COLLABORATIVE DEFENCE FOR DISTRIBUTED ATTACKS (CASE STUDY OF PALESTINIAN INFORMATION SYSTEMS)

International Journal of Network Security & Its Applications (IJNSA) Vol.6, No.6, November 2014
DOI : 10.5121/ijnsa.2014.6603 31
COLLABORATIVE DEFENCE FOR DISTRIBUTED
ATTACKS (CASE STUDY OF PALESTINIAN
INFORMATION SYSTEMS)
Hani Qusa
Information Technology Department
University College of Applied Science - Palestine
ABSTRACT
In this paper, we develop a comprehensive approach for protecting national Palestinian information
systems. We do not restrict our attention to protecting each individual organization, but rather focus on the
entire ecosystem as a whole. Therefore, the developed system will be opened for participation for all
Palestinian governmental and non-governmental organizations who are interested in improving their
security and protection against current threats and security attacks targeting Palestinian information
systems. The results will help in raising the awareness about information security for participating
organization.
KEYWORDS
collaborative environments, MapReduce, critical infrastructures, Intrusion detection system, DoS Attack.
1. INTRODUCTION
The new technological trends, the need to provide a competitive services, and the overlap in the
businesses in several fields have brought about a shift from the internal organization’s
infrastructure to engagement in a truly global ecosystem characterized by many cross domain
interactions and heterogeneous systems and data. Therefore, the dependency of several
governmental and non-governmental organizations on cyber-infrastructure of critical
infrastructures encounters an increasing demands and undergoes a profound technological and
usage changes.
However, the sophistication of cyber attacks has increased over time and tools and techniques of
attacks are easy and widely spread. This comes to fact that the technical knowledge required to
exploit existing vulnerabilities is decreasing. For example, organizations that suffer from
Distributed Denial of Service (DDoS) attack, know that they have been attacked, but they cannot
easily distinguish the group of IP addresses that commit this attack alone, which create a big
challenge for organizations to defend themselves. The perpetrators of these attacks, whether
motivated by the prospect of financial gain or because they see such an attacks as means of
garnering publicity or otherwise pursuing a political cause, benefit from sharing technology and
other information among themselves. Therefore, protecting these infrastructures in the face of
faults and malicious attacks is crucial to ensure stability, availability, and continuity of the key
electronic services and individual businesses worldwide.

32
Collaboration on the data level by aggregating data distributed among different organizations is
considered as an important analysis in the context of the security for these organizations.
Collaboration for security implies combining and correlating large volume of data from
distributed organizations in order to have a more comprehensive view of malicious activities that
may occur separately and that would have gone undetected if considered in isolation.
In this research, we embark on a research agenda aiming to develop a comprehensive approach
for protecting national information systems of Palestinian organizations. We do not restrict our
attention to protecting each individual organizations, but rather focus on the entire ecosystem as a
whole. Therefore, the developed system will be opened for participation for all Palestinian
governmental and non-governmental organizations who are interested in improving their security
and protection against current threats and security attacks targeting Palestinian information
systems. The results will help in raising the awareness about information security for participating
organization.
Our specific objective in the research timeframe will be to devise a scalable distributed
monitoring system that will provide the relevant IT components of participating organizations
with early notifications about faults and other potentially malicious activity originating at remote
sites (possibly belonging to other critical infrastructures) thus enabling those components to
trigger the necessary protective mechanisms in a timely fashion.
2. RELATED WORKS
Several systems have been developed in order to protect organizations in collaborative way such
as [1][2][6][7][8][9][10][11]. However, customization of these solutions in order to provide the
necessary protection for the Palestinian information system is not feasible. This returns to the fact
that these information systems are not so advanced to have such protection with more capabilities
than required. In other words, one main attack that threatens the Palestinian information systems
is the Denial of Services attack. This attack has little impact on financial transaction on the
Palestinian information systems. However, the delay impact caused by this attack on the
Palestinian community desires to provide the required protection. Thus, customization of an
existing solutions or using commercial one will be considered as an overhead cost for these
organizations that cannot be afforded. While designing a special system that fits to the existing
infrastructure will have a good impact with minimum cost.
3. SYSTEM ARCHITECTURE
3.1 System properties and assumptions
In fact, the security tools used in Palestinian organizations are very simple and depends mainly on
commercial firewalls. This exposes them to several types of attacks that cannot be controlled by
only firewalls. Therefore, the development of Collaborative Intrusion Detection System (CIDS)
for building collaborative defence for Palestinian information system is very important issue as
we as it has several challenges.
Furthermore, the old infrastructure of most Palestinian organizations makes difficulty in using
existing collaborative intrusion detection systems which requires at least a good services to run on
it. We assume that using a Distributed File System (DFS) that can be installed on several
individual machines can help in processing the required data.

33
Finally, the geographical distribution nature of the Palestinian Territories imposes a real challenge
which desires a need to adopt distributed collaborative defence system. Another challenge is the
differences in the data model used in these organizations. This requires a special normalization
level that helps in unifying the form without losing the benefit of the data.
3.2 Threat Model
We consider threat model that takes into account one major type of attackers who try to evade to
systems by running port scanning against Palestinian organizations running information systems.
The objective of these attackers is to suspend the electronic services provided by targeted
organizations. This attack is well-known as the Denial of Service (DoS) attack. More
sophisticated attacks of DoS is the distributed one, where the attacker depends of several
controlled machines to execute the attack concurrently from these controlled devices, which in
know as Distributed DoS (DDos).
In order to carry out such an attack, these sophisticated attackers often run port scanning against
single organization at low rate. The attackers use this strategy against several organizations in the
same time. By adopting this strategy, it’s difficult for individual intrusion detection systems of
these organizations to discover such an attacks.
Fig. 1: System Architecture
3.3 System Model
We propose a system model that employs multiple individual detectors installed in different
organizations. In each organization a local intrusion detection system IDS is installed in order to
monitor the traffic of that organization. The proposed system consists of n organizations that use
these individual IDSs in order to form Collaborative Intrusion Detection System CIDS. The
system will exchange the suspected behaviours among the individual IDSs in order to
cooperatively discover malicious behaviours and detecting the sources of these malicious
activities. Thus, the system doesn’t rely on any external party nor has a centralized node for

34
processing the data. This form a hierarchical architecture of collaborative intrusion detection
system. Fig. 1 shows the architecture of the system.
Practically, the processing of the data in the system is carried out by set of components distributed
over the participating organizations. Each organization contains two components; 1) the local
Intrusion Detection System (LIDS) and, 2) The Circulation Unit (CU). The phases work in
pipeline and are described next in more details.
The first component (LIDS) monitors the network traffic of the organization in order to produce a
blacklist of suspicious behaviours. The suspicious behaviour means an IP address that carries out
a port scanning activities against that organization.
The second component, CU , are connected together using a routing table. The CUs are
responsible for exchanging the suspicious lists among organizations. These suspicious list are
generated by the local IDSs. Furthermore, the CUs are responsible for correlation process over the
collected data.
4. SYSTEM IMPLEMENTATION
The processing of the data is executed in two phases according to the system architecture. The
implementation of the data includes deploying two sub-systems; one for the local processing that
uses the LIDS component. And the Collaborative Processing which use the Circulation Units
(CU) to process the resulted data from the individual LIDS cooperatively. The processing in the
two phases are described next.
4.1 Local Processing Phase
During this phase, each participating organization processes the local traffic network in order to
check for anomalies. This phase is implemented by the private processing unit that carried out all
the necessary operations. The unit is a private cloud constructed for this purpose that uses a
Distributed File System (DFS). Specifically, the system adopt Hadoop DFS (HDSF) [3] in order
to store and to process the network traffic data. The system uses a high level query language in
order to define the processing logic. The language is compiled into a series of MapReduce jobs
[12].
We adopt SQL-like query processing called HIVE [4], that specifies the data patterns to be
discovered on the set of large input data. A query processing engine inside each private
processing unit retrieves the data in the storage elements and aggregates them according to one or
more SQL-like queries. The main output of the query is a filtered and aggregated according to the
suspicious IP addresses. All participating organization generate a general format of data that in
order to be used in the next phase. The general format of the output is a list of IP addresses and
the count of the appearance of this IP addresses in the network traffic of each organization, i.e.
<IP, IP_Count>.
4.2 Circulation Phase
This module is in charge of collecting the previously generated lists from all participating
organizations. The data collection occurs periodically, i.e., every fix time window. The circulation
units agrees on disseminating the list of IP addresses according to a specific threshold. For
example, if the threshold is set to 100, then all the aggregated IP records with IP_Count ≥ 100
will be disseminated. This threshold is denoted as local threshold (LThreshold).

