Search

Group communication allows a collection of transceivers to mutually share messages, from either a single or multiple sources to multiple sinks. This form of communication has numerous applications, including vehicular networking, content distribution and multimedia streaming. Reliability and low latency are highly desirable for these applications. This thesis addresses the problem of low latency group communication by developing lightweight distributed algorithms. Most previous work in this area focuses on wired point-to-point links. Motivated by the increased prevalence of wireless networks, this thesis focuses on broadcast channels. The first problem addressed in this thesis is broadcasting from a single source to large numbers of receivers. The channels to distinct receivers are assumed to be independent discrete memoryless erasure channels. Random Linear Network Coding (RLNC) is shown to achieve non-vanishing throughput, unachievable by ARQ, in exchange for delay scaling with the number of receivers n as O(log(n)). The tradeoff between its throughput and delay is quantified. The next problem considered is all-to-all (allcast) communication in which agents take it in turns to broadcast packets. Each broadcast is received correctly by only a subset of the agents, and is erased at the others. The network of correct receptions at each time-step is modelled by a random graph, and these graphs are correlated over time. Three algorithms to address this problem are proposed: one in which messages are randomly forwarded, and two RLNC algorithms in which random linear combinations of messages are broadcast. A rigorous mathematical analysis of these algorithms is presented for graphs which are constant over time. The RLNC algorithms are shown to complete in a constant number of time-steps, as opposed to O(log(n)) for the baseline, enabling better scalability to large networks. Finally the analysis is supplemented by simulations, which address the more general setting of graphs evolving as Markov chains.

In this thesis, I study the stellar angular momentum within one effective radius, λRe of a few thousand galaxies and consider how this property relates to other galaxy properties such as stellar mass, optical morphology and environment. I make use of data from the Sloan Digital Sky Survey's (SDSS) Mapping Nearby Galaxies at Apache Point Observatory (MaNGA) survey which is a large integral field spectroscopic survey. I classify early-type galaxies (ETGs) as slow rotators if λRe < 0.08 + 0.25ε and ε < 0.4, where ε is the galaxy flattening, and as fast rotators if otherwise. I present the (λRe,ε) diagram for about 2300 galaxies and provide the first conclusive evidence that fast and slow rotator ETGs form a bimodality in angular momentum suggesting that they are kinematically distinct populations. I show that they also form a bimodality in other properties such as stellar mass and kinematic morphology, which I attribute as being a direct consequence of the existence of two ETG formation channels, a gas rich (wet) accretion channel which produces FRs and a gas poor (dry) merger channel which produces SRs. I investigate the hypothesis that massive SRs have built up their stellar mass and decreased their angular momentum as a result of experiencing multiple dry minor mergers by studying their environment. By running a new group-finder algorithm on a large galaxy catalogue containing spectroscopic and photometric observations, I find the neighbouring galaxies for about 4600 MaNGA galaxies. For all neighbours that satisfy the criteria for massive SRs (M ≥ 2 × 1011 M⊙ and ε<0.4), I assign an angular momentum classification based on optical morphology to achieve a complete, volume-limited sample of about 14,000 galaxies. Using this sample, I reproduce the kinematic morphology-density (kT-Σ) relation for massive SRs, finding that they are about six times more abundant in the dense centres of clusters compared to the field. I show that when the kT-Σ relation is derived from multiple clusters, the individual cluster trends are washed out. I account for this smoothing effect by defining a relative density estimator, Σrel, that measures number density relative to the cluster/group peak density. I show that the kT-Σrel relation is a better predictor of the abundance of massive SRs than the kT-Σ relation. I find that these environmental trends persist at fixed stellar mass, although they are subtle because stellar mass is the main driver for slow rotation. I also present schematics for 10 large galaxy clusters highlighting the location of the massive SRs. In all cases, the density peaks contain at least one massive SR, while FRs are distributed evenly throughout the cluster. I conclude that these results provide definitive proof that the progenitors of modern-day massive SRs formed rapidly at high redshift in dense environments before accreting a significant fraction of their stellar mass via dry minor mergers. Finally, I suggest potential avenues for follow-up.

Diffusion-weighted MRI (DW-MRI) is a powerful, non-invasive imaging technique that allows us to infer the structure of biological tissue. It is particularly well suited to the brain, and is used by clinicians and researchers studying its structure in health and disease. High quality data is required to accurately characterise tissue structure with DW-MRI. Obtaining such data requires the careful optimisation of the image acquisition and processing pipeline, in order to maximise image quality and minimise artefacts. This thesis extends an existing MRI simulator to create a simulation system capable of producing realistic DW-MR data, with artefacts, and applies it to improve the acquisition and processing of such data. The simulator is applied in three main ways. Firstly, a novel framework for evaluating post-processing techniques is proposed and applied to assess commonly used strategies for the correction of motion, eddy-current and susceptibility artefacts. Secondly, it is used to explore the often overlooked susceptibility-movement interaction. It is demonstrated that this adversely impacts analysis of DW-MRI data, and a simple modification to the acquisition scheme is suggested to mitigate its impact. Finally, the simulation is applied to develop a new tool to perform automatic quality control. Simulated data is used to train a classifier to detect movement artefacts in data, with performance approaching that of a classifier trained on real data whilst requiring much less manually-labelled training data. It is hoped that both the findings in this thesis and the simulation tool itself will benefit the DW-MRI community. To this end, the tool is made freely available online to aid the development and validation of methods for acquiring and processing DW-MRI data.

