Back propagation recursive least squares algorithm for intrusion detection systems

Other Title(s)

خوارزمية الانتشار المرتد أصغر التربيعات التكرارية لأنظمة كشف التطفل

Dissertant

al-Sadun, Umar Hisham Rashid

Thesis advisor

al-Hammuz, Sadiq O.

Comitee Members

Bani Muhammad, Ashraf
Naum, Riyad Shakir
Nasir al-Din, Hibah H. O.

University

Middle East University

Faculty

Faculty of Information Technology

Department

Computer Science Department

University Country

Jordan

Degree

Master

Degree Date

2014

English Abstract

Since malicious attacks have increased on wire computer networks and wireless systems, in various known and unknown types, conventional security techniques such as firewalls and available anti-viruses programs are not adequate to provide reliable, integrated and secure networks.

Intrusion Detection Systems (IDSs) are considered one of the most reliable and tested technologies that used to control the network traffic in order to distinguish any kind of computer system network mishandling or identifying any kind of unauthorizing usage, which make the implementation of intrusion detection system a critical issue in computer networks.

Intrusion detection systems form a principal part of the network system defence; however, intrusion detection is not yet a perfect and mature technology.

This fact provided a great opportunity for data mining to create many leading contributions in the field of intrusion detection.

In this thesis, we have proposed three data mining techniques: K-Means algorithm, Novel K- nearest neighbours (KNN) algorithm and Recursive Least Squares Back Propagation Neural Network (RLSBP NN) for intrusion detection fortification.

The proposed techniques have three common major phases; in first NSL-KDD dataset, pre-processing is performed, and then part of processed dataset (8%) is used to train the proposed algorithms.

Finally, the other part of dataset (2%) is used in sake of performance evaluation.

Data set used is NSL-KDD dataset and we have used detection rate (recall), accuracy, FPR, TNR, TPR, and FNR as evaluation measurements.

Our proposed IDSs are implemented and tested in MATLAB 2012 environment, where the experimental results shows superiority and high stability of our proposed systems in terms of detection rate ,accuracy, sensitivity, and specificity for all types of intrusions attacks.

The results of our proposed techniques are compared with each other and with other existing schemes.

These comparisons proved the reliability and effectiveness of the proposed techniques.

Main Subjects

Information Technology and Computer Science

Topics

• Algorithms
• Computer networks
• Neural networks(Computer science)
• Security measures
• Data mining

No. of Pages

110

Table of Contents

• APA
• MLA
• AMA

Language

English

Data Type

Arab Theses

Record ID

BIM-698583