Publikationer från Linköpings universitet
Not a member yet
58266 research outputs found
Sort by
Analysis of Direct S3 Access: Authentication, Data Isolation, and Performance
No full textI takt med att molnbaserade lösningar blir alltmer integrerade i företags IT-infrastruktur uppstår nya krav på säker och effektiv datahantering. Detta examensarbete undersöker hur filöverföringsarkitekturen i ett affärssystem kan förbättras genom att ersätta centraliserad serverbaserad kommunikation med direktkommunikation mellan klientapplikationer och Amazon S3. Målet är att minska serverbelastning, förbättra nätverksprestanda och samtidigt upprätthålla hög säkerhet och dataisolering. Två metoder för direkt åtkomst har implementerats och utvärderats: användning av presigned URL för objektspecifik åtkomst och temporära säkerhetstoken via AWS Security Token Service (STS), där klienter tilldelas IAM-roller för bredare och mer dynamisk åtkomstkontroll. Prestandatester visar att presigned URL-metoden konsekvent ger kortare överföringstider, särskilt under nätverkspåverkan som latens och paketförlust. Multipart-presigned URL för uppladdning presterade dock avsevärt sämre, vilket i studien kopplas till användning av fast segmentstorlek, avsaknad av parallellisering och högt antal serveranrop. Dessa faktorer identifieras som områden som kan förbättras i framtida arbete. STS erbjuder å andra sidan mer flexibla och säkra åtkomstkontroller, men med något längre överföringstider. Studien visar även att direkt S3-kommunikation minskar trafikbelastningen på serversidan avsevärt, vilket öppnar för mer kostnadseffektiva och skalbara system. Arbetet ger praktiska riktlinjer för val av åtkomstmetod baserat på applikationens krav på säkerhet, prestanda, nätverksmiljö och systemarkitektur.As cloud-based solutions become increasingly integrated into enterprise IT infrastructures, new requirements arise for secure and efficient data management. This thesis investigates how the file transfer architecture in a business system can be improved by replacing centralized, server-based communication with direct communication between client applications and Amazon S3. The goal is to reduce server load, improve network performance, and at the same time maintain high levels of security and data isolation. Two methods for direct access have been implemented and evaluated: the use of presigned URLs for object-specific access, and temporary security tokens via AWS Security Token Service (STS), where clients are assigned IAM roles for broader and more dynamic access control. Performance tests show that the presigned URL method consistently yields shorter transfer times, particularly under network conditions affected by latency and packet loss. However, multipart uploads using presigned URLs performed significantly worse, which in the study is linked to the use of fixed segment sizes, the lack of parallelization, and a high number of server calls. These factors are identified as areas for improvement in future work. STS, on the other hand, offers more flexible and secure access control, though with somewhat longer transfer times. The study also shows that direct communication with S3 significantly reduces traffic load on the server side, which opens up possibilities for more cost-effective and scalable systems. The work provides practical guidelines for choosing an access method based on the application's requirements for security, performance, network environment, and system architecture
Shaping the market of tomorrow : Developing a market-shaping strategy to accelerate the transition to more sustainable plant-based food production
No full textMot bakgrund av klimatförändringar, global befolkningstillväxt och ett alltmer instabilt geopolitiskt läge har behovet av en omställning mot en mer resilient och hållbar livsmedelsproduktion blivit alltmer framträdande. Trots ökad medvetenhet hos konsumenterna kvarstår ett gap mellan intention och konsumtion, vilket tyder på att marknaden ännu inte lyckats erbjuda tillräckligt attraktiva alternativ. För att adressera denna utmaning undersöks begreppet marknadsformning, en strategi för att förändra marknadsstrukturer, normer och beteenden. Syftet med studien är att undersöka hur livsmedelsföretag kan använda marknadsformande strategier för att driva en omställning mot en mer hållbar och växtbaserad livsmedelsproduktion. För att besvara syftet används ett svenskt företag inom växtbaserade proteiner som fallföretag. Genom att analysera hur företaget kan forma marknadsdynamiker och konsumentbeteenden utifrån sin nuvarande marknadssituation genereras empiriska insikter som kompletterar befintlig teori om marknadsformning. Studiens referensram utgår från tre faser i den marknadsformande processen: initiering, mobilisering och bevarande. Inom dessa faser identifieras åtta marknadsformande element: värdeerbjudande, affärsmodell, vision, marknadsnätverk, marknadsidentitet, normer, resiliens och barriärer. Dessa komponenter integreras i en analysmodell som utgör grunden för att besvara studiens syfte. Utifrån analysmodellen är två preciserade frågeställningar formulerade: PFS1: Vilka konsument- och leverantörsbehov måste tillgodoses för att driva en ökad efterfrågan på växtbaserade proteiner? PFS2: Vilka marknadsformande aktiviteter är viktiga för att forma marknaden för livsmedel? Den kvalitativa empiriska insamlingen utgörs av 18 semistrukturerade intervjuer med nyckelaktörer i värdekedjan för växtbaserade proteiner. Analysen visar att marknadsformning är en komplex och dynamisk process som kräver att företag anpassar sina strategier utifrån sju identifierade behov inom marknaden. För att möta dessa behov identifieras 35 marknadsformande aktiviteter fördelade över sex marknadsformande element inom faserna för initiering och mobilisering. Studiens slutsats är att livsmedelsföretag kan driva omställningen mot en mer hållbar och växtbaserad livsmedelsproduktion genom att genomföra marknadsformande aktiviteter som bemöter marknadens behov. Baserat på slutsatsen är två rekommendationer framtagna. Den första rekommendationen är att ta hänsyn till samtliga sju identifierade behov. Den andra rekommendationen är att genomföra de marknadsformande aktiviteter som bedöms vara mest ändamålsenliga för att möta respektive behov.In the context of climate change, population growth, and an increasingly fragile geopolitical environment, the need for a shift towards more resilient and sustainable food production has become increasingly prominent. Despite rising consumer awareness, a gap between intention and consumption persists, suggesting that the market has not yet succeeded in offering sufficiently attractive alternatives. To address this challenge, the concept of market-shaping, a strategy aimed to changing market structures, norms, and behaviours, is explored. The purpose of this study is to identify how food companies can employ market-shaping strategies to drive the transition towards more sustainable and plant-based food production. To fulfil this objective, a Swedish plant-based protein company is used as a case company. By analysing how the company can shape market dynamics and consumer behaviour based on its current market situation, empirical insights are generated that complement existing theory on market-shaping. The study’s framework is based on three phases of the market-shaping process: initiation, mobilization, and maintenance. Within these phases, eight market-shaping elements are identified: value proposition, business model, vision, market network, market identity, norms, resilience, and barriers. These components are integrated into an analytical model that forms the foundation for addressing the study’s objective. Based on this model, two research questions are formulated: RQ1: What consumer and supplier needs must be addressed to drive increased demand for plant-based proteins? RQ2: What market-shaping activities are important to shape the food industry? The qualitative empirical data collection consists of 18 semi-structured interviews with key actors in the plant-based protein value chain. The analysis reveals that market-shaping is a complex and dynamic process requiring companies to continuously to adapt their strategies based on seven identified needs within the market. To meet these needs, 35 market-shaping activities are identified, distributed across six market-shaping elements within the phases of initiation and mobilization. The study concludes that food companies can play a key role in driving the transition toward more sustainable, plant-based food production by engaging in market-shaping activities that respond to specific market needs. Based on this conclusion, two recommendations are offered. First, companies should consider all seven identified market needs. Second, they should implement the market-shaping activities best suited to address each of these needs effectively
Assessing Cookstove Emissions in Lagos (Nigeria) Using Low-Cost Sensors
