Lund University Publications
Not a member yet
244244 research outputs found
Sort by
Natural language processing (NLP)-based frameworks for cyber threat intelligence and early prediction of cyberattacks in Industry 4.0 : a systematic literature review
No full textThis study provides a systematic overview of Natural Language Processing (NLP)-based frameworks for Cyber Threat Intelligence (CTI) and the early prediction of cyberattacks in Industry 4.0. As digital transformation accelerates through the integration of IoT, SCADA, and cyber-physical systems, manufacturing environments face an expanding and complex cyber threat landscape. Following the PRISMA 2020 systematic review protocol, 80 peer-reviewed studies published between 2015 and 2025 were analyzed across IEEE Xplore, Scopus, and Web of Science to identify methods that employ NLP for CTI extraction, reasoning, and predictive modelling. The review finds that transformer-based architectures, knowledge graph reasoning, and social media mining are increasingly used to convert unstructured data into actionable intelligence, thereby enabling earlier detection and forecasting of cyber threats. Large Language Models (LLMs) demonstrate strong potential for anticipating attack sequences, while domain-specific models enhance industrial relevance. Persistent challenges include data scarcity, domain adaptation, explainability, and real-time scalability in operational-technology environments. The review concludes that NLP is reshaping Industry 4.0 cybersecurity from reactive defense toward predictive, adaptive, and intelligence-driven protection, and it highlights the need for interpretable, domain-specific, and resource-efficient frameworks to secure Industry 4.0 ecosystems
Tumor Signatures of Physical Fitness : Insights from a Preclinical Model
No full textPurpose: Cardiorespiratory fitness (CRF) and muscle strength are associated with cancer risk/mortality in adults. However, there is yet no evidence for pediatric tumors. This study investigated the association of CRF and muscle strength with several tumor-related phenotypes in an aggressive childhood malignancy, high-risk neuroblastoma (HR-NB). Methods: Twelve mice bearing orthotopic HR-NB were studied. CRF and muscle strength were assessed using treadmill and grip strength testing, respectively. The following tumor-related outcomes were studied: survival, clinical severity, tumor weight/volume, metastasis, and intratumor immune infiltrates. Additionally, tumor samples underwent quantitative proteomic analysis via liquid chromatography-tandem mass spectrometry. Spearman correlations (or logistic regression) were performed between CRF/muscle strength and the abovementioned variables. Proteins significantly correlated with CRF or muscle strength were mapped into protein–protein interaction (PPI) networks using the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) database. Results: CRF was inversely correlated with clinical severity score (r=–0.657, p=0.020). Of 6,840 identified tumor proteins, 76 correlated significantly with CRF (19 positively, 57 negatively), whereas 194 correlated with muscle strength (97 positively, 97 negatively). Proteins correlated with CRF were primarily involved in metabolic and structural pathways, including angiotensinogen and elastin. In turn, muscle strength-associated proteins were more abundant, and included keratin family proteins (e.g., keratin, type I cytoskeletal 14 and type II cytoskeletal 5), proteins involved in cell adhesion (e.g., desmoglein-1-alpha), and translational regulators (e.g., eukaryotic initiation factor 4A). Network analysis revealed significant enrichment in structural organization and cellular adhesion pathways. Conclusions: Besides the association of CRF with clinical severity of the tumor, distinct novel tumor proteomic signatures associated with CRF and muscle strength were identified, highlighting potential mechanisms linking physical fitness with childhood cancer biology
Numerical studies of strong protein interactions
No full textProtein–protein interactions are fundamental to biological functions as well as the design of new functional food products and biomedicines. However, they can be challenging to computationally quantify due to strong forces and sensitivity to pronounced solution conditions. This thesis combines computational methods with Monte Carlo simulations to investigate strong protein interactions in diverse systems.Paper I uses Monte Carlo simulations to study heteroprotein complex coacervation between the milk proteins lactoferrin and β−lactoglobulin. Both implicit and explicit salt simulations are performed. The explicit ones with the help of parallel tempering, where agreement with experimental observations is obtained. A residue–residue contact analysis at the free energy minimum is performed, where key interaction sites on the proteins are identified.Paper II investigates the thermodynamic consequences of structural variability in haemoglobins from six mammalian species. One- and two-body Monte Carlo simulations quantify electrostatic properties and second osmotic virial coefficients, revealing differences inanisotropy and interaction strength.Paper III introduces a general computational framework for evaluating two-body partition functions and quantifying respective thermodynamic properties. By discretising relative orientations on a quasi-regular angular grid, the method efficiently captures anisotropic interactions, agrees with Monte Carlo results for a benchmarking system, and highlights limitations of a current coarse-grained force field when off-centre charge distributions are important.Conclusively, this work contributes to a better understanding of protein–protein interactions, provides tools for accurately computing thermodynamic properties of anisotropic biomolecules, and demonstrates applications ranging from food proteins to haemoglobins
