Kyriakos Stefanidis

Federated data spaces, established through initiatives such as IDSA and GAIA-X, enable organizations to share and monetize datasets under contractual terms. However, enforcing these contracts—particularly detecting unauthorized reuse or modification of datasets—remains an open challenge. We present the Off-Platform Contract Inspector, a component of the PISTIS framework, that implements a modular similarity-detection pipeline combining path-value Jaccard similarity, field-aware type-specific comparisons, and sentence-embedding-based semantic analysis across structured, semi-structured, and unstructured datasets. This contributes as follows: (i) an Inverse Document Frequency (IDF)-weighted structural similarity mechanism that discounts common domain vocabulary via Inverse Document Frequency weighting over the data space catalog, combined with a schema-evidence-gated fusion that reduces false positives from domain vocabulary overlap; (ii) an adaptive threshold optimization mechanism that learns modality-specific fusion weights and decision thresholds via cross-validated grid search; and (iii) a privacy-preserving similarity layer based on MinHash Locality-Sensitive Hashing signatures, Bloom filters with OR folding alignment, and Laplace noise for differential privacy, enabling cross-organizational dataset comparison without exposing raw data. Further, we contribute a threat taxonomy of seven dataset modification types ordered by detection difficulty, and evaluate the system on dataset pairs derived from real-world datasets across three smart-city application domains (Mobility, Energy, Automotive), with controlled augmentations applied to model adversarial behaviors. The IDF-weighted pipeline achieves high precision on intra-domain hard negatives—pairs of different tables from the same data space that share domain vocabulary—where text-similarity baselines produce false positives. The adaptive scheme learns per-modality fusion weights via cross-validated grid search. The privacy-preserving mode operates without accessing raw data and runs noticeably faster than the full pipeline, enabling screening while preserving data confidentiality.

This work explores how deep learning models, with different numbers of parameters, can be effectively applied to detect personal data within unstructured text using Named Entity Recognition (NER) techniques. We evaluate the performance of various architectures by leveraging a plethora of language models (LMs) consisting of Distilbert-base-uncased, Distilbert-base-cased, Bert-base-uncased, Bert-base-cased, Bert-large-uncased, Bert-large-cased, ModernBERT-base, ModernBERT-large, nomic-BERT-2048, RoBERTa-base, DistilRoBERTa-base, RoBERTa-large, Deberta-v3-xsmall, Deberta-v3-small, and Deberta-v3-base, which are evaluated using the performance indices of accuracy, precision, recall, and F1-score. Our experiments show that some Small Language Models (SLMs) compete equally with some corresponding LLMs (Large Language Models), based on the specific PII (Personally Identifiable Information) dataset, thus enhancing personal data detection, which is of paramount importance in financial applications. Moreover, we proposed a novel architecture based on an optimized transformer fine-tuning strategy to improve PII recognition across diverse contexts and conducted an extensive comparative analysis to evaluate the performance of our proposed architecture in relation to all relevant existing approaches reported in the literature. This evaluation, performed on the AI4Privacy PII 43 K dataset, encompasses every publicly available work we identified and provides a thorough benchmarking of our methods within the current research field. The results highlight both the strengths and limitations of existing solutions and demonstrate the effectiveness of SLMs in addressing the challenges of privacy-preserving information extraction.

Trustworthy Identity and Access Management (IAM) is a foundational requirement for federated data trading platforms, yet existing solutions often rely on centralized Identity Providers (IdPs), lack cross-border interoperability, and offer limited support for user-friendly authorization management. These limitations hinder secure onboarding, fine-grained access control, and regulatory compliance, especially within European Union (EU) data spaces governed by the Electronic Identification, Authentication, and Trust Services (eIDAS) 2.0 framework. This work presents a comprehensive IAM framework designed for federated data trading environments, developed within the EU-funded PISTIS project. The framework is based on Keycloak IAM and offers three major capabilities: (i) a novel IAM architecture tailored to distributed data trading scenarios; (ii) full integration of eIDAS-compliant cross-border authentication and initial support for European Digital Identity (EUDI) Wallets; and (iii) a standalone, web-based Access Policy Editor (APE) that abstracts Keycloak’s policy engine and enables non-technical users to define fine-grained, owner-driven access rules. The approach is evaluated across real-world mobility, energy, and automotive industry pilots, demonstrating its effectiveness in enhancing trust, interoperability, and usability within regulated data-sharing ecosystems.

Enabling content-based search in federated data marketplaces, where assets remain distributed and secure, is a significant challenge. This paper introduces a two-stage “retrieve, then re-rank” solution. We present a comparative analysis of Approximate Nearest Neighbor (ANN) and Learning to Rank (LTR) methods, justifying our selection of Locality-Sensitive Hashing (LSH) for retrieval and a weak supervision strategy for re-ranking. Finally, we detail the architecture of the PISTIS distributed query engine, a novel system that implements this approach to provide effective, policy-aware search across federated data sources.

Blockchain (BC) represents a disruptive technology that has been extensively used to ensure immutability of digital transactions. Starting as an underlying mechanism in the digital currency sector, it has been applicable in a wide range of sectors and application domains. Agriculture represents a sector of significance for overall sustainability challenges that is benefiting from digitalisation and technological evolution and the enforcement of Industry 4.0 paradigm shift towards precision agriculture. Introduction of Internet of Things, and Cyber-Physical Systems increase overall complexity, with Big Data analysis and Machine Learning technologies providing innovative applications. BC appears to be a promising technology for agriculture providing mechanisms for tracing of products and overall agricultural SC (SC) management from the farm to the fork. Authors investigate the challenges and open issues for the application of BC in agriculture performing a state of the art analysis along the PESTELS framework. A large number of challenges including technological ones, create big research potential for the evolution of the area.

A Transient Effect Ring Oscillator (TERO) is a special case of a Ring Oscillator (RO) design that exhibits increased sensitivity to intrinsic noise. It can serve as a basis for implementing a True Random Number Generator (TRNG) or a Physically Uncloneable Function (PUF). Also, as a digital sensor for detecting insertion of malicious hardware logic (Trojans) in digital circuits. Here, we explore the application of TERO for detecting hardware Trojans injected in FPGA implementations of the AES cryptographic algorithm. Experiments and comparisons are reported in terms of the frequency as a function of the TERO length. Our findings indicate that TERO-based digital sensors can be used to efficiently detect the presence of the Trojan.

The detection of malicious hardware logic (hardware Trojan) requires test patterns that succeed in exciting the malicious logic part. Testing of all possible input patterns is often prohibitively expensive. As an alternative, we explored previously the applicability of the combinatorial testing principles. In this paper, we turn our focus on the efficiency of this approach for triggering the hidden malicious logic. We present a series of experiments with Trojan designs of various activation patterns and lengths that target a cryptographic module performing AES cryptography. Our findings indicate that the available test suites succeed in triggering the malicious logic in all cases requiring only a very small number of test vectors. Thus, it is an efficient means for detecting malicious hardware logic.