arXiv subjects

Stefan Decker

Publications and source records attributed to Stefan Decker.

At least 19 recordsLinked to original sources

arXiv2026

THGFM: Dual-Branch Temporal Heterogeneous Graph Fusion Model

Temporal heterogeneous graphs offer a natural abstraction for dynamic relational systems in which diverse node and relation types co-exist and evolve over time. Learning on such graphs requires jointly modeling cross-type structural heterogeneity and the temporal dynamics of interactions, yet existing methods still struggle to reconcile parameter-efficient cross-type transfer with relation-aware specialization, and typically inject time only as additive features outside the attention kernel. We propose \textbf{THGFM}, a web-scale temporal heterogeneous graph fusion model that addresses both limitations within a unified dual-path architecture. THGFM couples a \textit{Shared-Space Temporal Attention} branch for parameter-efficient cross-type transfer with a \textit{Relational Type-Partitioned Temporal Attention} branch for relation-aware specialization, and integrates them through \textit{Dual-Path Relational--Shared Fusion}, instantiated with \textit{Type-Conditioned Non-Competitive Gated Sum Fusion}: a adaptive mechanism that assigns independent, type-conditioned feature-wise gates to the shared and specialized branches, allowing both to be amplified or suppressed without zero-sum competition. To directly incorporate relative time into the attention score, THGFM further introduces \textit{Rotary Temporal Attention}, which rotates queries and keys by half-phases of relative time before matching. THGFM consistently outperforms baseline graph transformer models on academic graphs benchmarks, delivering a $+3.25\%$ six-task mean gain, with peak relative gains of $+12.37\%$ on OAG-CS PV, $+4.87\%$ on PF-$L_2$, and $+1.18\%$ on PF-$L_1$, and $+4.24\%$, $+3.73\%$, and $+4.61\%$ on OGBN-MAG, HTAG-ArXiv, and HTAG-DBLP, respectively.

cs.LG↗

arXiv2026

TELLER: Dual-Path Iterative Preference Optimization for Table Entity Linking

Entity linking in tables matches short and ambiguous cell mentions to their corresponding knowledge-base entities. Existing approaches typically rely on data preprocessing pipelines that retain either compact or extensive table content as contextual evidence, and then formulate entity linking as a language generation task for instruction-tuned models; recent systems further incorporate explicit reasoning to disambiguate challenging mentions. However, their training supervision is usually static: fixed preference data cannot adapt to the residual errors of an evolving model, while variations in reasoning length can bias sequence-level preference learning. To address these limitations, we present TELLER: Table Entity Linking through Learning from Errors and Reasoning. We first retrieve and rank Wikidata candidates and retain reduced table evidence in the prompt. The direct-answer path applies iterative direct preference optimization and refreshes its preference data with residual errors from the updated model. The reasoning path uses filtered and compressed chain-of-thought rationales for supervised fine-tuning, followed by our iterative length-normalized regularized preference optimization. On the TableInstruct entity-linking subset, the direct-answer path improves accuracy from 94.35\% to 94.50\%; on the MammoTab V2 evaluation set, it improves accuracy from 87.59\% to 88.20\%. The reasoning path improves accuracy from 92.90\% to 92.95\% on TableInstruct and from 79.09\% to 81.85\% on MammoTab V2, while maintaining high rates of complete reasoning generation. These results show that iterative preference learning benefits both concise entity prediction and explicit reasoning.

cs.CL↗

arXiv2026

What Does ODRL Mean? A Cross-Level Ontological Grounding of Permissions, Prohibitions, and Duties in UFO-L

ODRL policy evaluators produce verdicts, but say nothing about the normative positions a policy brings into existence, the authority structures those positions presuppose, or who holds the power to declare a norm violated. We formulate the Cross-Level Design Principle: any normative language with violable, consequential norms requires both conduct-level positions (Permission, Duty, Right, No right) and competence-level positions (Power, Subjection, Immunity, Disability). Applying this to ODRL, we establish that prohibition is sanctioned (violation possible and consequential), that permission is underspecified across its behaviour parameter (open vs. closed world), and that the formal semantics covers achievement obligations only. We ground ODRL in UFO-L, mapping each activated rule to a simple legal relator and extending coverage from two to eight legal positions; violation-declaration authority, implicit in every existing evaluator, becomes an explicit Power-Subjection pair. All axioms are mechanically verified in Isabelle/HOL and across a 39-problem benchmark under Vampire, E, and Z3.