No full textHousehold air pollution (HAP) from traditional cookstoves remains a significant public health challenge in Lagos, Nigeria, particularly among vulnerable urban households. However, critical gaps persist, including inadequate longitudinal data on household exposure patterns. socioeconomic barriers, variability across kitchen / fuel types, and limited assessments of agespecific respiratory risks, limits adopting cleaner cookstoves. Therefore, this study employed low-cost sensors across 33 households in the Ijaiye and Ahmadiya district of Lagos, Nigeria to monitor particulate matter (PM1, PM2.5) and gaseous pollutants (CO, CO2, NO2, SO2) during cooking events. Additionally, 55 socioeconomic surveys were conducted to identify factors hindering the adoption of sustainable cookstoves. The data were analyzed using MATLAB and statistical tools, followed by Multi-Path Particle Deposition (MPPD) modeling to assess respiratory deposition across children, teenagers, and adults. The results present an unexpectedinsight: LPG is not free from emissions, and no significant differences were observed among LPG, firewood, and charcoal. This challenges common assumptions and offers a unique sensorbased perspective, raising important questions about fuel sustainability and energy equity. Moreover, there was no significant difference in PM2.5 deposition rates among adults, children, and teenagers, indicating similar respiratory exposure risks. Additionally, socioeconomic analysis found occupation and financial constraints were not significant barriers to cookstove adoption, suggesting other factors beyond traditional constraints more strongly influence sustainable cookstove usage in the studied communities. These findings provide evidence for implementing targeted policies promoting affordable and cleaner cooking technologies and improved air quality in urban households, bridging the gap between HAP research and public health policy
Heritage, Modernisation, & Urban Imaginaries: A Comparative Case Study of Stockholm & Norrkoping
No full textThis thesis investigates how urban imaginaries shape heritage conservation and urban development in Stockholm and Norrköping. Through comparative analysis of planning documents Such as architectural policy, building regulations, and comprehensive plans along with interviews, it explores how different visions of the city influence planning practices and policy implementation. The study finds that Stockholm reflects a formalist conservation model, characterised by visual regulation, monumentality, and symbolic coherence. In contrast, Norrköping takes a more flexible and creative approach, using its industrial heritage in new ways through reuse and cultural projects. While both cities focus on sustainability and identity, there are clear tensions between inclusive planning and the reality of decisions being made by experts. Voices from migrants, women, and local community groups are mostly missing from the documents. Framed by urban imaginaries and critical heritage studies, the thesis argues that imaginaries function as instruments of technocratic governance. Heritage becomes not just a site of preservation but a tool for narrating and negotiating urban futures. Keywords: Urban imaginaries, heritage, adaptive reuse, conservation models, Stockholm, Norrköping, planning policy.
Comparative Analysis Between AI-Generated Text and Human-Written Text : An Analysis Using Shannon Entropy
No full textStudien undersöker skillnader mellan AI-genererad text och text skriven av människa med hjälp av shannonentropi, ett mått från informationsteorin som kvantifierar osäkerhet i språk. Genom att analysera 57 texter skrivna av lärarstudenter samt 57 motsvarande texter genererade av språkmodellen Gemini 2.5 Pro, har bland annat normaliserad shannonentropi beräknats för att minimera felkällorna som kan tillkomma med olika textlängder. Resultatet visar en statistiskt signifikant skillnad i entropivärden mellan de två texttyperna, där mänskligt skrivna texter uppvisar högre entropi och större variation i språkbruk. Studien visar att shannonentropi är ett potentiellt verktyg för att särskilja mellan AI-genererade och mänskligt producerade texter. Vidare forskning krävs för att generalisera resultatet till andra typer av text samt andra språkmodeller
Traffic Classification in LTE Networks : Leveraging Machine Learning on Unencrypted Control Traffic
No full textThis thesis, conducted at the Swedish Defence Research Agency (FOI), investigates the feasibility of classifying user activities—such as music and video streaming, website browsing, video calls, and social networking—in a Long Term Evolution/4G mobile network by analyzing Downlink Control Information (DCI) messages. A dataset was generated through controlled experiments, capturing DCI data during various application scenarios. The collected data was used to create plots illustrating important characteristics and differences between traffic cases; these visualizations guided the feature selection process by highlighting which features best separated the application types. From the DCI messages, statistical and traffic-based features were extracted based on transport block size and allocation timings. Feature selection and