Abstract

With the rapid explosion of Internet of Things (IoT), IoT enabled devices have reached every corner of the globe, connecting billions of devices to the global internet. Botnets remain as one that can profit the most from IoT security susceptibility among other threats. Machine Learning based techniques to detect malicious traffic has been a widely researched topic in the last few years. Most of the studies simply use these techniques in an isolated fashion, where the machine learning algorithms are trained and tested on the same stream of labeled traffic. This study will attempt to analyze the use of supervised machine learning technique from practical aspects. To be practical and effective, a machine learning algorithm should be able to go beyond that and be able to classify traffic coming from a different sources, networks or software. Moreover we see that network traffic based datasets are usually in the order of millions of records which are then used to train the machine learning models. This study will also explore the impact of training size to find out how much data is actually required to train an effective machine learning model.

Publication Date

9-20-2020

Document Type

Thesis

Student Type

Graduate

Degree Name

Networking and System Administration (MS)

Advisor

Ali Raza

Advisor/Committee Member

Muhieddin Amer

Campus

RIT Dubai

Share

COinS