Program

Data Science for Healthcare I

Chair:

Michelangelo Ceci, Università degli Studi di Bari "Aldo Moro"

Diabetes Management: Non-Invasive Glucose Monitoring via ECG and Machine Learning Matteo Ruggiero, Alessia Auriemma Citarella, Fabiola De Marco, Luigi Di Biasi, Genoveffa Tortora

Improving Preliminary Diagnosis in Italian Through a Retrieval-Augmented Medical Chatbot
Mariano Barone, Gian Marco Orlando, Marco Perillo, Giuseppe Riccio, Antonio Romano, Diego Russo, Ferdinando Tammaro, Valerio La Gatta, Marco Postiglione, Vincenzo Moscato

DART: A Structured Dataset of Regulatory Drug Documents in Italian for Clinical NLP
Mariano Barone, Antonio Laudante, Giuseppe Riccio, Antonio Romano, Marco Postiglione, Vincenzo Moscato

Homomorphic Encryption for EDSS Classification: Safeguarding Patient Privacy in MRI-Based Assessment of Multiple Sclerosis Stefano Cirillo, Vincenzo Deufemia, Luigi Di Biasi, Giuseppe Polese, Giandomenico Solimando, Genoveffa Tortora

Large Language Models

Chair:

Nicola Bena, Università degli Studi di Milano

Program

Investigating Edge Fine-Tuning of Large Language Models in a Federated Environment Lorenzo Colombi, Michela Vespa, Francesco Resca, Sara Cavicchi, Edoardo di Caro, Elena Bellodi, Mauro Tortonesi, Cesare Stefanelli

Question Answering through Retrieval-Augmented Large Language Models: an Experimental Evaluation Claudia Diamantini, Alex Mircoli, Alessia Pagnotta, Domenico Potena, Maria Cristina Recchioni, Emanuele Storti

PARSAL: Pipeline for Automatic Retrieval and Structuring of Academic Literature
Salvatore Contino, Irene Siragusa, Giovanni Sciortino, Roberto Pirrone

Advances in Data Science and AI

Chair:

Claudia Diamantini, Università Politecnica delle Marche

Program

Space-efficient private estimation of quantiles Massimo Cafaro, Angelo Coluccia, Italo Epicoco, Marco Pulimeno

HybridKAN: Leveraging Multi-Sized Sub-MLPs for Enhanced Performance Fabio D'Andreamatteo, Andrea D'Angelo, Giovanni Stilo

BoXoR-C: Coherence and Directionality of Counterfactual Explanations Antonio Maratea, Domenico Lugubre

Non-Euclidean Geometries for Prototype-Based Image Classification: a Reality Check
Samuele Fonio, Silvia Grosso, Sara Bouchenak

Security, Privacy, and Beyond

Chair:

Chiara Boldrini, CNR, IIT

Program

Bridging LLMs and Symbolic Reasoning in Educational QA Systems: Insights from the XAI Challenge at IJCNN 2025
Long S. T. Nguyen, Khang H. N. Vo, Thu H. A. Nguyen, Tuan C. Bui, Duc Q. Nguyen, Thanh-Tung Tran, Anh D. Nguyen, Minh L. Nguyen, Fabien Baldacci, Thang H. Bui, Emanuel Di Nardo, Angelo Ciaramella, Son H. Le, Ihsan Ullah, Lorenzo Di Rocco, Tho T. Quan

Network Traffic-Based Malware Detection for Android Devices Dincy R. Arikkat, Hasna Jabeen N. K., Mirko Lai, Serena Nicolazzo, Antonino Nocera, Anafa P., Vinod P., Rafidha Rehiman K.A.