analysis revealed that certain statistical descriptors, such as skewness and kurtosis, had limited impact on model performance and were excluded. Four machine learning models—K-Nearest Neighbors, Logistic Regression, Support Vector Machine (SVM), and Random Forest—were trained and evaluated. The results showed that SVM and Random Forest achieved the highest performance, with accuracies of 0.92 and 0.94, respectively, due to their ability to model non-linear class boundaries. In contrast, K-NN and Logistic Regression struggled, particularly with overlapping classes like music streaming with Spotify and social media usage. Additionally, the impact of training set size on classification performance was analyzed. The experiments showed that all models maintain good average performance even with limited data, but certain classes, such as Spotify traffic, suffer significantly when the amount of training data is reduced. The thesis concludes that user activities can be effectively identified from DCI data using supervised learning, provided that the model is capable of handling complex feature interactions
Användningen av artificiell intelligens i svenska kommuner : En analys av kommunala tjänstepersoners resonemang och förhållningssätt
No full textArtificiell intelligens (AI) har på senare tid fått en allt mer framträdande roll inom den svenska offentliga sektorn och beskrivs ofta som ett verktyg för att öka effektiviteten och främja innovation i kommunal förvaltning. Samtidigt väcker AI viktiga frågor om rättssäkerhet, transparens och demokratiska värden, vilket gör ämnet särskilt relevant ur ett statsvetenskapligt perspektiv. Denna studie undersöker hur kommunala tjänstepersoner reflekterar över AI:s nuvarande och framtida användning inom kommunal verksamhet, med särskilt fokus på vilka offentliga värden som lyfts fram och hur tjänstepersoner förhåller sig till de utmaningar och möjligheter som AI innebär. Studien utgår från ett nyinstitutionalistiskt analytiskt ramverk och bygger på empiriskt material insamlat genom fokusgruppsintervjuer med kommunala tjänstepersoner, främst utvecklingschefer och digitaliseringsstrateger från olika svenska kommuner. Analysen visar att frågor om legalitet och efterlevnad av lagstiftning är centrala för deltagarna. Det råder en utbredd osäkerhet kring juridisk kompetens och institutionella strukturer för AI, vilket bidrar till en försiktig hållning till användningen av AI i beslut som direkt påverkar medborgare. Skepsis mot AI-drivna beslut bottnar främst i upplevd brist på legitimitet, transparens och robusta valideringsprocesser. Samtidigt identifierar deltagarna en stor potential för AI i administrativa stödprocesser och effektiviseringsarbete, där användningen uppfattas som mer förenlig med etablerade institutionella logiker och värden relaterade till effektivitet och rationalitet. Svenska kommuner befinner sig således i en försiktig position, där ambitioner att använda AI ofta begränsas av organisatoriska och normativa begränsningar. Trots nationella riktlinjer och stöd kvarstår osäkerhet kring datadelning, transparens och opartiskhet. Sammanfattningsvis visar studien att informanterna uppvisar en stark innovationsvilja, samtidigt som analysen tydliggör att AI-användningen fortfarande begränsas av institutionell osäkerhet och en återhållsam attityd till AI i kommunal verksamhet
Methodology for BRDF Measurements in Outdoor Conditions with Solar Illumination
No full textUnderstanding and modeling the reflectance properties of materials is essential in various fields, for example during the development and evaluation of camouflage materials. An important concept for this purpose is the Bidirectional Reflectance Distribution Function, BRDF, which describes how light is reflectedfrom a surface as a function of both the incident and viewing angles. While BRDF is typically measured in indoor laboratories, there is significant value indeveloping methods for outdoor measurements under natural illumination, particularly for studying large or context specific samples. However, this introduces challenges in managing environmental variables. The aim of this project was to design a first useful methodology for measuring BRDF using the sun as the light source. A theoretical model was proposed to isolate the direct sunlight component by comparing radiometric measurements of both reference and sample materials under sunlit and shaded conditions. Measurements were performed with a high dynamic range camera sensitive to the visible spectrum, mounted onto an adjustable boom to alter the zenith angle. Images of the sample materials and the reference material were captured and the data was then used to calculate the BRDF of the different sample materials. Ingoing