cs.LO↗

arXiv2026

Sort-Stratified Semantics for Temporal Conflict Detection in ODRL Policies

In the Open Digital Rights Language (ODRL), temporal constraints range over two sorts, instants and durations, but the comparison operators do not distinguish them. The same operator thus means "earlier instant" or "shorter duration," leaving conflict detection between two policies unsound. We resolve this by sort stratification: each temporal operand is typed to one of two ordered domains, points in time or amounts of time. Each constraint then denotes an interval, and conflict reduces to interval comparison under a three-valued verdict (Conflict, Compatible, Unknown). We characterise the check's decidability across a static and a runtime fragment, prove it sound, and evaluate it on a benchmark of policy problems compiled to TPTP and SMT-LIB, available as an artefact.

cs.LO↗

arXiv2026

Data Architectures and their Technical Requirements (DATER)

Modern organizations generate and consume massive volumes of heterogeneous data at high speed. This requires a continuous development of new techniques for more efficient and reliable data management. Designing appropriate data architectures has therefore become a strategic necessity, as they shape how data is integrated, governed, and made available for analytics and decisionmaking. This paper introduces a conceptual framework - Data Architectures and their Technical Requirements (DATER) - to systematically describe and evaluate data architectures based on technical requirements. Six modern architectures are examined: data warehouse, (semantic) data lake, data lakehouse, data fabric, and data mesh. Each is analyzed by historical context, defining features, and conformance to DATER dimensions. The study supports researchers and practitioners in navigating architectural paradigms, clarifying overlaps, and highlighting strengths, limitations, and use-case suitability.

cs.DB↗

arXiv2026

Axis-Aligned Semantics for ODRL: Resolving Dimensional Ambiguity in Policy Constraints

The Open Digital Rights Language (ODRL) represents policy constraints as triples of a left operand, an operator, and a value. Several spatial operands, however, range over multi-axis domains such as width, height, and depth, while the constraint syntax provides no explicit axis identity. As a result, policy engines cannot determine whether multiple constraints apply to the same axis or different ones, making conflict detection unsound or incomplete. We resolve this ambiguity by axis decomposition, replacing multi-axis operands with axis-specific scalar operands over totally ordered domains. Each constraint then denotes an interval per axis and each policy an axis-aligned box, reducing conflict detection to box comparison. We define a three-valued semantics (Conflict, Compatible, Unknown), prove the decomposition sound and backward compatible with ODRL, instantiate it as ODRL Axis-Aligned Profile (OAAP), and validate it on a benchmark of 256 ODRL policy problems, each expressed in Turtle and compiled to first-order (TPTP) and SMT-LIB form, using Vampire, E, Z3, and cvc5.

cs.CL↗

arXiv2025

Unconsciously Forget: Mitigating Memorization; Without Knowing What is being Memorized

Recent advances in generative models have demonstrated an exceptional ability to produce highly realistic images. However, previous studies show that generated images often resemble the training data, and this problem becomes more severe as the model size increases. Memorizing training data can lead to legal challenges, including copyright infringement, violations of portrait rights, and trademark violations. Existing approaches to mitigating memorization mainly focus on manipulating the denoising sampling process to steer image embeddings away from the memorized embedding space or employ unlearning methods that require training on datasets containing specific sets of memorized concepts. However, existing methods often incur substantial computational overhead during sampling, or focus narrowly on removing one or more groups of target concepts, imposing a significant limitation on their scalability. To understand and mitigate these problems, our work, UniForget, offers a new perspective on understanding the root cause of memorization. Our work demonstrates that specific parts of the model are responsible for copyrighted content generation. By applying model pruning, we can effectively suppress the probability of generating copyrighted content without targeting specific concepts while preserving the general generative capabilities of the model. Additionally, we show that our approach is both orthogonal and complementary to existing unlearning methods, thereby highlighting its potential to improve current unlearning and de-memorization techniques.

