UUM ETD | Universiti Utara Malaysian Electronic Theses and Dissertation
FAQs | Feedback | Search Tips | Sitemap

Intrusion Detection in Mobile Ad Hoc Networks Using Transductive Machine Learning Techniques

Farhan, Farhan Abdel-Fattah Ahmad (2011) Intrusion Detection in Mobile Ad Hoc Networks Using Transductive Machine Learning Techniques. PhD. thesis, Universiti Utara Malaysia.

[img] PDF
Farhan_Abdel-Fattah_Ahmad_Farhan.pdf
Restricted to Registered users only

Download (1MB)
[img]
Preview
PDF
1.Farhan_Abdel-Fattah_Ahmad_Farhan.pdf

Download (411kB) | Preview

Abstract

This thesis presents a research whose objective is to design an intrusion detection model for Mobile Ad hoc NETworks (MANET). MANET is an autonomous system consisting of a group of mobile nodes with no infrastructure support. The MANET environment is particularly vulnerable because of the characteristics of mobile ad hoc networks such as open medium, dynamic topology, distributed cooperation, and constrained capability. Unfortunately, the traditional mechanisms designed for protecting networks are not directly applicable to MANETs without modifications. In the past decades, machine learning methods have been successfully used in several intrusion detection methods because of their ability to discover and detect novel attacks. This research investigates the use of a promising technique from machine learning to designing the most suitable intrusion detection for this challenging network type. The proposed algorithm employs a combined model that uses two different measures (nonconformity metric measures and Local Distance-based Outlier Factor (LDOF)) to improve its detection ability. Moreover, the algorithm can provide a graded confidence that indicates the reliability of the classification. In machine learning algorithm, choosing the most relevant features for each attack is a very important requirement, especially in mobile ad hoc networks where the network topology dynamically changes. Feature selection is undertaken to select the relevant subsets of features to build an efficient prediction model and improve intrusion detection performance by removing irrelevant features. The transductive conformal prediction and outlier detection have been employed for feature selection algorithm. Traditional intrusion detection techniques have had trouble dealing with dynamic environments. In particular, issues such as collects real time attack related audit data and cooperative global detection. Therefore, the researcher is motivated to design a new intrusion detection architecture which involves new detection technique to efficiently detect the abnormalities in the ad hoc networks. The proposed model has distributed and cooperative hierarchical architecture, where nodes communicate with their region gateway node to make decisions. To validate the research, the researcher presents case study using GLOMOSIM simulation platform with AODV ad hoc routing protocols. Various active attacks are implemented. A series of experimental results demonstrate that the proposed intrusion detection model can effectively detect anomalies with low false positive rate, high detection rate and achieve high detection accuracy.

Item Type: Thesis (PhD.)
Uncontrolled Keywords: Computer Networks, Intrusion Detection, Mobile Ad Hoc Networks, Transductive Machine Learning Techniques
Subjects: T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5101-6720 Telecommunication
Divisions: College of Arts and Sciences (CAS)
Depositing User: Mr Husni Ismail
Date Deposited: 22 Nov 2011 09:05
Last Modified: 27 Apr 2016 03:11
URI: http://etd.uum.edu.my/id/eprint/2540

Actions (login required)

View Item View Item