and outgoing angles were determined using a solar position algorithm program together with a mechanical angle measurement device on the boom. The results showed strong agreement between outdoor BRDF measurements and indoor lab-oratory measurements, performed with a 633 nm laser scatterometer, with only minor deviations. A limitation of the project was the boom that only allowed measurements within a fixed plane, which did not align with the incident plane of the sunlight. As a result, the measurement predominantly captured diffuse reflection, limiting characterization of specular materials. For future work, a drone based system flying in a hemispherical pattern could enable a broader angular sampling, allowing for detailed BRDF mapping of more specular or inhomogeneous materials in three dimensions
Mathematical Modeling and Simulation of Glucagon-Insulin Dynamics for Dual-Hormone Closed-Loop Systems in Type 1 Diabetes
No full textType 1 diabetes is a chronic disease that is becoming increasingly common worldwide, especially among children and young adults. Despite technological advances, maintaining stable blood glucose levels remains a major challenge, including preventing hypoglycemia. This thesis presents the development of a mathematical model of glucose dynamics, which was used to simulate and design a dual-hormone closed-loop control system for individuals with type 1 diabetes. Traditional insulin-only systems lack the ability to actively prevent hypoglycemia, making bi-hormonal approaches that combines insulin and glucagon a promising alternative. The aim of this work was to model the dynamic effects of insulin and glucagon on blood glucose regulation, and to use this model to develop and evaluate various dual-hormone control algorithms. A modified version of the glucose model by Herrero et al. was implemented in Python using the SUND toolbox. Key improvements included reducing the number of individual-specific parameters and removing time-window dependencies to enhance model robustness. The model was validated using clinical data and evaluated through uncertainty analysis. Multiple control systems were developed, including a derivative-based algorithm with meal bolus support, and a biologically inspired -cell algorithm. Simulation results showed that the dual-hormone systems improved glucose stability and reduced hypoglycemic episodes compared to single-hormone approaches. The derivative-based controller with meal bolus resulted in the best performance compared to the other systems. These results demonstrate that mathematical modeling can be a powerful tool for designing and evaluating closed-loop glucose control systems, and that dual-hormone systems have strong potential to improve glucose stability and support safer and more autonomous diabetes management.
Metodstudie för Återskapning av Tidsserier för Maskinsystem
No full textTime-series reconstruction aims to infer sequential data from limited or degraded inputs, with broad relevance across scientific, industrial, and data-driven domains. Legacy machine systems, such as wheel loaders, are equipped with hardware that stores data in coarse categories without accurate measurements and sequential order. Potential downstream use cases and future innovations in the field require a level of data accuracy currently not achieved. Machine learning (ML) offers a potential avenue for time-series reconstruction by estimating high-resolution sequences from such coarse inputs, without requiring replacement of the underlying foundations of machine systems, thereby reducing industry costs. Previous work lacks broad analysis in time-series reconstruction from non-sequential data in machine systems. This thesis investigates the feasibility of reconstructing high-resolution time-series data from coarse, non-sequential cell-based machine log data using a deterministic and a generative neural network model. Specifically, a recurrent neural network (RNN) and a conditional variational autoencoder (VAE) were implemented and compared to a baseline that samples values uniformly within predefined interval bins. The results demonstrated that both RNN and VAE models struggled to outperform the baseline in multiple quantitative metrics, including log-likelihood. Although the ML models produced smoother sequences, these outputs did not align with the expected distributions. Based on these observations, there is little to no justification for applying ML models of this nature to the given reconstruction problem, highlighting key limitations in applying neural network models for vector-to-sequential reconstruction tasks. Major methodological refinements or hybrid approaches are needed to achieve plausible and reliable results for this problem