35
The start and the end of each period is determined according a special threshold. Practically, let Li
be the list produced by Participant Pi. and let all participating organizations be connected as a
logical ring running over the Internet. Then, participant Pi starts the collection of lists Lj from all
the other participants by sending a start-collection message in the form of an empty token Ti. The
token is circulated along the ring until all the participants have appended their lists to it. After
that, Pi removes the token. This protocol is an enhancement of a previous protocol that was
introduced in [1].
In order to satisfy a level of privacy that avoids link-ability between the data and the owner of this
data, the first participant that puts the list inside the token is determined at random. Specifically,
when participant Pj (where i != j) receives Ti for the first time then, if the token is empty it adds
its list Lj to Ti with a probability pstart; otherwise, it adds the list for sure. Pi removes the token
from the ring when the token passes through it for the second time nonempty and unmodified.
This ensures that all participants have added to the token their lists. All participants executes this
protocol in the same window of time. The full algorithm is listed in Algorith1
Finally, the circulation units in all participating organizations starts the correlation process over
collected data. Another agreed threshold called collaborative threshold (CThreshold) is used to decide
the black list of IP addresses that have carried out malicious behaviour against participating
organizations.
Algorithm1: Circulation Protocol
Input: = probability of start sending
Output: Each participant will compute:
=
Uses:
= = i
A[n] = ;
upon event <init> do
foreach j 1 → n do
A[j] = 0;
= Φ
.send();
upon event <TokenReceived> do
= .id;
if == then
if ( == ) then
= ;
else
= = ;
else
if ( == Φ ) then
= random();
if ( ≥ ) then
= ;
else
= ;
else
if (A[ ] == 0) then
= ;
A[ ] = 1;
else
= ;
.send();
5. PERFORMANCE EVALUATION
This section provides an evaluation of a prototype implementation of the proposed system. The
proposed architecture has been implemented in Java and run on a cluster of four 2.3 GHz dual
processor physical machines, equipped with 4GB of RAM each one. The physical machines are
connected to a LAN of 1Gbit, running ubuntu linux.
The network traffics of the participating organizations were simulated using a log generator that
produces a traces for the network traffic in the same format as generated by web server. For

36
example, the generated log contains TCP connection in a trace which is represented as a tuple
<source, destination> with an associated timestamp with granularity up to seconds, and state
information (Dropped or Accepted). The generated logs are injected to the LIDS of the
participating organizations.
Fig. 2: (a) LIDS processing time, (b) Circulation throughput
Inside the processing unit, the data is processed by a collection of Hadoop's MapReduce jobs
inside each participant. The processing logic, which represents our data pattern, is expressed in
HIVE-QL statements. Specifically, the language supports SQL-like query constructs that can be
combined into flow and specify the data patterns. The query engine retrieves the data in the
storage elements and aggregates them according to one or more SQL-like queries. The HIVE-QL
statements are compiled into series of MapReduce jobs.
We evaluate the system throughput as a function of the number of records and number of
participants that the system can process and serve in a specific period of time with and without
applying the proposed privacy-preserving mechanism.
Figure 2(a) shows the throughput of this phase. The time for processing the data is an average of
≈16 sec in all cases. This is because HIVE is used mainly for processing large dataset and it
requires a high start-up time, while in our experiments, the dataset comprises 1000 records in case
of 8 participants with 1000 record per each one which is not big enough in terms of HIVE. While
Figure 2(b) shows the throughput of the circulation phase. Depending on the local network for
experimentation indicates mostly linear and very slow increasing in the delay regarding the size
of the data.
6. CONCLUDING REMARKS
In this paper we described a prototype for collaborative defense against coordinated attacks. The
system is proposed to protect national information systems of the Palestinian organizations. The
proposed system is able to collaboratively correlating raw data coming from web logs of
Palestinian organizations in order to detect port scanning activities which are used in preliminary
phase of conducting DDoS attacks.
We are currently conducting an experimental evaluation of our system implementation. The
initial results show that ability of the system to detect suspicious port scanning activities. We are
planning to carry out an extensive experimental evaluation to the accuracy of the detection and to
consider other types of attacks that threaten Palestinian information system. The future work will