Challenges, risks and costs of food sharing. Insights from a systematic literature review
No full textCurrent food systems are unsustainable, exacerbating environmental and socio-economic challenges. With 80 % of food consumed in cities, transforming urban food systems is vital. Urban food sharing initiatives (FSIs), which involve collective acts of growing, cooking, eating, and redistributing food, offer a promising alternative in sustainable urban transformations. However, FSIs often struggle to develop and scale. Academic literature has not adequately examined the factors hindering FSIs' establishment, operation, and scaling up. This study fills this gap through a systematic literature review of 55 articles, providing a comprehensive analysis of the challenges, risks, and costs faced by FSIs. It identifies six categories of challenges and risks each, as well as seven categories of costs. The study proposes nine mitigation strategies to address these issues and emphasises the need for a more detailed breakdown of costs and investments across different phases of FSI development. Additionally, it highlights the importance of support from governments, policymakers, health authorities, retailers, and charity organisations. The transformative potential of FSIs in addressing inefficiencies in current food systems is also underscored, alongside the need for a deeper understanding of the multi-dimensional challenges they face. The findings are particularly valuable for urban FSIs seeking to develop their practices, build capacity, and contribute to more resilient and sustainable urban food systems
ECG-Based Detection of Acute Myocardial Infarction Using a Wrist-Worn Device
No full textBackground: A wrist-worn wearable device for acquiring limb and chest ECG leads (wECG) may constitute a promising approach to detection of acute myocardial infarction (AMI). However, it remains to be demonstrated whether the information conveyed by the wECG is sufficient for AMI detection. Objective: To explore explainable machine learning models for detecting AMI using the wECG. Methods: Two types of machine learning models are explored: a convolutional neural network (CNN) using the raw ECG as input and a gradient-boosting decision tree (GBDT) using clinically informative features. 123 participants were included, divided into patients with AMI, patients with other cardiovascular diseases, and healthy individuals. A wrist-worn device equipped with three biopotential electrodes was used to acquire two ECG leads with a single touch: limb lead I and another lead involving a specific body site, i.e., either the V3 or V5 electrode positions, or the abdomen. Results: The best performance on the test dataset is obtained using models that incorporate all four leads. The CNN model performs slightly better than the GBDT model, with a sensitivity of 0.77 and specificity of 0.75 compared to 0.77 and 0.72, respectively. When distinguishing between AMI and healthy participants, the specificity increases to 0.94 for the CNN model and 0.90 for the GBDT model. Feature importance analysis shows that the GBDT model primarily relies on the J point, while the CNN model primarily relies on the QRS complex. Conclusions: wECG-based AMI detection shows considerable promise in out-of-hospital settings. However, caution is needed as CNN explanations rarely agree with the ECG intervals typically analyzed in clinical practice
Cross-scale convergence in the carbon balance of managed boreal forests in Northern Sweden
No full textBoreal forests are globally important carbon (C) sinks, but strategies for maximising their climate benefit remain under debate. Major uncertainties in this discussion arise from contrasting sink-source estimates, which largely emanate from inherent limitations of standard measurement techniques to distinct spatio-temporal scales. Here, we use a spatially-nested measurement framework that integrates bottom-up (forest-plot inventory and chamber-based fluxes) and top-down (eddy-covariance; atmospheric observations and atmospheric transport modelling) approaches to reconcile the C balance of actively managed boreal forests in Northern Sweden across plot-, ecosystem-, landscape-, and regional scales during 2016–2018. We found that 3-year mean estimates of the net ecosystem production (NEP) across plot-, landscape-, and regional scales did not differ significantly, converging into a mean (± 95 % confidence interval) C sink of 118 ± 27 g C m-2 yr-1. We also noted a convergence across these scales for the 3-means of the NEP components, i.e., gross primary production (908 ± 48 g C m-2 yr-1) and ecosystem respiration (790 ± 40 g C m-2 yr-1). However, estimates of the inter-annual variations in NEP and its components were inconsistent among most scales and measurement approaches. Furthermore, our results indicate a scale-dependency in the NEP response to the 2018 European summer drought, with a greater reduction of NEP observed in bottom-up compared to top-down estimates. Thus, this study consolidates the C sink-strength of managed boreal forests and advocates the need for cross-scale assessments to constrain forest C cycle-climate feedbacks
Core and periphery in the EU legal space
No full textThis chapter aims to establish a conceptual framework for analyzing the EU legal space and its core-periphery dynamics, notably the extension of EU legal frameworks and policies beyond the territorial borders of the Union. It investigates how the EU legal space is defined, how its cores and peripheries are structured, and how they interact, evolve, and influence each other. The chapter argues that the EU legal space is not homogeneous but consists of multiple legal, geographical, and policy-based layers. It explores how legal norms extend beyond the EU’s borders, the role of third countries in norm adoption, and the fluidity of core-periphery relationships in shaping European legal integration
