dc.description.abstract |
The world is embracing the idea of Internet of Things and Industrial Revolution 4.0. However, this acceptance of computerised evolution is met with a myriad of challenges, where consumers of this technology are also growing ever so anxious about the security of their personal data as well as reliability of data collected by the millions and even billions of sensors surrounding them.
Wireless sensor networks are the main baseline technology driving Internet of things; by their very inherent nature, these networks are too vulnerable to attacks and yet the network security tools designed for conventional computer networks are not effective in countering these attacks. Wireless sensors have low computational resources, may be highly mobile and in most cases, these networks do not have a central point which can be marked as an authentication point for the sensors, any node can join or leave whenever they want. This leaves the sensors and the internet of things applications depending on them highly susceptible to attacks, which may compromise consumer information and leave security breaches in situation that need absolute security such as homes or even the cars they drive. There are many possibilities of things that could go wrong when hackers gain control of sensors in a car or a house.
There have been many solutions offered to address security of Wireless Sensor Networks; however, most of those solutions are often not customised for African context. Given that most African countries have not kept pace with the development of these underlying technologies, blanket adoption of the solutions developed for consumption in the developed world has not yielded optimal results. The focus of this research was the development of an Intrusion Detection System that works in a hierarchical network structured Wireless Sensor Network, where cluster heads oversee groups of nodes and relay their data packets all the way to the sink node. This is a reactive Intrusion Detection System (IDS) that makes use of a fuzzy logic based algorithm for verification of intrusion detections. This system borrows characteristics of traditional Wireless Sensor Networks in that it is hosted external to the nodes; that is, on a computer or server connected to the sink node. The rational for this is the premise that developing the system in this manner optimises the power and processing resource of nodes because no part of the IDS is found in the nodes and they are left to focus purely on sensing.
The Intrusion Detection System makes use of remote Over The Air programming to communicate with compromised nodes, to either shut down or reboot and is designed with the ZigBee protocol in mind. Additionally, this Intrusion Detection System is intended to being part of a larger Internet of Things integration framework being proposed at the Central University of Technology. This framework is aimed at developing an Internet of Things adoption strategy customised for African needs and regionally local consumers.
To evaluate the effectiveness of the solution, the rate of false detections being picked out by the security algorithm were reduced through the use of fuzzy logic systems; this resulted in an accuracies of above 90 %. The algorithm is also very light when asymptotic notation is applied, making it ideal for Wireless Sensors. Lastly, we also put forward the Xbee version of the Triple Modular Redundancy architecture, customised for Wireless sensor networks in order to beef-up on the security solution presented in this dissertation. |
en_US |