. Bourgeois, J. (2008). Defining a simple met-
ric for real-time security level evaluation of multi-sites
networks. Parallel and Distributed Processing, 2008.
IPDPS 2008. IEEE International Symposium, 14-18:1
– 8.
Northcutt, S. and Novak, J. (2002). Network Intrusion De-
tection. ISBN: 0-73571-265-4. New Riders, third edi-
tion edition. September.
Staniford-Chen, S., Tung, B., and Schnackenberg, D.
(1998). The common intrusion detection framework
(cidf). In Information Survivability Workshop, Or-
lando.
SECRYPT 2009 - International Conference on Security and Cryptography
408