cs.CV↗

arXiv2025

From Legacy to Standard: LLM-Assisted Transformation of Cybersecurity Playbooks into CACAO Format

Existing cybersecurity playbooks are often written in heterogeneous, non-machine-readable formats, which limits their automation and interoperability across Security Orchestration, Automation, and Response platforms. This paper explores the suitability of Large Language Models, combined with Prompt Engineering, to automatically translate legacy incident response playbooks into the standardized, machine-readable CACAO format. We systematically examine various Prompt Engineering techniques and carefully design prompts aimed at maximizing syntactic accuracy and semantic fidelity for control flow preservation. Our modular transformation pipeline integrates a syntax checker to ensure syntactic correctness and features an iterative refinement mechanism that progressively reduces syntactic errors. We evaluate the proposed approach on a custom-generated dataset comprising diverse legacy playbooks paired with manually created CACAO references. The results demonstrate that our method significantly improves the accuracy of playbook transformation over baseline models, effectively captures complex workflow structures, and substantially reduces errors. It highlights the potential for practical deployment in automated cybersecurity playbook transformation tasks.

cs.CR↗

arXiv2025

A Service Architecture for Dataspaces

Dataspaces are designed to support sovereign, trusted and decentralized data exchange between participants forming an ecosystem. They are standardized by initiatives such as the International Data Spaces Association or Gaia-X and have gained adoption in several domains such as mobility, manufacturing, tourism or culture. In dataspaces, participants use connectors to communicate peer-to-peer. The Eclipse Dataspace Components (EDC) Connector is a broadly adopted, open-source implementation that adheres to the standards and is supported by a large community. As dataspaces in general, it focuses on the exchange of data assets with associated usage policies and does not support services. In practice, however, there is demand for dataspace-based services and conceptual arguments support their inclusion in dataspaces. In this paper, we propose an abstraction layer for providing generic services within dataspaces. Adopters can use this layer to easily develop own services, seamlessly integrated with the existing dataspace technology. Besides, we present an initial implementation of this service architecture for the EDC Connector and demonstrate its practical applicability.

cs.DB↗

arXiv2025

From Instructions to ODRL Usage Policies: An Ontology Guided Approach

This study presents an approach that uses large language models such as GPT-4 to generate usage policies in the W3C Open Digital Rights Language ODRL automatically from natural language instructions. Our approach uses the ODRL ontology and its documentation as a central part of the prompt. Our research hypothesis is that a curated version of existing ontology documentation will better guide policy generation. We present various heuristics for adapting the ODRL ontology and its documentation to guide an end-to-end KG construction process. We evaluate our approach in the context of dataspaces, i.e., distributed infrastructures for trustworthy data exchange between multiple participating organizations for the cultural domain. We created a benchmark consisting of 12 use cases of varying complexity. Our evaluation shows excellent results with up to 91.95% accuracy in the resulting knowledge graph.

cs.CL↗

arXiv2025

LogicAD: Explainable Anomaly Detection via VLM-based Text Feature Extraction

Logical image understanding involves interpreting and reasoning about the relationships and consistency within an image's visual content. This capability is essential in applications such as industrial inspection, where logical anomaly detection is critical for maintaining high-quality standards and minimizing costly recalls. Previous research in anomaly detection (AD) has relied on prior knowledge for designing algorithms, which often requires extensive manual annotations, significant computing power, and large amounts of data for training. Autoregressive, multimodal Vision Language Models (AVLMs) offer a promising alternative due to their exceptional performance in visual reasoning across various domains. Despite this, their application to logical AD remains unexplored. In this work, we investigate using AVLMs for logical AD and demonstrate that they are well-suited to the task. Combining AVLMs with format embedding and a logic reasoner, we achieve SOTA performance on public benchmarks, MVTec LOCO AD, with an AUROC of 86.0% and F1-max of 83.7%, along with explanations of anomalies. This significantly outperforms the existing SOTA method by a large margin.