The Cloud Security Alliance’s 2015 Cloud Adoption Practices and Priorities Survey reports that 73% of global IT professionals cite security as the top challenge holding back cloud services adoption. Malware with the capabilities to jump between the abstracted virtual infrastructures found within cloud service provider networks heightens the threat from botnet attack upon a cloud infrastructure. This research project aimed to provide a novel methodological approach for capturing communication traffic between botnets. The originality of this study comes from the application of standards-based IPFIX flow export protocol as a traffic capture mechanism. The first contribution to knowledge is a critical investigation into how IPFIX export overcomes the limitations of traditional NetFlow-based botnet communication traffic capture in cloud provider networks. The second contribution is the BotProbe IPFIX template, comprising eleven IANA IPFIX information elements. Field occupancy count and Spearman’s Rank correlation on 25 million botnet flows created an IPFIX template tailored specifically for botnet traffic capture. The third contribution is BotStack, a modular, non-intrusive IPFIX monitoring framework, created upon Xen hypervisor and virtual switched platforms, to incorporate IPFIX export into existing cloud stacks. The fourth contribution is compelling empirical evidence from weighted-factor observation across multiple network vantage points, that siting IPFIX exporters on the host hypervisor provides maximum traffic visibility. BotProbe performs on average 26.73%±0.03% quicker than traditional NetFlow v5, with 14.06%±0.01% less storage requirements. BotProbe can be extended with additional application layer attributes, for use in less privacy sensitive environments. Both novel IPFIX templates were tested on the BotStack framework, capturing four distinct traffic profiles in the life cycle of a Zeus botnet. The techniques developed in this research can be repurposed to create IPFIX traffic capture templates for most Cybersecurity threats, including DDoS and spam, turning behavioural-based traffic capture from a big data challenge into a manageable data solution.

This thesis examines the evolution of home-working within British-based telecommunications giant, BT; looking at the evolution and effects of this policy on both the management practices of the corporation and the working practices of home workers. It is argued that essential to understanding homeworking in BT is an analysis of the changing nature of corporate culture and governance within the organisation. These issues are analysed through both a detailed examination of BT policy documents and in-depth semi-structured interviews with28 home workers at various levels within the organization. The thesis examines why senior management has thought it necessary to change corporate culture within BT since privatization, the numerous manifestations of these changes, and the role of home working within these transformations. The study examines whether there is a distinctive culture associated with home working and the effect of this on individual's identities. A central concern of the thesis is to understand how it is possible for a corporation to maintain control over a workforce that is geographically dispersed and no longer within its direct scrutiny. What forms do resistance to change take and how can these be overcome? Whilst Foucault's notion of surveillance and its development into the concept of the e-Panopticon, were found to be wanting as an explanation for control, other notions of Foucault's combined with the insights of Allen were found to be useful. In particular, the Foucauldian notions of governance highlight the techniques used to drive change within BT. In addition, the analysis reveals how trust is central to running a 'flexible' organization. Finally, a synthesis of Foucauldian ideas on ethics and Allen's work on the spatiality of power is used to formulate notions of 'ethical modes of power' and understand the ways in which employees come to accept corporate change. It is argued that identity and governance interact in such a way that modes of power can have an effect on the employee wherever they may be based.