A Privacy-Preserving Schema Matching Technique based on k-Anonymity
Domenico Beneventano, Sonia Bergamaschi, Carmelo La Corte

Data Science: Technical Advances and Societal Perspectives

Chair:

Matteo Zignani, Università degli Studi di Milano

Program

Dividi et Impera: Enhancing Synthetic Data Fidelity through Data Partitioning Yordanos Nebiyou Yifru, Salvatore Distefano

Exploring the Dataset Landscape for Automated Propaganda Detection: A Data-Centric Insight Marco Usai, Davide Antonio Mura, Andrea Loddo, Manuela Sanguinetti, Luca Zedda, Cecilia Di Ruberto, Maurizio Atzori

The BotBid Project: Data Education at Work Giorgio Delzanno, Giovanna Guerrini, Enrica Roccotiello, Angela Sugliano, Giovanni Zanone

S-PIC4CHU: Semantics-Enriched Techniques for Data Preparation in Data Science Gianvincenzo Alfano, Ilaria Bartolini, Diego Calvanese, Paolo Ciaccia, Sergio Greco, Davide Lanti, Pasquale Leonardo Lazzaro, Emilia Lenzi, Davide Martinenghi, Cristian Molinaro, Marco Patella, Letizia Tanca, Riccardo Torlone, Irina Trubitsyna

Storage and Querying

Chair:

Antonio Maratea, Università degli Studi di Napoli "Parthenope"

Program

From Data Sources to User Screens: a Multimodal hEalth Data lakehouse for ITAly (MEDITA)
Maria Paola Priola

Logic-based Semantics and Query Entailment for Knowledge Graphs
Roberto Maria Delfino, Maurizio Lenzerini, Antonella Poggi

A new framework for Integrated distributed data management, processing, and advanced analytics supporting Digital Twin systems: a Proof of Concept Nicola Masi, Chiara Maria Capizzi, Francesco Patanè, Giuseppe Veneziano, Davide Profeta, Sergio Comella, Giada Sajeva, Lorenzo Gori, Susanna Bonura

Controlled Query Evaluation under Epistemic Protection Policies
Gianluca Cima, Domenico Lembo, Lorenzo Marconi, Riccardo Rosati, Domenico Fabio Savo

Data Science and AI for Good II

Chair:

Roberto Esposito, Università degli Studi di Torino

Program

Climate Data IT: towards a National Data Hub for Climate Science
Sandro Fiore, Paola Nassisi, Ludovica Sacco, Gabriele Padovani, Fabrizio Antonio, Donatello Elia, Italo Epicoco, Nicola Giuseppe Marchioro, Marco Robol, Konstantinos Zefkilis, Flavio Vella, Carolina Sopranzetti, Andrea Alessandri, Annalisa Cherchi, Franco Catalano, Gianmaria Sannino, Gabriella Scipione, Marco Puccini, Erika Coppola, Ivan Girotto, Marco Reale, Stefano Salon, Antonella Galizia, Marcello Iotti, Hilary J Oliver, Valentine Anantharaj, Silvio Gualdi

Variational Inference for the Partial Credit Model Paolo Frazzetto, Nicolò Navarin, Alessandro Sperduti

Urban Region Embeddings from Service-Specific Mobile Traffic Data Giulio Loddi, Chiara Pugliese, Francesco Lettich, Chiara Renso, Fabio Pinelli

Analyzing and visualizing open urban data to study city accessibility and segregation
Mirko Lai, Anna Sapienza, Salvatore Vilella, Massimo Canonico, Federica Cena, Giancarlo Ruffo

Data Science and AI for Good III

Chair:

Domenico Beneventano, Università degli Studi di Modena e Reggio Emilia

Program

A Deep Learning-based Anomaly Detection System for Railway Equipment Michele Bevilacqua, Walter Didimo, Alessandro Eugeni, Fabrizio Montecchiani, Giovanni Spinelli

Aircraft data analysis and predictive maintenance: The experience of the 'Arrowhead-fpvn' project Gianluca Boleto, Gianluca Sommariva, Luca Oneto, Stefano Rovetta, Alberto Calvo, Luca Mortorano, Claudio Porretti, Davide Anguita

