Please use this identifier to cite or link to this item: http://theses.ncl.ac.uk/jspui/handle/10443/5107
Title: Physical Unclonability Framework for the Internet of Things
Authors: Goutsos, Konstantinos
Issue Date: 2020
Publisher: Newcastle University
Abstract: The rise of the Internet of Things (IoT) creates a tendency to construct unified architectures with a great number of edge nodes and inherent security risks due to centralisation. At the same time, security and privacy defenders advocate for decentralised solutions which divide the control and the responsibility among the entirety of the network nodes. However, spreading secrets among several parties also expands the attack surface. This conflict is in part due to the difficulty in differentiating between instances of the same hardware, which leads to treating physically distinct devices as identical. Harnessing the uniqueness of each connected device and injecting it into security protocols can provide solutions to several common issues of the IoT. Secrets can be generated directly from this uniqueness without the need to manually embed them into devices, reducing both the risk of exposure and the cost of managing great numbers of devices. Uniqueness can then lead to the primitive of unclonability. Unclonability refers to ensuring the difficulty of producing an exact duplicate of an entity via observing and measuring the entity’s features and behaviour. Unclonability has been realised on a physical level via the use of Physical Unclonable Functions (PUFs). PUFs are constructions that extract the inherent unclonable features of objects and compound them into a usable form, often that of binary data. PUFs are also exceptionally useful in IoT applications since they are low-cost, easy to integrate into existing designs, and have the potential to replace expensive cryptographic operations. Thus, a great number of solutions have been developed to integrate PUFs in various security scenarios. However, methods to expand unclonability into a complete security framework have not been thoroughly studied. In this work, the foundations are set for the development of such a framework through the formulation of an unclonability stack, in the paradigm of the OSI reference model. The stack comprises layers propagating the primitive from the unclonable PUF ICs, to devices, network links and eventually unclonable systems. Those layers are introduced, and work towards the design of protocols and methods for several of the layers is presented. A collection of protocols based on one or more unclonable tokens or authority devices is proposed, to enable the secure introduction of network nodes into groups or neighbourhoods. The role of the authority devices is that of a consolidated, observable root of ownership, whose physical state can be verified. After their introduction, nodes are able to identify and interact with their peers, exchange keys and form relationships, without the need of continued interaction with the authority device. Building on this introduction scheme, methods for establishing and maintaining unclonable links between pairs of nodes are introduced. These pairwise links are essential for the construction of relationships among multiple network nodes, in a variety of topologies. Those topologies and the resulting relationships are formulated and discussed. While the framework does not depend on specific PUF hardware, SRAM PUFs are chosen as a case study since they are commonly used and based on components that are already present in the majority of IoT devices. In the context of SRAM PUFs and with a view to the proposed framework, practical issues affecting the adoption of PUFs in security protocols are discussed. Methods of improving the capabilities of SRAM PUFs are also proposed, based on experimental data.
Description: Ph. D. Thesis
URI: http://theses.ncl.ac.uk/jspui/handle/10443/5107
Appears in Collections:School of Engineering

Files in This Item:
File Description SizeFormat 
Goutsos Konstantinos ethesis.pdfThesis1.99 MBAdobe PDFView/Open
dspacelicence.pdfLicence43.82 kBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.