21684 research outputs found
Sort by
Securing intelligent networks: federated learning approaches for privacy-conscious anomaly detection
Get PDFAs intelligent systems increasingly rely on distributed data generated at the edge, traditional centralized machine learning paradigms face critical limitations related to data privacy, band-width constraints, and vulnerability to single points of failure. Federated Learning (FL) has emerged as a promising distributed alternative that enables collaborative model training across devices without sharing raw data, preserving privacy while leveraging collective intelligence. In smart energy systems, Short-Term Load Forecasting (STLF) exemplifies a critical application where FL’s privacy-preserving capabilities offer significant advantages, supporting efficient grid operation. However, deploying FL in dynamic and heterogeneous environments poses unique challenges, including security vulnerabilities from untrusted participants, communication bottlenecks in low-bandwidth networks, and managing data heterogeneity across clients. This research systematically addresses these hurdles by targeting four key challenges to advance the practical deployment of FL in energy-centric applications. The first challenge (C1) involves model attacks, where adversarial clients attempt to compromise the global model. Existing attacks do not fully exploit FL’s vulnerabilities and can be captured with current defence frameworks. This necessitates the development of stealth attack strategies. This research introduces novel stealth model poisoning techniques, including the Federated Communication Round Attack (Fed-CRA), which increases communication rounds without degrading model performance but at the cost of higher resource consumption. These vulnerabilities highlight the need for stronger defense mechanisms, driving the development of more robust frameworks. The second challenge (C2) focuses on robust aggregation, as traditional FL methods often struggle to filter out adversarial updates effectively. To address this challenge, we introduce four innovative frameworks: (a). Federated Random Layer Aggregation (FedRLA), which enhances security by aggregating only a randomly selected layer during each round, which reduces the attack surface as attack can only attack single layer of local model thereby mitigating the impact of adversarial updates; (b). Layer-Based Anomaly Aware Federated Averaging (LBAAFedAvg), which detects and isolates compromised layers while ensuring that valid updates are preserved with the help novel clustering criteria to identify good and back clients, improving the overall integrity of the aggregation process; (c). Federated Incentivized Averaging (Fed-InA), specifically designed for Fed-CRA, which is based on game theory, it incentivizes honest clients by rewarding them and penalizes malicious ones, promoting a healthier collaborative environment; and Decentralized Federated Learning (DFL), which distributes the aggregation process across multiple clients, minimizing the risk of single points of failure and eliminate the need of server. Furthermore, (d). Decentralized Federated Random Layer Aggregation (DRLA) combines DFL with FedRLA to significantly enhance robustness against adversarial attacks by aggregating a single layer in peer to peer communication manner. The third challenge (C3) concerns communication and computational efficiency, as FL’s iterative updates can strain bandwidth and processing resources, especially in energy-constrained environments. The proposed frameworks optimize efficiency by minimizing transmitted data and computational overhead. FedRLA significantly reduces communication costs by limiting shared model information, while Adaptive Single Layer Aggregation (ASLA) leverages quantization and adaptive stopping criteria to ensure minimal resource usage. Other robust frameworks, LBAA-FedAvg, Fed-InA, and DFL, are designed to require minimal resources for model training. The fourth challenge (C4) addresses data heterogeneity, a fundamental issue in energy networks where clients possess diverse consumption patterns. Two frameworks tackle this problem: (a). FedBranched, which clusters clients based on data similarity to enhance local model convergence, and (b). ASLA, which selectively aggregates the most effective layer across clients, improving generalization across varied datasets. Through addressing these interconnected challenges, this research enhances the robustness of the model, communication and computational efficiency and fixes the issue of data heterogeneity between different clients, paving the way for more resilient and privacy-preserving intelligent systems
AI-driven operations for flexible and secure microgrids and multi-energy microgrids
No full textAbstract not currently available
On the monodromies of strata of (half-)translation surfaces
Get PDFThe present work consists of two independent parts devoted to the study of topological monodromy maps of strata of (half-)translation surfaces. As the name suggests, topological monodromies provide some important topological data for such strata (Lemma 2.4.1), considered by many authors as the loci of fundamental dynamical actions. Inspired by Calderon–Salter’s description of the images of the topological monodromy maps of non-hyperelliptic strata of translation surfaces ([Cal20], [CS21] and [CS22]), we estimated the size of the respective kernels in low genera and in particular in the case of the non-hyperelliptic strata Hodd(4) and H(3, 1) in genus 3 and Heven(6) in genus 4. We prove that the kernel contains a non-abelian free group of rank 2. In the second part, we study the topological monodromy maps of strata of half-translation surfaces, a generalisation of the concept of translation surface. Here, we improve Walker’s result [Wal09] and provide a promising candidate for the image of the topological monodromies as subgroups of the mapping class groups
The impact of preeclampsia on fetal nutrition and offspring metabolism and cardiovascular function
Get PDFThe incidence of pregnancy complications, including preeclampsia (PE), is increasing nowadays. PE is one of the main causes of maternal and fetal morbidity and mortality worldwide. PE can also affect the later-life health of both mother and baby by increasing the risk for the development of CVD, although the mechanisms for this remain unknown.