Automated Patent Landscaping via Citation Expansion and Semantic Filtering for Twin (Green and Digital) Patents
Francesca Ghinami

On effective ways to warn people against misleading contents Luigi Catuogno, Emanuel Di Nardo, Clemente Galdi, Gennaro Ragucci

A Service-Oriented Architecture for the Digital Thread of Smart Products Devis Bianchini, Massimiliano Garda, Anisa Rula

SLO-aware orchestration and autoscaling for microservice based applications
Orazio Tomarchio, Angelo Marchese

Fairness and Applications of Data Science

Chair:

Valerio Bellandi, Università degli Studi di Milano

Program

Reconciling Statistical and Causal Metrics of Fairness in Machine Learning on Data-Driven Systems Chiara Criscuolo, Davide Martinenghi, Jing Huang

Supporting human evaluation for the Fundamental Rights Impact Assessment of an AI-based hiring system
Roberta Savella, Roberto Trasarti, Francesca Pratesi, Marco Di Cristo

From Wearable Data to Computational Notebooks Anna Khristodulo, Giancarlo Mascetti, Giorgio Delzanno, Marta Bassi, Nicola Minuto

The ego network of words Kilian Ollivier, Chiara Boldrini, Andrea Passarella, Marco Conti

Data Science for Healthcare II

Chair:

Nicola Bena, Università degli Studi di Milano

Program

A Federated Learning Architecture for Prostate MRI Image Segmentation
Alice Bovio, Mario Barile, Francesco Pallotta, Lorenzo Pede, Alessandro Maiocchi, Marco Alì, Fatemeh Darvizeh, Deborah Fazzini, Francesco Lacavalla, Massimo Banzi, Gabriele Gianini, Filippo Berto, Ruslan Bondaruc, Ernesto Damiani, Saman Fouladi

A secure AI-enabled platform to support the automated Alzheimer's disease diagnosis
Claudio A. Ardagna, Veronica Buttaro, Michelangelo Ceci, Marco Luzzara, Antonio Pellicani, Gianvito Pio, Antongiacomo Polimeno

Health Data Sharing for Secondary Use: a Critical Overview Alessio Famiani, Ruggero G. Pensa

Automatic Evaluation of Cancer Reduction during Radiotherapy
Valerio Bellandi, Stefano Siccardi

BigHPC2025 - Special Track on Big Data and High-Performance Computing

Chair:

Alessia Antelmi and Massimo Cafaro, Università degli Studi di Torino and Università del Salento

Program

dqsweep: Parameter Sweeps for Benchmarking Distributed Quantum Computing Applications
Quang-Phong Tran (IIT-CNR), Claudio Cicconetti (IIT-CNR)*, Marco Conti (IIT-CNR), and Andrea Passarella (IIT-CNR)

The Quantum Bottleneck: An Analysis of Data Balancing in QML for Security
Mansur Ziiatdinov (UNIME)*, and Salvatore Distefano (UNIME)

Exploring an implementation of quantum learning pipeline for support vector machines
Mario Bifulco (UNITO), and Luca Roversi (UNITO)

Panel: What's Next for BigHPC? Parallel, Distributed, and Quantum Horizons
Claudio Cicconetti (IIT-CNR), Biagio Cosenza (UNISA), and Sandro Fiore (UNITN)

BigHPC2025 - Special Track on Big Data and High-Performance Computing

Chair:

Alessia Antelmi and Massimo Cafaro, Università degli Studi di Torino and Università del Salento

Program

Hiding Latencies in Network-Based Image Loading for Deep Learning
Francesco Versaci (CRS4)*, and Giovanni Busonera (CRS4)

The impact of ECS logic on parallel performance in agent-based model simulations
Daniele De Vinco (UNISA)*, Antonio Ambrosio (UNISA), Francesco Foglia (UNISA), Carmine Spagnuolo (UNISA), and Vittorio Scarano (UNISA)

Federated Learning for Distributed Weather Forecasting: A Practical Approach on Real Multidimensional Georeferenced Data
Giuseppe Fiorillo (UNIParthenope)*, Attilio Di Vicino (UNIParthenope), Luigi Galluccio (UNIParthenope), and Raffaele Montella (UNIParthenope)

Benchmarking Federated Learning Frameworks Scalability
Gianluca Mittone (UNITO), Samuele Fonio (UNITO)*, Robert Birke (UNITO), and Marco Aldinucci (UNITO)

Tutorial "Tutorial Mining and learning on temporal networks"

Abstract

In many real-world systems, the relationships between entities are not static but evolve over time and across multiple dimensions. Examples include interactions in social media platforms, financial transactions, communication networks, and decentralized blockchain infrastructures. These systems are best modeled as temporal networks, where nodes and edges are associated with time-varying properties, and often as heterogeneous networks, where multiple types of nodes and relations coexist. Modeling these dynamics is crucial for uncovering the underlying mechanisms that govern the evolution of complex systems.

This tutorial is structured in three interconnected parts, each balancing theoretical foundations with practical implementation. It presents a progression from fundamental modeling of temporal dynamics to interpretable pattern mining, and finally to advanced machine learning techniques for prediction tasks.

Part I - Temporal Network: Models and Analysis

This section introduces the foundations of temporal networks and their practical analysis using the Raphtory framework, a scalable and expressive library for analyzing evolving graphs. Participants will learn how to model systems where edges appear, disappear, or change attributes over time, and how to compute metrics that reflect the temporal evolution of network structure.

Topics include:

Representations of temporal networks (event streams, time snapshots, interval graphs).
Temporal extensions of network metrics (e.g., time-aware degree, betweenness).
Use of Python library Raphtory to construct and query large-scale temporal graphs.
Applications in domains such as social platforms and Web3 ecosystems.

The hands-on session guides participants through end-to-end workflows, including data ingestion, time slicing, and metric computation, using open datasets.

Part II - Mining Temporal Networks through Graph Evolution Rules

Many existing approaches to modeling network dynamics rely on simplistic assumptions — often attributing structural changes to a single parameter or rule. However, empirical systems exhibit heterogeneous and multifaceted behaviors, making such models insufficient.

Graph Evolution Rules (GERs) offer a novel and interpretable framework for discovering multiple, overlapping mechanisms that govern temporal graph changes without assuming a priori models.

This section presents GERs as a formalism to:

Capture statistically significant and human-readable rules from evolving networks.
Uncover heterogeneous behaviors across network regions or time periods.
Explore how distinct mechanisms interact or compete over time.

Participants will use the Geranio library to extract GERs from real datasets, validate them against null models, and visualize the impact of different rules. Use cases will include communication platforms, dynamic communities in social networks, and behavioral patterns in blockchain transactions.

Part III - Machine Learning on Temporal (Heterogeneous) Networks

The final part of the tutorial focuses on predictive modeling of temporal graphs using machine learning, particularly deep learning methods such as Graph Neural Networks (GNNs). The tutorial addresses the unique challenges posed by temporal and heterogeneous data, including variable-length sequences, evolving topologies, and the need to generalize across time. Key topics include:

Temporal graph learning paradigms: snapshot-based vs. event-based models.
Architectures for temporal GNNs: Temporal GraphSAGE, TGAT, TGN.
Learning node and edge dynamics with temporal random walks and message-passing.
Handling heterogeneous node/edge types in temporal GNNs.
Dynamic link prediction and evaluation protocols.

Participants will engage in a practical session using PyTorch Geometric (PyG), implementing link prediction models on annotated temporal datasets. Emphasis will be placed on reproducibility, interpretability, and evaluation of predictive performance under realistic temporal constraints.