37
take into consideration other issues with a deep investigation such as the privacy of the data and
the scalability of the system. Furthermore we aim to investigate more in the scalability of the
system by considering the number of the data manipulated and the number of participants in the
systems.
ACKNOWLEDGEMENTS
This work was supported and funded by the scientific research council in the ministry of
education & higher education in Palestinian national authority. The author would like to thank
reviewers for their thoughtful comments to improve this article.
REFERENCES
[1] Qusa, Hani, and Shadi Abudalfa. "SECURE COLLABORATIVE PROCESSING ARCHITECTURE
FOR MITB ATTACK DETECTION." International Journal of Network Security & Its Applications
5.5 (2013).
[2] Roberto Baldoni, Giuseppe Di Luna, and Leonardo Querzoni. Collaborative Detection of
Coordinated Port Scans. ICDCN proceeding , pp 102-117. 2013.
[3] Hadoop. http://hadoop.apache.org/, 2011.
[4] Hive. http://wiki.apache.org/hadoop/Hive, 2011.
[5] Jaql. http://www.jaql.org/, 2011.
[6] D. Bogdanov, S. Laur, and J. Willemson. Sharemind: A Framework for Fast Privacy-Preserving
Computations. In Proc. of the 13th European Symposium on Research in Computer Security:
Computer Security, ESORICS '08, pages 192-206, Berlin, Heidelberg, 2008. Springer-Verlag.
[7] D. Je_rey and S. Ghemawat. MapReduce: simplified data processing on large clusters. Commun.
ACM, 51(1):107-113, 2008.
[8] I. Roy, S. T. V. Setty, A. Kilzer, V. Shmatikov, and E. Witchel. Airavat: security and privacy for
MapReduce. In Proc. of the 7th USENIX conference on Networked systems design and
implementation, NSDI'10, Berkeley, CA, USA, 2010. USENIX Association.
[9] Y. Xie, V. Sekar, M. K. Reiter, and H. Zhang. Forensic analysis for epidemic attacks in federated
networks. In ICNP, pages 143-153, 2006.
[10] G. Zhang and M. Parashar. Cooperative detection and protection against network attacks using
decentralized information sharing . Cluster Computing, 13(1):67-86, 2010.
[11] Maher Salem and Ulrich Buehler. Mining Techniques in Network Security to Enhance Intrusion
Detection Systems. International Journal of Network Security & Its Applications (IJNSA), Vol.4,
No.6, November 2012.
[12] Gates, Alan F., et al. "Building a high-level dataflow system on top of Map-Reduce: the Pig
experience." Proceedings of the VLDB Endowment 2.2 (2009): 1414-1425.
[13] H. F. Lipson. Tracking and Tracing Cyber-Attacks: Technical Challenges and Global Policy, 2002.
AUTHORS
Hani Qusa is a Assistant professor in Information Technology Department at University
College of Applied Sciences. He received his master and PhD degrees in 2009 and 2013
respectively in computer engineering from university of Rome “La Sapienza”.
Additionally, he published several topics in the area of network security, participated in
international workshops and exhibitions, and contributed with national and international
projects. His research interests include privacy preserving in collaborative environments,
secure multiparty computation protocols, distributed computing and secure distributed
systems.

COLLABORATIVE DEFENCE FOR DISTRIBUTED ATTACKS (CASE STUDY OF PALESTINIAN INFORMATION SYSTEMS)

More Related Content

Similar to COLLABORATIVE DEFENCE FOR DISTRIBUTED ATTACKS (CASE STUDY OF PALESTINIAN INFORMATION SYSTEMS) (20)

Recently uploaded (20)

COLLABORATIVE DEFENCE FOR DISTRIBUTED ATTACKS (CASE STUDY OF PALESTINIAN INFORMATION SYSTEMS)