The development of an insulin-resistant state is a normal adaptation during late pregnancy which leads to increased adipose tissue lipolysis and presents a source of both short chain saturated fatty acids as an energy source and long-chain polyunsaturated fatty acids (LC-PUFAs) as a key nutrient for the fetus. LC-PUFAs are essential for fetal neural development and their supply depends on the delivery from the mother through the placenta. Mothers with PE have exacerbated levels of insulin resistance that lead to increased maternal adipose tissue lipolysis leading to maternal excess of plasma non-esterified fatty acids (NEFAs) and triglycerides (TGs) as well as reduced maternal and fetal levels of LC-PUFAs. There is preliminary evidence for ectopic fat accumulation in the PE placenta which may affect LC-PUFA transport to the baby, although the presence of placental lipid droplets is yet to be confirmed. Reduced LC-PUFA transport may also impact the cardiovascular development of the offspring and, consequently, increase the risk of later-life CVD. Therefore, it was hypothesised that PE impairs essential fatty acid transfer by the placenta to the growing fetus thereby limiting cardiac and vascular development in the offspring.
Firstly, this thesis aimed to investigate the cardiovascular impact on the offspring of pregnancies complicated by PE in a rat model of superimposed PE. Pregnant stroke-prone spontaneously hypertensive rats (SHRSP) were implanted with osmotic minipumps for the continuous delivery of 0.9 % saline (vehicle control) or 700 ng/kg/min angiotensin II (ANGII). Wistar Kyoto (WKY) rats were also implanted with vehicle minipumps and used as a normotensive control. Fetal growth restriction (FGR) was observed in the ANGII neonates which were significantly smaller than the SHRSP vehicle control neonates (birth weight: 5.11 ± 0.05 vs 6.00 ± 0.12 g; mean ± SEM, P<0.001). Phenotyping of the offspring by echocardiography and tail-cuff plethysmography between week 5 and 17 of age showed thatANGII-exposed offspring present a worsened cardiovascular phenotype compared to controls, but without an impact on their systolic blood pressure. Fractional shortening (FSwas significantly lower in the ANGII offspring compared to vehicle controls (ANGII: 48.51 ± 0.97 ± Control: 52.12 ± 0.95; mean ± SEM, P<0.001).Relative wall thickness (RWT) was also smaller in the ANGII offspring compared to vehicle controls (ANGII: 0.51 ± 0.04 vs Control: 0.58 ± 0.03 mm; mean ± SEM, P=0.024). Lower FS and RWT in the ANGII are signs of systolic dysfunction. Additionally, the early to late mitral valve flow velocity ratio (E/A) was significantly higher in the ANGII offspring compared to SHRSP vehicle control (ANGII: 1.94 ± 0.06 vs Control: 1.53 ± 0.04; mean ± SEM, P<0.001), which is evidence of diastolic dysfunction.
A lipidomics analysis of human and rat placentas was carried out to investigate the potential lipid dysregulation in PE and to further validate the ANGII animal model. Placental samples were collected at Caesarean section from women with PE and healthy pregnancies. Placental lipids were extracted and analysed via shotgun lipidomic techniques using high-resolution mass spectrometry. Lipid droplets from PE and healthy human placental cryosections were stained with Oil red O and quantitated by image analysis. TGs and cholesteryl esters (CEs)
concentrations were significantly higher in PE placentas compared to controls (TG: PE 453.6 ± 262.1 vs healthy 129.3 ± 109.4; P<0.001; CE: PE 1063 ± 1004 vs healthy 425.9 ± 432.7 nmol/g; median ±IQR, P<0.001), which is evidence of neutral lipid accumulation. PE placentas had a significantly higher lipid droplet area compared to healthy controls (5.56 ± 5.15 vs 1.94 ± 1.70 %; median ±IQR, P=0.009). Gene expression of lipid-droplet associated proteins perilipin 4 and 5 (PLIN4 and PLIN5) were also significantly higher in PE compared to healthy placentas, which further corroborated the presence of ectopic fat in the PE placenta. Lipidomic analyses of placentas from ANGII-treated dams and vehicle controls did not show any differences in lipid content between groups. Therefore, the ANGII model of superimposed PE does not mimic the elevated placental lipid storage accumulation observed in human PE pregnancies.
Gene expression of placental markers of fatty acid transport, lipolysis and synthesis were examined in PE in order to obtain further insights into impaired placental lipid metabolism. PE placentas had higher expression of genes involved in de novo lipogenesis (DGAT2, FASN and SREBF-1) compared to healthy controls which suggests that the elevated ectopic placental fat presence in PE is due to greater lipid synthesis. Elevated angiopoietin-like 4 (Angptl4) expression in the PE placenta was also observed, which may indicate decreased fatty acid uptake by the placenta through the inhibition of lipoprotein lipase (LPL), consequently contributing to the lower LC-PUFA status in PE. LIPE expression was higher in the PE placenta compared to controls, which may be a reflection of the elevated TG stored in intracellular lipid droplets. This dysregulated fatty acid homeostasis in the PE placenta may be caused by increased fatty acid supply due to severe insulin resistance in the mother.
Another condition characterised by elevated maternal insulin resistance during pregnancy, namely gestational diabetes mellitus (GDM), was used as a comparator to PE. GDM pregnancy did not demonstrate ectopic fat accumulation on the placenta. Placental neutral lipid content, PLIN expression and lipid droplet content were not different between healthy and GDM women, although several differences in other lipid species were identified and warrant further investigation.
Overall, the data presented in this thesis indicated that impaired placental fatty acid metabolism and storage may contribute to the pathology of PE. Despite some phenotypic similarities with PE, GDM did not have elevated ectopic fat storage in the placenta, which suggests excess fatty acids may be stored elsewhere. Due to the role of hyperlipidaemia and insulin resistance in the aetiology of PE, reducing lipid levels in the mother could potentially improve neonatal and offspring later-life outcomes. The offspring of a model of superimposed PE showed impaired cardiovascular development after ANGII intrauterine exposure, which suggests that PE offspring may benefit from long-term monitoring to ameliorate CVD risk
An assessment of the criminal law’s protection of privacy and reputation rights
Get PDFThis thesis seeks to provide a critical assessment of the criminal law’s protection of two fundamental interests: privacy and reputation. While protection of these interests is traditionally rooted in the civil law and there has been considerable scholarship in this context, the extent to which the criminal law upholds and protects these interests remains under-researched. This is particularly so in Scotland, where much of the criminal law continues to be part of the common law and has developed from historical principles and precedents.
This is an appropriate time at which to carry out this research. Privacy and reputation interests have become increasingly valuable and are recognised as fundamental rights, receiving protected status in international human rights instruments. These interests have nevertheless become more vulnerable to harmful conduct and subject to greater levels of interference. Technological advancements, the growth of social media, and the collection and retention of personal data all pose significant threats to privacy and reputation.
These developments have been accompanied by some legislative intervention in Scots law (e.g. the introduction of offences combatting stalking, the non-consensual distribution of intimate images, stirring up hatred). While these offences may have some bearing on privacy and reputation, there has been no principled consideration of how these interests arespecifically protected (nor how they ought to be protected) by the criminal law. Moreover, despite recent reform of defamation law in Scotland resulting in the introduction of the Defamation and Malicious Publication (Scotland) Act 2021, this did not take account of the role that may be played by the criminal law.
This research seeks to address these shortcomings by establishing a normative basis for protecting these interests through the criminal law. It argues that there is scope for the criminal law to play a greater role in regulating violations of three interests: informational privacy, physical privacy, and reputation. This argument is developed, firstly, by reference to fundamental criminalisation principles (wrongfulness, harmfulness and culpability) and, secondly, the impact that this may have in terms of conveying the seriousness of such wrongsand facilitating practical improvements for victims of privacy and reputation violations.It is concluded that this would overcome some of the existing limitations in the legal mechanisms through which these rights are protected, as identified through a detailed evaluation of relevant criminal and civil wrongs undertaken in this thesis
Polyphonic novels and binary referenda: A Bakhtinian analysis of Indyref Fiction
No full textAbstract not currently available
Investigating natural analogues for carbon dioxide removal in Scottish rock formations: insights for scaling CDR technologies
No full textAbstract not currently available
Exploring the pattern of DNA double strand breaks in glioma cancer stem cells before and after radiation
Get PDFIntroduction: Glioblastoma (GBM) is an incurable brain malignancy with a median survival of 15 months. Glioma stem cells (GSCs) are a radioresistant cell subpopulation promoting recurrence with aberrant DNA damage response (DDR). Radiotherapy prolongs survival and induces DNA double strand breaks (DSBs),however the genome-wide distribution of endogenous and ionising radiation (IR) induced DSBs remain unexplored.