Clair Groups and underlying basement contain cataclastic faults, cement-sealed faults/veins and open fractures. The dominant fracture cements are calcite and pyrite, although fluorite (with sulphides and native silver) occurs in minor amounts in some Clair Group fractures, marcasite is often associated with Clair Group fracture-hosted pyrite, and some basement fractures contain epidote/chlorite and quartz/feldspar/hematite. Calcite cements in many Clair Group fractures may have crystallised from porewaters whose calcium and bicarbonate were mostly remobilised from Clair Group calcretes. Many of these cements may predate Late Cretaceous/Early Tertiary oil charge. Some may have crystallised from Early Tertiary meteoric porewaters, the fluorite with which they are associated perhaps having crystallised from porewaters hydrothermally affected by local igneous activity. Calcite cements within some basement fracture may have crystallised from porewaters whose calcium and bicarbonate came from Clair Group calcretes, as may most non-ferroan calcite cement within Clair Group sandstones. Calcite cements in some Clair Group fractures contain a high proportion of non-calcrete carbonate. These cements may have crystallised after oil charge began, from Early Tertiary meteoric porewaters or meteoric/basinal-derived brine mixtures. Non-calcrete bicarbonate may have been generated via bacterial activity (organic matter oxidation and methanogenesis) within the Clair Group. Bacteria could have utilised hydrocarbons, substrates generated via aerobic hydrocarbon oxidation, or externally-sourced substrates such as acetate. Authigenic iron sulphides often predate the calcite cements. The sulphur within these was probably externally-sourced, H2S either having come direct from external sources, or having been generated in situ via bacterial sulphate reduction. Sulphur may have undergone redox cycling through contact with oxygenated meteoric water in some fractures and marcasite genesis may have resulted from this. Calcite cements within some basement fractures may have crystallised from Early Tertiary porewaters containing bicarbonate generated in situ via the oxidation of hydrocarbons, and these cements are often predated by authigenic pyrite.

It is commonly argued that only a handful of technology firms control the infrastructure that underpins contemporary forms of artificial intelligence (AI). However, little is known about how this concentration of computational resources manifests in new production geographies. This thesis asks: What forces enact, and are enacted by, AI's geographies of production? In answering this question, it develops the original framework of neural production networks: geographically dispersed but computationally enveloped production arrangements powered by artificial neural networks. Due to the impacts of AI on economic, political, and cultural interactions, neural production networks already infiltrate infrastructural life. Drawing on a paradigmatic case study design that is primarily informed by document analysis, the thesis applies this new framework to answer three research questions: First, how do technology giants act as lead firms within AI's geographies of production? Second, how does state action shape, and is shaped by, AI's geographies of production? Third, what do AI's geographies of production imply for platformised cultural production? The thesis answers these questions by disassembling neural production networks into three of their mutually conditioning forces. The first force is envelopment: Acting as industry-dominating lead firms, Amazon, Google, and Microsoft decentralise the production of AI and centralise its provision, thereby enclosing users into their proprietary ecosystems. The second force is sovereignty: Although the EU's Digital Single Market Strategy attempts to push back against an infrastructural dependency on the computational resources of American lead firms, its agency to reconfigure power relations in neural production networks remains limited. The third force is hyperproduction: Underlying the penetration of cultural life with AI-generated media, new revenue streams emerge that purport to destabilise the boundary between reality and simulation. By opening out a conceptual space that probes how those forces relate to each other, the thesis expounds a new understanding of production geographies. This conceptual space contributes to cross-disciplinary scholarship at the intersection of platform studies and economic geography. Additionally, the arguments and insights of this thesis are relevant to policymakers and regulators facing the task of governing the diffusion of neural production networks into ever-more societal spheres.

Proponents of the decentralised database technology blockchain suggest it will revolutionise the aid sector by allowing money and data to be exchanged more equitably, efficiently, transparently, and securely. Blockchain is widely expected to improve refugee identification, cross-border remittances, supply chain management and more. However, humanitarian blockchain projects are nascent and there is a lack of critical scholarship on how the promises for blockchain are playing out in practice. This thesis provides the first comprehensive, theoretically informed, empirically grounded case study of the use of blockchain in humanitarianism. I investigate the imaginaries, uses and effects of blockchain in anthropological terms, based on close-up work with the key stakeholder communities connected to a pilot project in Jordan's refugee camps, which in 2019 began using blockchain to deliver financial aid to Syrian refugee women. Drawing together a range of perspectives on the same project, I show how the aspirations for this new technology materialise, in support of some priorities-those of UN donors, agencies and their corporate and government partners-at the expense of others-aid workers and refugees. Building on critical data studies, infrastructure studies, and aid and migration studies, I put forward the theoretical frame of infrastructure justice. With this, I advance scholarship on socio-economic injustices and political hegemonies in the digital age. I argue that injustices are experienced, extended, and contested through three interlocking dimensions of infrastructure: subjectivities, timescapes, and materialities. I demonstrate how strongly affective cultural and subjective experiences with blockchain-involving suspicion, ignorance, mystification, and faith-connect with the asymmetrical regulation of information, material resources and strategic interests over space and time. This thesis gives fresh insight into how radical, emergent technology implemented at socio-economic margins both challenges and is incorporated into the structures of global governance, and in particular how it is experienced by its presumed beneficiaries.

