Exploring the use of conversational agents to improve cyber situational awareness in the Internet of Things (IoT).

The Internet of Things (IoT) is an emerging paradigm, which aims to extend the power of the Internet beyond computers and smartphones to a vast and growing range of "things" - devices, processes and environments. The result is an interconnected world where humans and devices interact with each other, establishing a smart environment for the continuous exchange of information and services. Billions of everyday devices such as home appliances, surveillance cameras, wearables and doorbells, enriched with computational and networking capabilities, have already been connected to the Internet. However, as the IoT has grown, the demand for low-cost, easy-to-deploy devices has also increased, leading to the production of millions of insecure Internet-connected smart devices. Many of these devices can be easily exploited and leveraged to perform large-scale attacks on the Internet, such as the recently witnessed botnet attacks. Since these attacks often target consumer-level products, which commonly lack a screen or user interface, it can be difficult for users to identify signs of infection and be aware of devices that have been compromised. This thesis presents four studies which collectively explored how user awareness of threats in consumer IoT networks could be improved. Maintaining situational awareness of what is happening within a home network is challenging, not least because malicious activity often occurs in devices which are not easily monitored. This thesis evaluated the effectiveness of conversational agents to improve Cyber Situational Awareness. In doing so, it presented the first study to investigate their ability to help users improve their perception of smart device activity, comprehend this in the context of their home environment, and project this knowledge to determine if a threat had occurred or may occur in the future. The research demonstrated how a BLSTMRNN with word embedding could be used to extract semantic meaning from packets to perform deep packet inspection and detect IoT botnet activity. Specifically, how the models use of contextual information from both the past and future enabled better predictions to be made about the current state (packet) due to the sequential nature of the network traffic. In addition, a cross-sectional study examined users' awareness and perception of threats and found that, although users value security and privacy, they found it difficult to identify threats and infected devices. Finally, novel cross-sectional and longitudinal studies evaluated the use of conversational agents, and demonstrated them to be an effective and efficient method of improving Cyber Situational Awareness. In particular, this was shown to be true when using a multi-modal approach and combining aural, verbal and visual modalities.

Name dc:type.qualificationname

PhD

Level dc:type.qualificationlevel

Doctoral

Grantor dc:publisher.institution

Robert Gordon University

Year dc:date.issued

2020

Author dc:creator

• McDermott, Christopher D.

Advisor dc:contributor.advisor

• John P. Isaacs and Andrei V. Petrovski

• Internet of Things (IoT)
• Cyber attacks
• Cyber situational awareness
• Consumer awareness
• Malware detection
• Botnet detection
• Hacking detection

Language dc:language

en

Identifier

oai:rgu-repository.worktribe.com:1239552

Author Identifier

0000-0001-8718-0988

OAI identifier oai:identifier

oai:rgu-repository.worktribe.com:1239552