Toronto Metropolitan University
Browse

Network intrusion detection using machine learning

Download (3.16 MB)
thesis
posted on 2021-05-22, 11:59 authored by Seyed Pedrum Jalali Mosallam
In this research we have studied the use of machine learning techniques in detecting network intrusions. Most research in the field has used the very outdated dataset (KDDCup99) which consists of a set handcrafted features. In our research we present models that work well on both the older dataset and on newer datasets such as ISCX2014 and ISCX2012. We also present methods for extracting features from these datasets. Another issue we found with most research in this field is that they do not study the effect of surges in regular network traffic and how that might affect the model. We put our model to test in 10x traffic and show its effectiveness under these conditions. We also study how semi-supervised models can be used in training NIDS models without directly showing them labeled data.

History

Language

English

Degree

  • Master of Science

Program

  • Computer Science

Granting Institution

Ryerson University

LAC Thesis Type

  • Thesis

Year

2018

Usage metrics

    Computer Science (Theses)

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC