Date of Award

2020

Degree Type

Thesis

Degree Name

Doctor of Philosophy

School

School of Science

First Advisor

Dr Ahmed Ibrahim

Second Advisor

Dr Zubair Baig

Third Advisor

Professor Craig Valli

Abstract

Internet of Things (IoT) is poised to transform the quality of life and provide new business opportunities with its wide range of applications. However, the bene_ts of this emerging paradigm are coupled with serious cyber security issues. The lack of strong cyber security measures in protecting IoT systems can result in cyber attacks targeting all the layers of IoT architecture which includes the IoT devices, the IoT communication protocols and the services accessing the IoT data. Various IoT malware such as Mirai, BASHLITE and BrickBot show an already rising IoT device based attacks as well as the usage of infected IoT devices to launch other cyber attacks. However, as sustained IoT deployment and functionality are heavily reliant on the use of e_ective data communication protocols, the attacks on other layers of IoT architecture are anticipated to increase. In the IoT landscape, the publish/- subscribe based Message Queuing Telemetry Transport (MQTT) protocol is widely popular. Hence, cyber security threats against the MQTT protocol are projected to rise at par with its increasing use by IoT manufacturers. In particular, the Internet exposed MQTT brokers are vulnerable to protocolbased Application Layer Denial of Service (DoS) attacks, which have been known to cause wide spread service disruptions in legacy systems. In this thesis, we propose Application Layer based DoS attacks that target the authentication and authorisation mechanism of the the MQTT protocol. In addition, we also propose an MQTT protocol attack detection framework based on machine learning. Through extensive experiments, we demonstrate the impact of authentication and authorisation DoS attacks on three opensource MQTT brokers. Based on the proposed DoS attack scenarios, an IoT-MQTT attack dataset was generated to evaluate the e_ectiveness of the proposed framework to detect these malicious attacks. The DoS attack evaluation results obtained indicate that such attacks can overwhelm the MQTT brokers resources even when legitimate access to it was denied and resources were restricted. The evaluations also indicate that the proposed DoS attack scenarios can signi_cantly increase the MQTT message delay, especially in QoS2 messages causing heavy tail latencies. In addition, the proposed MQTT features showed high attack detection accuracy compared to simply using TCP based features to detect MQTT based attacks. It was also observed that the protocol _eld size and length based features drastically reduced the false positive rates and hence, are suitable for detecting IoT based attacks.

Share

 
COinS