cs.CV↗

arXiv2024

Encryption-Aware Anomaly Detection in Power Grid Communication Networks

The shift to smart grids has made electrical power systems more vulnerable to sophisticated cyber threats. To protect these systems, holistic security measures that encompass preventive, detective, and reactive components are required, even with encrypted data. However, traditional intrusion detection methods struggle with encrypted traffic, our research focuses on the low-level communication layers of encrypted power grid systems to identify irregular patterns using statistics and machine learning. Our results indicate that a harmonic security concept based on encrypted traffic and anomaly detection is promising for smart grid security; however, further research is necessary to improve detection accuracy.

cs.CR↗

arXiv2024

PASTA-4-PHT: A Pipeline for Automated Security and Technical Audits for the Personal Health Train

With the introduction of data protection regulations, the need for innovative privacy-preserving approaches to process and analyse sensitive data has become apparent. One approach is the Personal Health Train (PHT) that brings analysis code to the data and conducts the data processing at the data premises. However, despite its demonstrated success in various studies, the execution of external code in sensitive environments, such as hospitals, introduces new research challenges because the interactions of the code with sensitive data are often incomprehensible and lack transparency. These interactions raise concerns about potential effects on the data and increases the risk of data breaches. To address this issue, this work discusses a PHT-aligned security and audit pipeline inspired by DevSecOps principles. The automated pipeline incorporates multiple phases that detect vulnerabilities. To thoroughly study its versatility, we evaluate this pipeline in two ways. First, we deliberately introduce vulnerabilities into a PHT. Second, we apply our pipeline to five real-world PHTs, which have been utilised in real-world studies, to audit them for potential vulnerabilities. Our evaluation demonstrates that our designed pipeline successfully identifies potential vulnerabilities and can be applied to real-world studies. In compliance with the requirements of the GDPR for data management, documentation, and protection, our automated approach supports researchers using in their data-intensive work and reduces manual overhead. It can be used as a decision-making tool to assess and document potential vulnerabilities in code for data processing. Ultimately, our work contributes to an increased security and overall transparency of data processing activities within the PHT framework.

cs.CR↗

arXiv2024

Towards Enabling FAIR Dataspaces Using Large Language Models

Dataspaces have recently gained adoption across various sectors, including traditionally less digitized domains such as culture. Leveraging Semantic Web technologies helps to make dataspaces FAIR, but their complexity poses a significant challenge to the adoption of dataspaces and increases their cost. The advent of Large Language Models (LLMs) raises the question of how these models can support the adoption of FAIR dataspaces. In this work, we demonstrate the potential of LLMs in dataspaces with a concrete example. We also derive a research agenda for exploring this emerging field.

cs.CL↗

arXiv2023

Interpreting Black-box Machine Learning Models for High Dimensional Datasets

Deep neural networks (DNNs) have been shown to outperform traditional machine learning algorithms in a broad variety of application domains due to their effectiveness in modeling complex problems and handling high-dimensional datasets. Many real-life datasets, however, are of increasingly high dimensionality, where a large number of features may be irrelevant for both supervised and unsupervised learning tasks. The inclusion of such features would not only introduce unwanted noise but also increase computational complexity. Furthermore, due to high non-linearity and dependency among a large number of features, DNN models tend to be unavoidably opaque and perceived as black-box methods because of their not well-understood internal functioning. Their algorithmic complexity is often simply beyond the capacities of humans to understand the interplay among myriads of hyperparameters. A well-interpretable model can identify statistically significant features and explain the way they affect the model's outcome. In this paper, we propose an efficient method to improve the interpretability of black-box models for classification tasks in the case of high-dimensional datasets. First, we train a black-box model on a high-dimensional dataset to learn the embeddings on which the classification is performed. To decompose the inner working principles of the black-box model and to identify top-k important features, we employ different probing and perturbing techniques. We then approximate the behavior of the black-box model by means of an interpretable surrogate model on the top-k feature space. Finally, we derive decision rules and local explanations from the surrogate model to explain individual decisions. Our approach outperforms state-of-the-art methods like TabNet and XGboost when tested on different datasets with varying dimensionality between 50 and 20,000 w.r.t metrics and explainability.

