21684 research outputs found
Sort by
A novel privacy-preserving data sharing system based on attributed-based encryption and zero knowledge proof
Get PDFThe exponential growth of digital data across various sectors, such as healthcare, finance, and e-commerce, has underscored critical concerns regarding data privacy, security, and ownership. Centralised data storage systems are inherently vulnerable to cyber-attacks, raising significant privacy risks and compliance challenges, despite regulatory frameworks like the General Data Protection Regulation (GDPR). This research introduces a decentralised, privacy-preserving data-sharing framework leveraging blockchain technology, Ciphertext-Policy Attribute-Based Encryption (CP-ABE), and Zero-Knowledge Proofs (ZKP).
By employing CP-ABE, the proposed system enables fine-grained access control, ensuring that only authorised entities can access sensitive data based on specified attributes. The integration of Zero-Knowledge Proofs preserves user privacy by allowing verification of access rights without revealing the underlying attributes. The system architecture is underpinned by decentralised storage, with smart contracts managing secure access verification.
Performance evaluations demonstrate that the system effectively handles dynamic policies and attribute sets, demonstrating its adaptability to real-world applications. This framework represents a significant advancement in privacy-preserving data-sharing technologies, offering a scalable and secure solution for safeguarding sensitive users’ attributes in decentralised environments
Federated learning for next generation intelligent applications
Get PDFThe rapid proliferation of smart devices and Internet of Things (IoT) technologies has revolutionised data collection for artificial intelligence (AI)-driven applications, enabling rapid training and near real-time inference. However, the traditional centralised learning approaches require transferring vast amounts of raw data from end devices to a central server. This process leads to substantial network overhead, increased latency, and significant privacy concerns, hindering the scalability and responsiveness of intelligent applications. This thesis exploits federated learning (FL) as a distributed, on-device learning framework that enables collaborative model training without raw data sharing. The distributed architecture of FL offers privacy by design and reduces communication costs by exchanging the model parameters that align with principles of data sovereignty and regulatory compliance. Despite its advantages, FL faces significant challenges in real-world applications, and this thesis aims to address the following three critical challenges: C1) data diversity; C2) robust aggregation ensuring privacy and security in the training process; and finally, C3) energy efficiency. The first contribution introduces the similarity-driven truncated aggregation (SDTA) framework, designed to tackle challenges C1 and C2. SDTA measures the similarity among the model updates to identify and filter the anomalous updates, mitigating the impact of attacks and overfitting without accessing client data. Additionally, it incorporates differential privacy (DP) to strengthen training privacy. The second contribution introduces the semantic-aware federated blockage prediction (SFBP) framework, addressing challenges C1 and C3. Using multi-modal fusion and a lightweight computer vision model for edge-based semantic extraction, the proposed framework reduces communication costs and inference delays while maintaining high prediction accuracy. Additionally, SFBP incorporates a filter mechanism to minimise the effects of noisy or adversarial updates. The third contribution addresses C1 and C3 and develops a hybrid neuromorphic federated learning (HNFL) framework for outdoor human activity recognition (HAR) using wearable sensors. The proposed spiking-long short-term memory (S-LSTM) model combines the energy-efficient spiking neural networks with the sequential data handling strengths of LSTM networks. This approach improves the accuracy while ensuring data privacy and reducing computational costs, making it suitable for deployment on resource-constrained edge devices. Finally, to address challenges C2 and C3, the federated fusion quantisation (FFQ) framework is proposed to improve HAR models in indoor settings. FFQ combines FL with edge-based preprocessing, feature engineering, and model compression to achieve a low false positive rate, essential for applications like fall detection. A customised FedDist algorithm is used for global model aggregation, effectively reducing overfitting in diverse data. Additionally, FFQ applies model compression and quantisation-aware training to lower communication overhead without compromising accuracy. These contributions advance FL by enhancing scalability, robustness, and efficiency, paving the way for next-generation intelligent systems
The genetic and cellular basis of viviparity
Get PDFReproductive mode is an important topic in evolutionary biology, as the evolution of viviparity is a major evolutionary transition that has evolved repeatedly in different animal groups. Squamate reptiles are recognised as excellent models for the evolution of viviparity, with over 100 such transitions known from this class of animals, representing a broad range of phenotypes on the oviparity-viviparity continuum, particularly in the case of reproductively bimodal species such as the Eurasian common lizard Zootoca vivipara. In this thesis, I present the results of a series of experiments building on and expanding the existing body of work using Z. vivipara as a model system. In Chapter 1 I present a review of the current state of the literature on the evolution of oviparity-viviparity transitions in squamate reptiles across every major squamate group, and discuss the status of Z. vivipara as an emerging model organism while summarising previous work on reproductive mode in this and other squamate species. In Chapters 2-3 I present the results of two new sequencing experiments designed to investigate the uterine transcriptomic changes which accompany pregnancy in viviparous Z. vivipara, describing changes in gene expression and alternative-splicing of genes before, during and after pregnancy, as well as exploring gene expression of known viviparity-related genes at the cellular level. In Chapter 4 I then present a new model for the investigation of reproductive mode in the form of cultured Z. vivipara oviduct cells, and characterise and evaluate this system as a potential tool for future research. Finally in Chapter 5 I discuss the potential for these results to enable a new program of research for the functional validation of candidate genes linked to viviparity, in a new reverse-genetic paradigm for reproductive mode research
The effects of adipose-derived regenerative cells in preclinical kidney injury models
Get PDFChronic kidney disease is a common health issue affecting approximately 10% of the UK population. Advanced kidney failure necessitates kidney transplantation or dialysis in affected patients. The optimal treatment for end-stage kidney failure is kidney transplantation. Although the number of deceased donor kidney transplants has increased, the number of patients waiting for kidney transplantation continues to rise. Therefore, to make more kidneys available for transplantation, it is important to improve kidney transplantation techniques by using effective perfusate solutions, developing methods for organ storage, and applying additional treatments to the transplant kidney to optimise outcomes.
Adipose-derived regenerative cells (ADRCs) are a heterogeneous cell population derived from adipose tissue. These cells are composed of endothelial progenitor cells, pericytes, and adipose stromal cells. Previous studies have revealed the potential therapeutic effects of ADRCs. These effects might help kidney grafts recover from ischaemia-reperfusion injury in kidney transplantation. My project aims to examine the therapeutic effects of ADRCs on injured kidney models to assess their potential utility for optimising kidneys for transplantation. I used three approaches for investigation: 1) a systematic review to investigate the effect of mesenchymal stem cells (MSCs) on solid organ transplantation models; 2) testing kidney organoids as a complex structural model for examining the effects of ADRCs on cellular communication and organoid protection; and 3) proximal tubular cell culture to study the therapeutic effects of ADRCs. Since oxidative stress is a hallmark of injury in ischemia-reperfusion injury (IRI), hydrogen peroxide was used as an injury inducer to mimic kidney transplantation by increasing oxidative stress in both the kidney organoid model and the proximal tubular cell model.
A systematic review was conducted using the search engine program Rayyan. Data were collected from 1991 to 2020 using specific terms. Two researchers assessed these data for validity. The results showed that MSCs have anti-apoptotic, anti-fibrotic, tubular injury protection, and anti-oxidative stress properties. These findings reveal the effects of MSCs on organ transplantation in the preclinical and clinical literature to date.
In studies of ADRCs on injured kidney organoid models, embryonic kidneys at days 13.5–14.5 were collected and processed into cell pellets. These pellets were cultured for 14 days for organogenesis. Organoids were assigned to four groups for study: non-injury, injury, ADRC-treated non-injury, and ADRC-treated injury groups. Injured organoids were induced using 10 µM hydrogen peroxide for 60 minutes, and ADRCs at 10% of embryonic cells were used in the ADRC-treated groups. After assignment, organoids were cultured for 14 days before collection and downstream assays. The results showed organoid protection by maintenance of organoid size after injury, revealing the pro-survival properties of ADRCs.
In proximal tubular cell culture experiments, NRK-52E cells were used and assigned to four groups (similar to the previous study). These cells were incubated with 500 µM hydrogen peroxide for 4 hours, and ADRCs were added at a 10% ratio of total NRK-52E cells. Laboratory assays including trypan blue exclusion test, western blot, real-time polymerase chain reaction (RT-PCR), flow cytometry, caspase-3 activity assay, and nanoparticle tracking analysis were performed after 24 hours of ADRC treatment. The results showed that ADRCs tended to decrease cell death markers and increase live cell markers in the ADRC-treated injured group and decreased NGAL gene expression. These findings reveal tubular protection from injury by ADRCs in injured proximal tubular cells.
This thesis shows that ADRCs had positive effects in organ transplant models in systematic reviews, exhibited pro-survival effects on kidney organoids, and reduced kidney injury markers in NRK-52E cells. Therefore, ADRCs are a promising additional treatment to diminish IRI in transplant kidneys
Effectively cornering new physics at colliders and beyond
Get PDFThe Standard Model has achieved remarkable success, yet growing empirical and theoretical tensions point to the need for new physics at the TeV scale. As collider experiments enter a precision era, even small deviations from Standard Model predictions can provide crucial clues about the underlying structure of fundamental interactions. This thesis explores these possibilities through both model-dependent studies and the effective field theory approach. Multi-Higgs production is used to probe extended scalar sectors, offering insight into the nature of electroweak symmetry breaking and the dynamics of possible phase transitions. Electroweak-scale triplet models are examined through collider signatures and flavour constraints, presenting a realistic mechanism for radiative neutrino mass generation. In the top-quark sector, momentum-dependent width effects are implemented in a gauge-consistent way, leading to more accurate predictions for SMEFT constraints. To address the challenge of high-multiplicity final states, machine learning techniques, including graph neural networks, are applied to identify hidden correlations and enhance signal sensitivity. Together, these studies sharpen existing bounds and provide complementary strategies to guide future experimental efforts at the High-Luminosity LHC and beyond
1,1-dithiolate ligands in coordination chemistry: a study of their electronic and molecular structures
Get PDFDithio-ligands are known for their versatility in coordinating to various metal centres. Among them, 1,1- and 1,2-dithiolates stand out for their extended conjugated backbone and unusual redox properties upon coordination. While early studies have shown clear differences in bonding and redox behaviour between the two systems, 1,1-dithiolates remain far less explored.
A series of new NiII and CuII bis-1,1-dithiolate complexes were prepared and compared alongside other homoleptic dithio-complexes. The 1,1-dithiolate ligand scope was expanded to the unprecedented para-R- phenylacetonitrile dithiolate R-padt2− derivatives (R = OMe, H, Cl, meta-CF3, para-CF3, CN, NO2), enabling a systematic variation of the electron-donating and electron-withdrawing substituents on the backbone. All complexes were characterised by ligand-specific bands in their IR spectra and chemical shifts in their 1H/13C NMR spectra. In their EPR spectra, the characteristic profile of a CuS4 centre was observed and their electronic absorption spectroscopy was characterised by ligand-to-ligand charge transfer (LLCT) as well as ligand-field (LF) transitions. Additionally, a reversible oxidation process was observed during cyclic voltammetry experiments and was proved to be formally metal-centred. The square-planar structure was confirmed by X-ray crystallography, and DFT calculations provided further insight into their MO energy levels and bonding. By using the Hammett and pKa values, predictable substituent effects were confirmed across the series, creating a correlation between the molecular and electronic structure of these complexes. Oxidation of the CuII 1,1-dithiolate complexes stabilised the formal CuIII state and the potentials were directly affected by the substituent effects of the backbone groups of each 1,1-dithiolate.
The final part of the thesis extended to studying the chemistry of 1,1-dithiolate ligands in heteroleptic NiII and CuII complexes containing α-diimines and the non-conjugated N-donor ligand tetramethyl ethylenediamine (TMEDA). The NiII 1,1-dithiolate–diimine complexes exhibited distinct LLCT transitions containing the HOMO and LUMO, as well as diimine-based reductions. The energy of the LLCT and the potentials of the reductions showed correlations with the Hammett and pKa values corresponding to the α-diimine and 1,1-dithiolate coordinated ligands. Electron-withdrawing substituents on the α-diimine ligands stabilised the π* orbitals, leading to red-shifted LLCT transitions and more positive reduction potentials, whereas electron-donating groups had the opposite effect
Integrating pre-trained language models into novel query expansion pipelines
Get PDFIn our increasingly digital society, proficiency in finding valuable and useful information is crucial within everyday personal and professional life. Information Retrieval (IR) is the academic f ield that focuses on discovering useful information (documents) that fulfil a user’s information need (query). In particular, search systems process a user query and return a ranked list of documents determined by the query-specific relevance calculated by a retrieval model. Retrieval models have traditionally been based on query-document term overlap; however, dense embedding models are becoming increasingly prevalent with the emergence of Pre-Trained Language Models (PLMs).
Lexical mismatch is a classic problem within information retrieval, whereby a user query fails to capture their complete information need, leading to retrieval models failing to find relevant documents. A common approach for this issue is query expansion, involving the augmentation of the query with supplementary information to enhance the retrieval of relevant documents. This process is usually done automatically through pseudo-relevance feedback (PRF), where a set of documents from a first-pass retrieval algorithm are assumed relevant, and are used to expand the query with useful context. This approach has proven beneficial for both sparse and dense retrieval models. In this thesis, I hypothesise that integrating PLMs into multi-stage query expansion pipelines can improve performance over current sparse and dense expansion methods. Leveraging the capabilities of PLMs to generate and rank relevant content to build better expansion models, which should particularly help queries that require reasoning or contextualisation.
This thesis examines existing retrieval and expansion models, identifying their shortcomings to focus the contributions. This includes constructing a formal definition of complex queries and constructing new datasets, such as CODEC, designed to evaluate the effectiveness of our proposed retrieval models on this query type. I first show that by simply using PLMs to re-rank our first-pass candidate set of documents before query expansion improves sparse and dense retrieval by 5–8%. This motivates the development of a new fine-grained expansion model, Latent Entity Expansion (LEE), that achieves a further 2-8% gain in NDCG by explicitly modelling knowledge using terms and entities. Furthermore, I introduce a novel expansion pipeline, termed “adaptive expansion”, that iterates between retrieving new batches of documents and updating the expansion model through PLM re-ranking. Adaptive expansion leads to state-of-the-art effectiveness gains without requiring any additional re-ranking computation.
This thesis’s second central research thread explores not using pseudo-relevance feedback at all; instead, leveraging the generative capabilities of PLMs to build our query expansion models directly. I introduce Generative Relevance Feedback (GRF), which shows that sparse expansion using PLM-generated content improves MAP between 5-19% over traditional PRF. I also show that GRF is highly effective when combined with dense and learned sparse retrieval. Furthermore, I increase the retrieval effectiveness of these pipelines by incorporating Generative Relevance Modelling (GRM) to mitigate hallucination by scaling the weight of generated documents in our expansion model. We propose Relevance-Aware Sample Estimation (RASE) to ground the generated documents to the target corpus and use PLMs to estimate relevance.
Overall, this body of work demonstrates the potential of query expansion when combined with novel pipelines that leverage the capabilities of PLMs. This paradigm shift provides a foundation for future advancements, promising more efficient and effective search systems
Advancing drought understanding and prediction in the Vietnamese Mekong Delta
Get PDFDrought, one of the most destructive climate related natural hazards, affects millions of people worldwide and poses substantial challenges in the Vietnamese Mekong Delta (VMD), one of Southeast Asia’s largest deltas. In recent decades, particularly during 1991 1994, 1998, 2005, 2010, 2015 2016, and 2019 2020, the VMD has suffered from severe and prolonged droughts that resulted in significant socioeconomic impacts. In this delta, droughts often result in severe clean water shortages and extensive damage to cropland. Despite these profound impacts, the mechanisms driving these droughts, including anomalies in the atmospheric moisture transport and land atmosphere (LA) interactions, and the prediction of droughts in the VMD remain underexplored. Addressing these gaps is crucial for enhancing drought preparedness and developing effective drought mitigation strategies.