This comprehensive tutorial provides participants with a modern and integrated perspective on the challenges and opportunities in analyzing temporal and heterogeneous networks. By connecting graph theory, pattern mining, and deep learning, it enables attendees to effectively model complex systems and to extract actionable insights from their dynamic structure.

Presenters

Alessia Galdeman. She received a degree in Computer Science for digital communication in 2018, and a master's degree in Computer Science in 2021. She is currently a PostDoc at IT University of Copenhagen in the NEtwoRks, Data, and Society — NERDS — research group on the COllective COordination through Online Social media (COCOONS) project, supervised by Prof. Luca Maria Aiello. She is a member of the Executive Committee of NetPlace. Her current research interests include temporal networks, subgraph mining, Web3 platforms.

Manuel Dileo. He received a master's degree in Computer Science in 2022. He is currently a senior Ph.D. student at the Computer Science Department of the University of Milan, where he is also a tutor of machine learning courses. He has published works on machine learning for temporal networks, link prediction in online social networks, and temporal knowledge graphs. He also contributed to PyTorch Geometric, the main library for the development of machine learning on graph architectures.

Matteo Zignani. Matteo Zignani is an associate professor of the Connets Lab in the Department of Computer Science at the University of Milan, where he teaches machine learning and network analysis. His research activity takes place within data and network science, with a focus on data mining and machine learning applications on different networked systems, from blockchain and online social networks to human mobility and communication networks.

Sabrina Gaito. Sabrina Gaito is a full professor in the Department of Computer Science at the University of Milan, where she leads the Connets Lab and teaches machine learning, network science, and social media mining. Her research activity takes place within the network science and machine learning fields with applications to social, economic, and blockchain-based networks. She is regularly a member of the program committee of main international conferences and of the editorial board of prominent journals.

Tutorial "Open Data and Air Quality Data for Personalized Route Planning in Smart Cities"

Abstract

Urban mobility planning is evolving towards a more sustainable, health-conscious, and data-driven approach. This hands-on tutorial will guide participants through a practical methodology to integrate heterogeneous open data sources (GTFS, OpenStreetMap, air quality APIs) into actionable insights for personalized active mobility routing. Attendees will learn to model multi-layer transport networks, design adaptable routing algorithms, and evaluate trade-offs between safety, pollution exposure, and quality of experience specifically for pedestrians and cyclists.

Session 1: Foundations of Smart Mobility Data (30 min)

Data Ecosystem Deep Dive: Explore the role of open data (GTFS, OpenStreetMap, Open Data portals) in modern mobility solutions
Graph Modeling Techniques: How to represent multimodal transport networks as interconnected layers (pedestrian paths, cycling lanes, road network) with dynamic weights (e.g., real-time pollution levels) for routing purposes

Session 2: Multi-objective Routing for Smart Cities (30 min)

Multi-Objective Optimization: Trade-off analysis between classic (distance, time) and well-being metrics (air pollution, safety, green area coverage).
Personalization Frameworks: Adapting algorithms for diverse user profiles (e.g., commuters prioritizing speed vs. tourists favoring green/low-pollution paths).
Case Studies from Modena and Ferrara: Real-world challenges in integrating heterogeneous data into routing logic.

Session 3: Hands-on Implementation (50 min)

Live Coding Demo: Build a proof-of-concept routing system using Python (NetworkX, OSMnx) through a pre-configured Jupyter notebook.

Step 1: Data fusion – creating a new graph combining OSM geometries with contextual data (air pollution, road accidents).
Step 2: Graph weighting strategies – assigning costs based on dynamic factors.
Step 3: Visualizing results through maps comparing route alternatives.

Participant Interaction: Modify costs to define new objectives (e.g., 'Avoid schools during pickup hours') or integrate new data layers.

Session 4: Q&A and Discussion (10 min)

Open discussion on scalability, integration with municipal services, and future research directions.