cs.LG↗

arXiv2023

From Large Language Models to Knowledge Graphs for Biomarker Discovery in Cancer

Domain experts often rely on most recent knowledge for apprehending and disseminating specific biological processes that help them design strategies for developing prevention and therapeutic decision-making in various disease scenarios. A challenging scenarios for artificial intelligence (AI) is using biomedical data (e.g., texts, imaging, omics, and clinical) to provide diagnosis and treatment recommendations for cancerous conditions.~Data and knowledge about biomedical entities like cancer, drugs, genes, proteins, and their mechanism is spread across structured (knowledge bases (KBs)) and unstructured (e.g., scientific articles) sources. A large-scale knowledge graph (KG) can be constructed by integrating and extracting facts about semantically interrelated entities and relations. Such a KG not only allows exploration and question answering (QA) but also enables domain experts to deduce new knowledge. However, exploring and querying large-scale KGs is tedious for non-domain users due to their lack of understanding of the data assets and semantic technologies. In this paper, we develop a domain KG to leverage cancer-specific biomarker discovery and interactive QA. For this, we constructed a domain ontology called OncoNet Ontology (ONO), which enables semantic reasoning for validating gene-disease (different types of cancer) relations. The KG is further enriched by harmonizing the ONO, metadata, controlled vocabularies, and biomedical concepts from scientific articles by employing BioBERT- and SciBERT-based information extractors. Further, since the biomedical domain is evolving, where new findings often replace old ones, without having access to up-to-date scientific findings, there is a high chance an AI system exhibits concept drift while providing diagnosis and treatment. Therefore, we fine-tune the KG using large language models (LLMs) based on more recent articles and KBs.

cs.CL↗

arXiv2023

Facilitating the sharing of electrophysiology data analysis results through in-depth provenance capture

Scientific research demands reproducibility and transparency, particularly in data-intensive fields like electrophysiology. Electrophysiology data is typically analyzed using scripts that generate output files, including figures. Handling these results poses several challenges due to the complexity and interactivity of the analysis process. These stem from the difficulty to discern the analysis steps, parameters, and data flow from the results, making knowledge transfer and findability challenging in collaborative settings. Provenance information tracks data lineage and processes applied to it, and provenance capture during the execution of an analysis script can address those challenges. We present Alpaca (Automated Lightweight Provenance Capture), a tool that captures fine-grained provenance information with minimal user intervention when running data analysis pipelines implemented in Python scripts. Alpaca records inputs, outputs, and function parameters and structures information according to the W3C PROV standard. We demonstrate the tool using a realistic use case involving multichannel local field potential recordings of a neurophysiological experiment, highlighting how the tool makes result details known in a standardized manner in order to address the challenges of the analysis process. Ultimately, using Alpaca will help to represent results according to the FAIR principles, which will improve research reproducibility and facilitate sharing the results of data analyses.

q-bio.NC↗

arXiv2023

Überprüfung von Integritätsbedingungen in Deduktiven Datenbanken

Advancements in computer science and AI lead to the development of larger, more complex knowledge bases. These are susceptible to contradictions, particularly when multiple experts are involved. To ensure integrity during changes, procedures are needed. This work addresses the problem from a logical programming perspective. Integrity violations can be interpreted as special operations on proofs of integrity constraints, with SLDNF proofs being the focus. We define a proof tree as a special data structure and demonstrate the implication of the existence of an SLDNF proof through such a tree. Proof trees are more convenient than SLDNF trees and allow set-oriented considerations of proofs. They also present the proof structure more clearly, enabling further applications. Using this structure, we determine a minimal set of conditions that specify when a change in the knowledge base affects the validity of an integrity constraint. Additionally, this approach allows for the reuse of large parts of the old proof when searching for a new one, which reduces the effort compared to previous approaches.

cs.DB↗