Search the digital library catalog

Abstract: With the increased complexity of applications and systems, threat modelling struggles to keep pace with the evolution of risks. This article addresses this challenge by exploring how large language models (LLMs) can be leveraged to create comprehensive threat models across different risk assessment methodologies. We examine whether a single generic prompt can support frameworks such as LINDDUN, PASTA, and STRIDE, despite their different requirements. Through this comparative analysis, we identify components that enable AI-based assessments, while acknowledging that privacy, regulatory, and dynamic risks require adaptation of the frameworks. Our findings show that a universal guideline is feasible for broad applications, but adaptation is necessary for effective use. Overall, LLM-based threat modelling improves the accessibility, repeatability, and effectiveness of risk analysis and supports stronger and more sustainable practices.
Keywords: cybersecurity, large language models, threat modelling, sustainability, resilient infrastructure, SDG 9
Published in DKUM: 02.12.2025; Views: 0; Downloads: 2
Full text (875,62 KB)

Abstract: Aim To evaluate and compare human-led and artificial intelligence-automated critical appraisal of evidence. Background Critical appraisal is essential in evidence-based practice, yet many nurses lack the skills to perform it. Large language models offer potential support, but their role in critical appraisal remains underexplored. Design We conducted a comparative study to evaluate the performance of five commonly used large language models versus two human reviewers in appraising four systematic reviews on interventions to reduce medication administration errors. Methods We compared large language models and two human reviewers in independently appraising four systematic reviews using the JBI Critical Appraisal Checklist. These models were Perplexity Sonar (Pro), Claude 3.7 Sonnet, Gemini 2.0 Flash, GPT-4.5 and Grok-2. All models received identical full texts and standardized prompts. Responses were analyzed descriptively and agreement was assessed using Cohen’s Kappa. Results Large language models showed full agreement with human reviewers on five of 11 JBI items. Most disagreements occurred in appraising search strategy, inclusion criteria and publication bias. The agreement between human reviewers and large language models ranged from slight to moderate. The highest level of agreement was observed with Claude (κ = 0.732), while the lowest level was observed with Gemini (κ = 0.394). Conclusion Large language models can support aspects of critical appraisal evidence but lack contextual reasoning and methodological insight required for complex judgments. While Claude 3.7 Sonnet aligned most closely with human reviewers, human oversight remains essential. Large language models should serve as adjuncts and not substitutes for evidence-based practice.
Keywords: artificial intelligence in healthcare, multimodal large language models, nursing, evidence-based practice
Published in DKUM: 12.11.2025; Views: 0; Downloads: 1
Link to file

Abstract: This thesis examines the success rate of large language models (LLM) in detecting macOS malware through Endpoint Security logs. A literature review and 144 experiments with three ChatGPT variants and six prompt types evaluated accuracy, precision, recall, specificity, and F1-score. Results show that prompt wording is crucial: zero-shot and chain-of-thought prompts performed best, while conservative prompts minimized false positives but missed threats. GPT-4o and o1 outperformed o4-mini but showed similar results. Findings suggest LLMs can support, but not replace, traditional detection, with prompt design proving as important as model choice.
Keywords: malware, large language models, detection
Published in DKUM: 03.11.2025; Views: 0; Downloads: 6
Full text (29,72 MB)

Abstract: The automotive industry is increasingly confronted with challenges in managing complex requirements and test cases arising from the integration of advanced electronic systems, software functionalities, and compliance with international standards. Conventional manual validation of requirements is time-consuming, error-prone, and resource-intensive, underscoring the need for more efficient and reliable approaches. This thesis investigates the automation of test case generation through the application of Retrieval-Augmented Generation (RAG) in combination with Large Language Models (LLMs). A complete RAG workflow was implemented in Python, incorporating LangChain, LangGraph, Ollama, and ChromaDB to facilitate indexing, retrieval, and generation. The system was trained and evaluated on datasets comprising automotive requirements and test cases, with experiments examining embedding quality, retrieval strategies, prompt engineering techniques, and generative model parameters. The results demonstrate that RAG is capable of generating high-quality, contextually relevant test cases on consumer-grade hardware, thereby significantly enhancing efficiency, consistency, and productivity relative to manual methods. Furthermore, the findings suggest that RAG-based systems are best positioned as complementary tools that support, rather than replace, human engineers. This research provides a foundation for future work on hybrid retrieval methods, advanced embedding techniques, and the integration of more powerful LLMs into requirement and test case management processes.
Keywords: automotive requirements validation, test case generation, large Language Models, Retrieval-Augmented Generation
Published in DKUM: 01.10.2025; Views: 0; Downloads: 0
Full text (5,84 MB)

Abstract: Cellulose-based nanomaterials are fascinating renewable biosystems, yet low thermal conductivity and dielectric permittivity often limit their potential applications in flexible electronics. We report dielectric properties of composite films prepared by vacuum filtration or solvent casting method from the native (CNF) or carboxylated (TCNF) cellulose nanofibrils and high electrically and thermally conductive 2D titanium carbide (Ti3C2Tx) MXenes. Measurements over broad frequency and temperature ranges revealed the influence of preparation method and type of nanofibrils matrix on the overall dielectric response, as well as a notable impact of absorbed water, particularly on the cellulose’s secondary β and γ relaxations. A detailed investigation of material with the lowest amount of impurities, vacuum-filtered MXene/CNF composites, confirmed that the dielectric response follows the predictions of the percolation theory. The resulting strong enhancement of the dielectric permittivity on increasing MXene content demonstrates the potential of developed composites for applications in eco-friendly dielectric and piezoelectric devices.
Keywords: percolation theory, piezoelectric devices, dielectric properties, nanomaterials, carbides, biological systems, biological models, carbohydrates
Published in DKUM: 08.09.2025; Views: 0; Downloads: 3
Full text (10,11 MB)
This document has many files! More...