Presenters

Laura Po, PhD. Laura Po is a Web and Data Scientist based in Modena, Italy. Since 2020, she has served as an Associate Professor at the 'Enzo Ferrari' Engineering Department of the University of Modena and Reggio Emilia (UNIMORE), and is a senior researcher in the Big Data Research Group (http://www.dbgroup.unimo.it/).
Her research focuses on information representation and management, with particular interest in Big Data analytics, urban and environmental data integration, Web and Information Systems, Knowledge Engineering, AI and Machine Learning, Digital Twins, and Graph-based models. She has also worked extensively on Word Sense Disambiguation, Data Streams, Linked Open Data, and Smart Cities.
She earned a Ph.D. in Computer Engineering and Science from UNIMORE in 2009 and holds a cum laude MSc in Computer Engineering (2005). She obtained her professional engineering license in 2006.
She is co-founder and scientific committee member of DataRiver srl, a research spin-off offering semantic-based data integration solutions, recognized as a Qualified Research Organization by the Emilia-Romagna High Technology Network since 2009.
She is a member of the Board of the International Doctoral School in ICT at UNIMORE (since 2011), the Interdepartmental Research Center on AI (AIRI) (since 2021), and the DHMoRe Center for Digital Humanities (since 2020).
She has been a visiting professor at Université Claude Bernard Lyon 1 and CPE Lyon (France, 2023), and at Universidad de Zaragoza (Spain, 2022).
She has authored over 70 publications (Scopus), with an H-index of 17 and over 800 citations. Her work appears in high-impact journals such as Environmental Modelling & Software, Ecological Informatics, Knowledge Based Systems, and Future Generation Computer Systems.
She currently leads or contributes to several national and European projects: AIQS, a national project on AI-based sensor data correction for green urban routing (ECOSISTER, 2024-25); Lively Ageing, on active ageing and predictive monitoring (Ministry of Health, 2022-27); ECOSISTER (2022-26), supporting carbon-neutral urban mobility planning in Emilia-Romagna (NextGeneration EU).
Previously, she was Project Leader of the European project TRAFAIR - Understanding Traffic Flows to Improve Air Quality (2018-2021), and Task Leader in Re-Search Alps (2017-2019).

Federica Rollo, PhD. Federica Rollo is an Assistant Professor at the University of Modena and Reggio Emilia (Italy), specializing in AI for smart cities, NLP, and knowledge graphs. She holds a PhD in Computer Engineering (2021) and has worked on national and European projects like TRAFAIR (traffic/air quality analytics), ECOSISTER (smart mobility), and AIQS (air quality integration in mobility solutions).
Her research focuses on event extraction from text, sensor data analysis, Italian-language NLP, urban analytics, graph-based road network modeling, and multi-objective routing.
She is the author of 28 publications (Scopus H-index: 11) and has supervised multiple MSc/PhD theses.
She was a visiting researcher at the University of Galway (Ireland, 2021) and the Universidade da Coruña (Spain, 2025).

Implementing Trusted European Data Spaces: From Strategy to Practice

Abstract

In its European Strategy for Data, the European Commission emphasized that citizens, businesses, and public authorities can be empowered to make better decisions through the effective use of data. To unlock the full potential of the vast and growing volume of data generated daily across Europe, it is essential to build trust in the ways data are collected, accessed, and used.

In this context, the development of a robust European Data Spaces ecosystem is seen as a key opportunity to ensure better access to data and its responsible use. A Data Space is a structured framework that facilitates secure and trustworthy data sharing within a defined data ecosystem. It provides clear rules and technical means for participants to exchange, trade, and collaborate on data assets in a manner that complies with relevant legal and regulatory requirements, while guaranteeing fair treatment for all stakeholders (Source: Data Spaces Support Centre (DSSC) website).

The establishment of Common European Data Spaces across key sectors enables the seamless sharing and reuse of data both within and across domains. This contributes to the emergence of a competitive and dynamic European data market, underpinned by the EU's strong legal foundations in personal data protection, intellectual property rights, safety, and cybersecurity. In turn, this facilitates greater demand for data-driven products and services across the continent.

