An Enhanced Fuzzy ARM Approach for Intrusion Detection

An Enhanced Fuzzy ARM Approach for Intrusion Detection

Nasser S. Abouzakhar, Huankai Chen, Bruce Christianson
Copyright: © 2011 |Pages: 21
DOI: 10.4018/jdcf.2011040104
(Individual Articles)
No Current Special Offers


The integration of fuzzy logic with data mining methods such as association rules has achieved interesting results in various digital forensics applications. As a data mining technique, the association rule mining (ARM) algorithm uses ranges to convert any quantitative features into categorical ones. Such features lead to the sudden boundary problem, which can be smoothed by incorporating fuzzy logic so as to develop interesting patterns for intrusion detection. This paper introduces a Fuzzy ARM-based intrusion detection model that is tested on the CAIDA 2007 backscatter network traffic dataset. Moreover, the authors present an improved algorithm named Matrix Fuzzy ARM algorithm for mining fuzzy association rules. The experiments and results that are presented in this paper demonstrate the effectiveness of integrating fuzzy logic with association rule mining in intrusion detection. The performance of the developed detection model is improved by using this integrated approach and improved algorithm.
Article Preview


The wide spread use of the Internet and WWW in today’s society led to various security vulnerabilities and threats. Such security problems have become increasingly important and an international priority to many organizations and countries. Within the network security community, researching network intrusion and current attacks is one of the major challenges, particularly, researching denial-of-service attacks. Intrusion detection (Proctor, 2001) is the process of monitoring computer and/or network activities and events, analysing them for sings of security threats such as unauthorised access, malicious activities and violations for security policy. Intrusion Detection Systems (IDSs) are capable of observing patterns of activities in user accounts and detect malicious behaviour. Intrusion detection systems are usually divided into two types (Carter, 2002): misuse detection approach and anomaly detection approach. The misuse detection approach tends to look for events that match certain network behaviors against well-defined intrusive patterns that are precisely written in advance. The anomaly detection approach attempts to evaluate a user or system behavior and consider intrusive or irregular activities as some deviation from the normal patterns. Such approach is capable of identifying newly developed attacks for which a well-defined intrusive pattern does not exist.

Denial-of service (DoS) attack (Moore et al., 2006; Tague, et al., 2009) is a computer network-based attack where an attacker floods a computer system or a network (victim) with useless traffic. Such attacks are difficult to be avoided since it is hard to distinguish the “good” requests from the “bad” ones. The association rule mining (ARM) technique (Agrawal & Srikant, 1994) has been applied into anomaly detection to automatically mine abnormal patterns from network data and/or audit data. One of the major limitations of such a mining approach (Changguo et al., 2009; Bridges, et al., 2000) of dealing with quantitative features is the sudden boundary problem. For example, an intrusion that deviates only slightly from the normal acceptable patterns may not be detected or a small change in normal patterns may trigger a false alarm. In order to deal with such a problem and improve the flexibility of the system, fuzzy logic has been integrated with ARM technique for intrusion detection.

It is possible to integrate fuzzy logic (Luo, 1999) with ARM as many quantitative features that are involved in intrusion detection can be treated as fuzzy variables. An example of quantitative feature is the number of different SYN flags in a fixed-interval of 1 sec or 2 sec. In traditional association rule mining, given a quantitative boundary, the quantitative feature can be split into two levels of categories i.e. Low or High. Any values of the quantitative feature falling below the boundary will be categorised as Low. Those values falling above the boundary will be categorised as High. Regardless of their distance to the boundary all values are categorised as either Low or High. Such an approach leads to the sudden separation of Low and High, which can be smoothed by integrating the Fuzzy Logic.

We present a new FARM algorithm named Matrix Fuzzy Association Rule Mining (Matrix FARM) algorithm. This algorithm improves the performance of the proposed detection model which uses FARM techniques to mine new patterns from the network traffic dataset. The model calculates the similarity between the new patterns and the normal patterns mined from normal network traffic. If the similarity value is under a user-defined threshold, the model will generate an alarm indicate that there may be some anomalies occurred in the network traffic. The evaluation of the proposed new model is achieved by testing the performance of the developed Matrix FARM using standard methods.

Complete Article List

Search this Journal:
Volume 16: 1 Issue (2024): Forthcoming, Available for Pre-Order
Volume 15: 1 Issue (2023)
Volume 14: 3 Issues (2022)
Volume 13: 6 Issues (2021)
Volume 12: 4 Issues (2020)
Volume 11: 4 Issues (2019)
Volume 10: 4 Issues (2018)
Volume 9: 4 Issues (2017)
Volume 8: 4 Issues (2016)
Volume 7: 4 Issues (2015)
Volume 6: 4 Issues (2014)
Volume 5: 4 Issues (2013)
Volume 4: 4 Issues (2012)
Volume 3: 4 Issues (2011)
Volume 2: 4 Issues (2010)
Volume 1: 4 Issues (2009)
View Complete Journal Contents Listing