Accordingly, this thesis aims to achieve three primary objectives : 1) to elucidate the sources of precipitation moisture and identify the dominant factors influencing these sources during drought periods in the VMD ; 2) to quantitatively assess the LA interactions in the VMD using advanced deep learning techniques; 3) to develop an accurate deep learning model capable of predicting droughts in the VMD on account of atmospheric conditions from the external precipitation source region. The first two objectives are designed to deepen understanding of the mechanisms and processes driving droughts in the VMD, while the third aims to utilize these insights to provide accurate drought predictions.
To better understand the process es of atmospheric moisture transport, the Water Accounting Model-2layers (WAM-2layers), an Eulerian based moisture tracking model, was employed to identify the primary moisture sources of precipitation in the VMD from 1980 to 2020. In addition, for the first time, the causal inference algorithms were introduced to analyze the causal relationships among variables involved in moisture transport, specifically, to identify which factor drives the moisture transport process and dominates the amount of tracked moisture. The analysis revealed that (1) over 60% of precipitation in the VMD originates from external moisture sources (60.4 93. 3%), with local recycling contributing from 1.2 % to 27.1%; (2) seasonal shifts in monsoon patterns strongly influence the origins of moisture: during the dry season, the South China Sea (northeast) serves as the dominant source, while the Bay of Bengal (southwest) becomes the primary origin during the wet season; (3) based on the causal inference algorithms, atmospheric humidity and wind speed in the upwind area were identified as the principal factors influencing moisture transport during dry and wet seasons, respectively; (4) large scale forcings (e.g., El Niño and La Niña) were found to affect the processes of affect the processes of moisture transport significantly and these effects vary spatially and seasonally across the VMD’s precipitationshed; (5) local atmospheric conditions, including atmospheric instability (e.g., convective available potential energy, CAPE) and local atmospheric humidity, also play a crucial role in modulating moisture recycling efficiency.
As for the interactions among LA variables, the Long- and Short-term Time-series Network (LSTNet) was applied to model these dynamics over the VMD. The key findings are as as followsfollows: (1) the LSTNet model demonstrated superior performance compared to the traditional regional climate model in simulating key variables (i.e., precipitation, soil precipitation, soil moisture, sensiblemoisture, sensible and and latent heat) during both dry and wet seasons. It exhibited higher accuracy and lower bias, underscoring its suitability for modeling LA interactions in the VMD; (2) this deep learning model effectively captured the relative importance of key variables within the LA interactions, highlighting the critical roles of soil moisture and sensible heat, particularly during dry periods when their negative anomalies substantially reduce precipitation. For example, anomalies in sensible heat were found to decrease precipitation by up to 20% during dry periods, primarily through interactions with temperature and convective inhibition (CIN). Similarly, soil moisture strongly influences precipitation in both dry and wet periods, with deficits leading to reductions in precipitation of up to 30%; (3) projected declines in soil moisture coupled with increases in sensible heat are expected to exacerbate precipitation deficits under changing climatic conditions. By 2075-2099, a 10% increase in sensible heat could reduce precipitation by 3.76% in dry seasons.
Finally, exploring the utility of atmospheric conditions from external precipitation source regions, the deep neural network, Convolutional Gated Recurrent Unit (ConvGRU) was developed to enhance accuracy in drought prediction. The ConvGRU model exhibited exceptional performance in predicting drought conditions at a 3-month lead time, which successfully predicts approximately 90% of meteorological drought events and about 80% of agagricultural drought events, with ricultural drought events, with fewer than 10% false predictions for drought months and events. Furthermore, ConvGRU predicts about 70% and 80% compound dry-hot months and events, respectively. The outstanding performance of the ConvGRU model in drought prediction at the 3-month lead time largely attributed to the delayed impacts of external atmospheric conditions, including specific humidity and U- and V-wind, on the VMD’s drought conditions through the water vapor transport process. Incorporating the atmospheric data from these external precipitation source regions significantly improvess the ConvGRU model’s ’s predictive capability, particularly at the lead time of 3 months.