Aims: This will be the first study to define the DSB distribution (‘breakome’) of GSCs and their differentiated progeny before and after radiation.
Methods: Breaks ligation in-situ sequencing (BLISS) was performed in GSCs and matched differentiated progeny before and after IR exposure and correlated with RNA-seq, ATAC-seq data and public neural cell and cancer cell BLISS datasets. High DSB density locations, long genes, transcription start and termination sites (TSS & TTS) as well as euchromatin-rich regions were mapped for DSBs. Data were validated using an alternative DSB mapping technique (INDUCE-seq).
Results: Few differences in DSB distribution between GSC and differentiated progeny were observed, however DSB frequency was higher in GSCs. MALAT1 gene had the highest DSB density across all GBM lines. Long genes showed higher absolute DSBs but not higher DSB density. GSCs showed a significantly higher accumulation of DSBs at TTS compared to neural cell lines. Highly transcribed genes showed an increase in DSB density. Both BLISS and INDUCE-seq data displayed higher DSB frequency following IR in differentiated GBM R10 cells but lower DSB frequency following IR in R10 GSCs.
Conclusions: DSBs in GSCs and differentiated progeny are not random but show distinct endogenous break patterns. Additionally, DSB frequency following IR is increased in radiosensitive R10 differentiated cells but decreased in radioresistant R10 GSCs relative to unirradiated GSCs. This is likely due to changes in cellular activities including global transcriptional downregulation and improved DDR in GSCs resulting in radioresistance
The consequences of urbanisation and artificial light at night on wildlife health and physiology: insights from avian systems
Get PDFUrbanisation is accelerating at an unprecedented rate globally, exposing urban wildlife to a suite of novel anthropogenic stressors that can significantly impact health and fitness. One of the most pervasive of these stressors is Artificial light at night (ALAN) which can impact species at multiple levels of the ecosystem through the disruption of natural light-dark cycles that are essential for the regulation of many biological and ecological processes. Despite growing awareness of these threats, our understanding of how urbanisation and specific urban stressors like ALAN affect wildlife health remains limited, as does our knowledge of the mechanisms driving these relationships. A deeper understanding is vital for developing effective mitigation strategies that balance human needs with the conservation of vulnerable species in the face of urban expansion. As urbanisation continues to intensify, assessing its effects on wildlife health and fitness becomes increasingly urgent. This requires the use of reliable health biomarkers to generate consistent and biologically meaningful evidence in ecological studies. This thesis integrates experimental, and meta-analytical approaches to provide a comprehensive assessment of how urbanisation influences physiology and health in avian study systems. It also addresses the methodological challenges of measuring health in ecological contexts, with a focus on high intraindividual variation in physiological biomarkers, which can obscure meaningful patterns.
To address these knowledge gaps, I first conducted a phylogenetically controlled global meta-analysis to evaluate the relationship between urbanisation and avian health, identifying the biological and methodological factors that influence this relationship. While some studies suggest urbanisation negatively impacts avian health, findings have been inconsistent, and this is likely due to variability in study locations, focal species, and biomarker types. By collating data across a variety of studies, I was able to identify key drivers of variation and reveal hidden patterns. Notably, urbanisation was found to negatively impact avian health only when it was measured as a continuous variable (urban score), rather than as a binary categorical variable (urban vs non-urban classification). This finding emphasises the importance of quantifying urbanisation in a way that captures its heterogeneity for the detection of subtle relationships. The results also highlight that the impacts of urbanisation on avian health are context dependent, influenced by factors such as life stage and biomarker type.
I then narrowed the focus to investigate the effects of ALAN exposure on multiple health biomarkers in birds, using both field and captive study designs. In the field, I exposed great tit (Parus major) nestlings to ALAN from hatching to fledging using artificial nest boxes, targeting the sensitive developmental period where exposure to stressors is likely to have stronger consequences. Additionally, in a captive experiment, I exposed adult zebra finches (Taeniopygia guttata) to ALAN over four months to assess its long-term physiological effects using a longitudinal design. To explore potential mitigation strategies, I also tested a partial light at night treatment in the captive experiment in which lights were turned off for half of the night, comparing its effects to those of the full night ALAN exposure.
