An enhanced resilient backpropagation artificial neural network for intrusion detection system

Dissertant

al-Sultani, Zaynab Nimah Abd Allah

Thesis advisor

Naum, Riyad Shakir

Comitee Members

Husayn, Abd al-Amir Khalaf
Shita, Ala F.

University

Middle East University

Faculty

Faculty of Information Technology

Department

Computer Science Department

University Country

Jordan

Degree

Master

Degree Date

2012

English Abstract

Over the last two decades, computer and network security has become a main issue, especially with the increase number of intruders and hackers, therefore systems were designed to detect or/and prevent intruders.

This research presents a hybrid intrusion detection system models, using k-Nearest Neighbor machine learning algorithm and an enhanced resilient backpropagation artificial neural network.

The proposed system is divided into five phases: environment phase, dataset features and pre-processing phase, feature classification k-Nearest Neighbor (kNN) phase, training the enhanced resilient backpropagation neural network phase and testing the hybrid system phase.

k- Nearest Neighbor as a machine learning algorithm was used in the first stage of classification using the first norm which demonstrates better results than the second norm.

A multilayer perceptron as the second stage of classification was trained using an enhanced resilient backpropagation training algorithm.

Best number of hidden layers, hidden neurons and training iterations were calculated to train the enhanced resilient backpropagation neural network.

One hidden layer with 34 hidden neurons was used in resilient backpropagation artificial neural network training process.

An optimal learning factor was derived to speed up the convergence of the resilient backpropagation neural network performance.

The experiments have shown that the hybrid system (kNN_ERBP) was able to classify normal class using the k- Nearest Neighbor, and the enhanced resilient backpropagation on the other hand was able to classify intrusions classes with high detection rate and with less time than the ordinary resilient backpropagation.

The evaluations were performed using the NSLKDD99 network anomaly intrusion detection dataset.

The experiments results demonstrate that the proposed system (kNN_ERBP) has a detection rate about 97.2% with an accuracy rate of 99%.

The proposed hybrid system (kNN_ERBP) was compared to other Intrusion Detection Systems that was designed using supervised learning such as ordinary backpropagation and unsupervised learning such as k-means and Kohonen self organizing maps.

Main Subjects

Information Technology and Computer Science

No. of Pages

88

Table of Contents

• APA
• MLA
• AMA

Language

English

Data Type

Arab Theses

Record ID

BIM-694110