In summary, this research not only advances the understanding of mechanisms driving drought dynamics including external atmospheric moisture transport and local LA interactions, but also establishes an innovative, effective model for drought prediction. These research developments are vital for improving drought resilience and adaptability in the VMD, and offering substantial benefits for regional drought management strategies
Factors influencing healthcare workers’ adherence to infection prevention and control in Saudi Arabia
Get PDFBackground: Infection Prevention and Control (IPC) practices have a unique role in reducing the risk of healthcare-associated infections in healthcare settings. However, adherence to IPC practices, including standard precautions, remains suboptimal among healthcare workers. In Saudi Arabia and across the Middle East, research into IPC is growing, especially following the increased global emphasis on infection control due to the COVID-19 pandemic. However, despite this growing interest, there is still a limited understanding of the factors contributing to IPC adherence among healthcare workers, particularly in intensive care units (ICUs) and medical wards. Therefore, the current study aims to address this gap by conducting a multi-method qualitative study to explore the factors affecting adherence to IPC practices among healthcare workers in ICUs and medical wards in Saudi Arabia.
Methods: The research comprised three phases and the empirical work was conducted in two selected hospitals within Saudi Arabia. A qualitative systematic review was first conducted to explore the factors influencing IPC adherence among healthcare workers in Middle Eastern countries (Phase 1). This was followed by two qualitative studies, designed to capture perspectives on IPC practices and potential factors influencing adherence to IPC practices from different perspectives. The first perspective was that of workers (n = 8) who worked in infection control teams in two hospitals in Saudi Arabia, sought through focus groups (Phase 2). This was followed by Phase 3 in which individual semi-structured interviews were conducted to seek perspectives on IPC practices, and their barriers and facilitators in practice, through the lens of healthcare workers (n = 20) delivering hands-on care and employed within ICU and medical wards of the two hospitals.
Findings: Phase 1 of the current study identified organisational and individual factors influencing adherence to IPC practices. Individual factors, including moral principles, ethical beliefs, and cultural habits, played a significant role in promoting IPC adherence. Organisational factors, including leadership, training gaps, and environmental challenges were also perceived to affect adherence. Phase 2 further explored the role of the infection control team in monitoring adherence and providing education and training on IPC practices. The findings from Phase 2 revealed the infection control teams’ perceptions of the main challenges associated with IPC adherence. These challenges included perceived differences in adherence among professional groups and across various components of IPC practices; staff stability; and the nature of each department, including its procedures and the acuity of patients. Phase 3 further supported the findings from the second phase. It highlighted poor leadership and managerial support, and the need for more training, and for involving all healthcare workers as well as patients and their relatives in this training, as major challenges that affected adherence to IPC practices. Overall, the study showed a notable increase in awareness of IPC following the emergence of COVID-19. It also highlighted the role of cultural and social factors in IPC adherence, along with persistent hierarchical challenges within the healthcare system.
Conclusion and implications: This study highlights the importance of organisational support for healthcare workers as well as improving the monitoring strategies in Saudi Arabia. The study recommends enhancing the involvement of family in IPC practices and fostering a supportive working environment through recognition and team-building initiatives. It also emphasises the development of culturally sensitive IPC policies, as well as the establishment of recognition programmes for IPC leaders. Addressing staffing issues and improving the physical work environment are also crucial for maintaining IPC practices. For future research, it is important to explore the impact of culturally sensitive IPC interventions, engaging family members in IPC education, and comparing IPC adherence across various healthcare settings and disciplines to gain a comprehensive understanding of IPC practices and to improve overall adherence
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! 👇