Detecting and defending against cyber attacks in a smart home Internet of Things ecosystem

Anthi, Eirini 2022. Detecting and defending against cyber attacks in a smart home Internet of Things ecosystem. PhD Thesis, Cardiff University. Item availability restricted.

Preview	PDF (PhD Thesis) - Accepted Post-Print Version Available under License Creative Commons Attribution No Derivatives. Download (8MB) | Preview
	PDF (Publication Form) - Supplemental Material Restricted to Repository staff only Download (152kB)

Abstract

The proliferation in Internet of Things (IoT) devices is demonstrated by their prominence in our daily lives. Although such devices simplify and automate everyday tasks, they also introduce tremendous security flaws. Current security measures are insufficient, making IoT one of the weakest links to breaking into a secure infrastructure which can have serious consequences. Subsequently, this thesis is motivated by the need to develop and further enhance novel mechanisms tailored towards strengthening the overall security infrastructures of IoT ecosystems. To estimate the degree to which a hub can improve the overall security of the ecosystem, this thesis presents a design and prototype implementation of a novel secure IoT hub, consisting of various built-in security mechanisms that satisfy key security properties (e.g. authentication, confidentiality, access control) applicable to a range of devices. The effectiveness of the hub was evaluated within a smart home IoT network upon which popular cyber attacks were deployed. To further enhance the security of the IoT environment, the initial experiments towards the development of a three-layered Intrusion Detection System (IDS) is proposed. The IDS aims to: 1) classify IoT devices, 2) identify malicious or benign network packets, and 3) identify the type of attack which has occurred. To support the classification experiments, real network data was collected from a smart home testbed, where a range of cyber attacks from four main attack types were targeted towards the devices. Lastly, the robustness of the IDS was further evaluated against Adversarial Machine Learning (AML) attacks. Such attacks may target models by generating adversarial samples which aim to exploit the weaknesses of the pre-trained model, consequently bypassing the detector. This thesis presents a first approach towards automatically generating adversarial malicious DoS IoT network packets. The analysis further explores how adversarial training can enhance the robustness of the IDS.

Item Type:	Thesis (PhD)
Date Type:	Completion
Status:	Unpublished
Schools:	Computer Science & Informatics
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science Q Science > QA Mathematics > QA76 Computer software
Date of First Compliant Deposit:	8 March 2022
Date of Acceptance:	3 March 2022
Last Modified:	09 Mar 2022 10:16
URI:	https://orca.cardiff.ac.uk/id/eprint/148044

Actions (repository staff only)

Edit Item