Recent years have witnessed a surge in the development of real estate data analytics platforms. These devices variously promise a revolution in investment and valuation, leading to a more transparent knowledge of market reality and more objective decision-making. To examine these claims, this research adopts a theoretical framework from STS and critical data studies to examine the ethnographic case study of one prototype data analytics product throughout the early stages of its development and marketing. This product, Dashboard X, was designed to provide large institutional investors with a way to better know which areas of London were currently undergoing gentrification for the purpose of investing in Build-to-Rent housing. This study was followed by a series of interviews and participation in several industry events. By examining the actors emerging around this would-be 'market device', this research explains that Dashboard X and similar platforms are in fact producing and circulating entirely new, subjective, and distinctly situated forms of real estate values, knowledges, and even assets. Furthermore, it is argued Dashboard X entails a series of theoretical and political implications that range from the co-performation of planetary gentrification and the expansion of rentier capitalism, to a 'new urban colonialism' and the greater ontological politics associated with bringing new real estate assets and markets into being. The key theoretical contribution is that these data analytics platforms do little to objectively increase investor rationality in the sense of knowing reality, but that the use of these technologies can nevertheless powerfully control the future trajectories and possibilities of a global real estate market by circulating the values and knowledges needed for its realisation. Hence, these new urban real estate values can precede their realisation in material marketplace reality through data analytics platforms as hyperreal estate.

Innovation hub organizations - or 'hubs' - have become a prevalent form of support for technology entrepreneurship in Africa. About 170 African hubs have been established, most since 2010. Practitioners have argued either that hubs are transformative network infra-structures for Africa's fledgling digital economy or that they are ineffective business incubators. This thesis steps back from this debate about whether hubs work. Instead, it asks how African hubs work, specifically how they shape relationships of technology entrepreneurs. Literature on intermediation and incubation is reviewed to establish a theoretical framework. The thesis then tests and extends the framework based on an extensive, grounded empirical inquiry. In-depth case study data (including 119 interviews with 133 participants) on six hubs were collected during field studies in Kigali, Harare, and Accra from September-December 2014. The thesis finds that the analyzed hub organizations were defined by nested, fluidly bounded entrepreneurial communities. Communities varied by their level of activation: mem-bers of active communities had concern for each other and recognized communities as social entities, while inactive community members only shared a loose purpose. The six hubs followed two distinct organizational patterns: the technology hub (depending on active core communities) and the entrepreneurship hub (relying on active peripheral communities). Based on these results, the thesis theorizes hubs as assemblers of technology entrepreneurs: hubs assemble previously distant and different actors into entrepreneurial communities. Assembly is unique to hubs: it is related to but different from incubation and most forms of intermediation. Assembly theory addresses important meso-level analytical gaps in prior research on the coordination and organization of entrepreneurship. The thesis underscores limitations in African technology entrepreneurship environments, advising hub practitioners to acknowledge that 'only what is there can be assembled.' Ultimately, it highlights that hubs have been critically misunderstood, and clarifies what hubs can and cannot do for technology entrepreneurs.

Wikipedia is no longer just another source of knowledge about the world. It is fast becoming a central source, used by other powerful knowledge brokers like Google and Bing to offer authoritative answers to search queries about people, places and things and as information infrastructure for a growing number of Web applications and services. Researchers have found that Wikipedia offers a skewed representation of the world that favours some groups at the expense of others so that representations on the platform have repercussions for the subjects of those representations beyond Wikipedia's domain. It becomes critical in this context to understand how exactly Wikipedia's representations come about, what practices give rise to them and what socio-technical arrangements lead to their expression. This ethnographic study of Wikipedia explores the values, principles and practices that guide what knowledge Wikipedia represents. It follows the foundational principles of Wikipedia in its identity both as an encyclopaedia and a product of the free and open source software and internet freedom rhetoric of the early 2000s. Two case studies are analysed against the backdrop of this ideology, illustrating how different sets of actors battle to extend or reject the boundaries of Wikipedia, and in doing so, affect who are defined as the experts, subjects and revolutionaries of the knowledge that is taken up. The findings of this thesis indicate that Wikipedia's process of decision-making is neither hierarchical nor is it egalitarian; rather, the power to represent on Wikipedia is rhizoid: it happens at the edges rather than in the centre of the network. Instead of everyone having the same power to represent their views on Wikipedia, those who understand how to perform and speak according to Wikipedia's complex technical, symbolic and policy vocabulary tend to prevail over those who possess disciplinary knowledge about the subject being represented. Wikipedians are no amateurs as many would have us believe; nor are they passive collectors of knowledge held in sources; Wikipedians are, instead, active co-creators of knowledge in the form of facts that they support using specially chosen sources. The authority of Wikipedia and Wikipedians is garnered through the performative acts of citation, through the ability of individual editors to construct the traces that represent citation, and through the stabilization and destabilization of facts according to the ideological viewpoints of its editors. In venerating and selecting certain sources among others, Wikipedians also serve to reaffirm traditional centres of authority, while at the same time amplifying new centres of knowledge and denying the authority of knowledge that is not codified in practice. As a result, Wikipedia is becoming the site of new centres of expertise and authoritative knowledge creation, and is signalling a move towards the professionalization of the expertise required to produce factual data in the context of digital networks.