In both experiments, I adopted a multi-biomarker approach to capture different dimensions of health, addressing its multivariate nature. In the field, ALAN exposure did not show any consistent negative effect on oxidative stress or corticosterone levels in nestlings, but it was shown to significantly reduce body condition. In the captive study, ALAN increased telomere loss and disrupted glucose circadian rhythms, although it had no detectable effects on the oxidative stress markers assessed. Importantly, these negative effects were absent in the partial light at night treatment group, suggesting that this mitigation strategy may provide a viable means of reducing the impact of ALAN on wildlife health. These findings highlight the complex and variable nature of the effects of ALAN on health and show the potential for mitigation through the use of altered light exposure regimes.
I then explored the methodological challenges involved in measuring health by focusing on oxidative stress biomarkers, which are commonly used to assess fitness and physiological condition, particularly in research focusing on ageing and ageing related diseases. Through a comprehensive meta-analysis, I demonstrated that oxidative stress markers exhibit generally low within-individual repeatability, regardless of taxa, sex, environment, biomarker type, or study design. Using a simulation study, I further showed that this low repeatability can obscure associations with other traits, including telomere length. While increasing samples sizes and the number of repeated measures per individual can help mitigate this issue, further work is needed to develop robust approaches that address these detection limitations. These findings highlight the importance of careful experimental design and cautious interpretation when using highly variable physiological biomarkers in ecological studies.
The findings of this thesis contribute to the growing body of literature looking into the effects of urbanisation and anthropogenic stressors on wildlife health and well-being. By examining these issues and considering species traits, life stages, environmental factors, and employing a multi-biomarker approach, this work advances our understanding how urban environments shape wildlife health. Furthermore, insights into the use and limitations of oxidative stress markers will inform future studies and support the development of reliable assessment of health and physiology in wild populations. This research enhances our understanding of the complex interactions between urbanisation, anthropogenic stressors, and wildlife health, providing a foundation for developing evidence-based mitigation strategies to reduce the negative effects of urban stressors such as ALAN. This research also lays the groundwork for future research on key areas, including the mechanistic pathways underlying these relationships, the cumulative long-term effects of urbanisation, and the development of realistic, scalable interventions to reduce negative impacts of urbanisation on biodiversity
21st Century energy and legitimation of new industries: Innovating for and legitimating the bioeconomy – A study of the European Green Transition
Get PDFThe literature around legitimation explores the creation and implementation of change at several levels. Associating legitimation to the green transition allows us to understand when it was created and where the implementation has occurred, or not, why, and who is responsible for it. Since legitimation is such an important topic in the academic world of business, the exploration of its association to such a current issue as the implementation of the green transition is a very relevant approach to understand it further. The connection between the green transition and legitimation is what is explored in this thesis, through the energy sector.
To do so, this research focused on identifying the historical presence of the green transition through an analysis of the most important documents and strategies affecting it. Associated to these documents are actors who create and implement them at several levels of society. Considering the transition affects all planes of society, including industry and institutions, the research tried to gather the insights from actors of said planes. Thus, regulators, innovators and integrators were defined and engaged with in order to understand how the green transition is being perceived by experts. Interviews were set up with representatives of the energy sector from higher international institutions and governments; successful, green oriented businesses; and activist or media related NGOs or institutions.
The insights from these experts showed the main factors associated with a potential delay in the green transition. Time and reaction are considered, in this thesis, as the two main factors contributing to the delay of the green transition’s energy sector. As such, the research focused on understanding the underlying events that cause such delays and whether the processes can be improved through the exposition of their framework.
Alongside the actors’ inputs, the theoretical lens of the institutional genealogy took a front seat in the observation and guidance of the train of thought in this research. Focusing on understanding the history of incumbents and their effect on allowing the current green transition to flourish, the institutional genealogical lens synergized with the profiles chosen and with the historical approach that set up the research. It gives a historical categorization, via policy proxies, of the events that marked the green transition and how the current events contrast with older ones.
Therefore, this research tries to look at the green transition’s energy sector through the eyes of the current actors’ insights while keeping in mind the incumbents that generated the current issues and which are still present in the transition. Without attributing any fault, this thesis attempts to show the current progress of the green transition’s energy sector and uncover potential factors that affect its speed. As a final contribution to knowledge, the thesis provides a tentative framework of the influential actors and their relationship with legitimation regarding the legitimation of the green transition through the energy sector
Data sources:
Access Repository Dashboard
Do you manage Open Research Online? Become a CORE Member to access insider analytics, issue reports and manage access to outputs from your repository in the CORE Repository Dashboard! 👇