Despite technological advancements in cybersecurity, phishing emails remain a primary vector for cyberattacks due to their sophisticated exploitation of human psychology. While existing research has examined phishing susceptibility, studies have typically relied on small scale datasets and text analysis methods, limiting our understanding of how these attacks evolve and target specific cultural contexts. This paper addresses this gaps of knowledge by investigating phishing emails targeting the Dutch population through an analysis of linguistic patterns, thematic content, and psychological manipulation techniques. Using a dataset of 8,000 phishing emails, we employ advanced natural language processing techniques, including BERTopic modelling and frequency analysis. Our findings reveal that phishing emails predominantly utilize business and financial terminology, emphasising essential Dutch services. Our thematic analysis of the email contents identifies eight distinct narratives, demonstrating the attackers' detailed understanding of Dutch institutional frameworks. The analysis of psychological triggers shows a sophisticated evolution beyond simple urgency-based tactics, combining elements of authority, trust, and urgency. These insights contribute to understanding modern phishing tactics and suggest the need for improved, contextually aware security training approaches.
Video generation models have opened new opportunities for simulating business processes through realistic visualizations. However, current video generation approaches often fall short of capturing the inherent dynamics and structure of business processes and tend to produce inconsistent simulations that lack the rigor provided by formal process models. To address these limitations, we introduce a novel method termed Petri Net structure-driven video generation, which integrates the inherent structural information from process models to tailor video simulations more closely to the dynamics of business processes. We explore multiple strategies for this tailoring, including i) the use of domain knowledge-rich prompting, ii) a storyboard employing image references extracted from process evidence data, and iii) generated image references informed by process models. We evaluate our method across diverse domains, and demonstrate that the Petri Net structure-driven approach improves the perceived usefulness and consistency of the simulated video, marking a step forward in the use of generative AI for more realistic business process simulation.
Data spaces have become a strategic pillar of Europe's digital agenda, enabling sovereign, legally compliant data sharing within decentralized ecosystems. As data space initiatives evolve, personalized recommendations are increasingly recognized as key use cases. However, traditional recommendation approaches typically rely on centralized aggregation of user behavior data-directly conflicting with the core ethos of data spaces: sovereignty, privacy, and trust. Federated recommendation systems offer a promising alternative by training models locally and exchanging only intermediate parameters to build a global model. Despite this potential, the integration of federated recommendation techniques and data space architectures remains largely underexplored in research and practice. This paper addresses this gap by designing and evaluating a prototype of a federated recommendation system specifically tailored for data spaces and compliant with their underlying infrastructure. Our findings highlight the viability of developing privacy-preserving, collaborative recommendation systems within data spaces, and contribute to the broader adoption of AI across these emerging ecosystems.
As data-driven applications gain traction in smart living environments, ensuring high data quality becomes a critical prerequisite for reliable analytics and Artificial Intelligence (AI)based services. However, due to the wide range of hardware and software providers as well as the variety of use cases in the smart living domain, the resulting data are characterized by highly heterogeneous structures and nesting levels. The diversity of systems, their strong reliance on context, and the rapid growth of Internet of Things (IoT) devices make it difficult to ensure consistency, accuracy, and reliability in data quality assessments within smart living environments. This paper addresses the need for a consistent assessment of data quality within heterogeneous smart living datasets. We present a prototype that generates quality reports for datasets with an unknown structure based on established data quality metrics A Large Language Model component generates structured metadata describing the dataset's structure and suggesting applicable quality checks, which then serve as input for an implemented code base to perform standardized quality checks. This approach not only supports the assessment of data quality within individual organizations but also facilitates cross-organizational assessments, enabling better comparability and evaluation of datasets from different sources or providers. The prototype was evaluated using a variety of synthetic and real-world smart living datasets. The results demonstrate the feasibility of the proposed approach, although certain limitations remain, which are discussed in detail within the paper.