By promoting standardized technologies and legal frameworks, Data Spaces aim to overcome barriers to data sharing among organizations and foster the creation of innovative, value-added services. Through a focus on interoperability and data sovereignty, Data Spaces play a pivotal role in advancing collaboration and innovation across the European data economy.

The aim of this tutorial is to provide an overview of the current landscape of Data Spaces initiatives and to introduce selected open-source platforms that support their practical deployment.

By the end of the tutorial, participants will be able to:

Understand the strategic role of Data Spaces in the European data economy.
Explain the core components, principles, behind Data Spaces.
Identify key EU initiatives supporting Data Space development.
Explore real-world use cases demonstrating the value and implementation of Data Spaces across sectors.
Evaluate open-source technologies and reference architectures that support the deployment of Data Spaces.
Gain hands-on familiarity with tools and platforms used to build and manage Data Spaces.
Discuss challenges and best practices for data sharing, interoperability, and data sovereignty within multi-stakeholder environments.

Presenters

Antonella Longo. Antonella Longo is an Associate Professor at the Department of Engineering for Innovation at the University of Salento, Italy, where she gained her Ph.D. in Information Engineering. She is the scientific coordinator of DataLab, the University Lab which promotes research and innovation related to big data management and governance in the AI era. Her research interests focus on data management, big data analytics, digital twins, and data spaces, with applications in smart cities, Industry 4.0, and digital transformation. She has been involved in numerous national and European research projects addressing data-driven innovation and the design of interoperable, secure digital ecosystems. Prof. Longo is also active in promoting data governance strategies that ensure data quality, interoperability, and sovereignty across distributed platforms. She regularly publishes in international journals and conferences and contributes to the academic community as a reviewer and member of scientific committees. Prof. Antonella Longo is also one of the founders of the Women in Big Data – Italy, the Italian chapter of the international association which promotes the use of data to reduce the gender gap.

Angelo Martella. Ph.D., Eng. Angelo Martella is a junior researcher at the DataLab of the Department of Engineering for Innovation, University of Salento. He also serves as an academic lecturer in Big Data Management and Data Engineering topics at the University of Salento. His research focuses on Digital Twins and Data Spaces. His areas of expertise therefore span all phases of the design, development, and deployment of digital twins for smart cities, with a particular emphasis on the use of data spaces as critical enablers of data ecosystems—managing data flows in distributed edge-cloud environments using machine learning and AI techniques. He is a Fiware expert and he is currently involved in several research projects within the National Center of High Performance Computing, Big Data, and Quantum Computing.

Cristian Martella. Eng. Cristian Martella received his M.Sc. in Computer Engineering from the University of Salento, where he is currently pursuing a Ph.D. in Engineering of Complex Systems. He is affiliated at the DataLab of the University of Salento. His research focuses on data spaces for digital twins, particularly in domains such as smart energy and smart cities. He explores federated learning architectures for data preparation across the edge-cloud continuum, aiming to enhance interoperability among heterogeneous distributed systems while ensuring data protection and sovereignty. By leveraging advanced machine learning techniques and digital twin technology, he seeks to enable seamless data exchange and collaboration in urban environments, ultimately fostering innovation in resource management and decision-making. Dedicated to promoting sustainable and resilient smart city infrastructures and smart grids, he is committed to developing solutions that empower stakeholders while safeguarding their data rights.

Intelligent Statistics: Applied Methods in Healthcare, Conflict Monitoring, and High-Frequency Trading

The tutorial aims to showcase recent contributions in the field of data analysis, where statistical modeling and machine learning intersect to address real-world challenges. Emphasis will be placed on the interpretability and practical implications of the proposed techniques, particularly in the analysis of high-dimensional, complex, and multisource datasets. Participants will gain insights into how advanced statistical tools and AI-driven models can enhance understanding and decision-making across domains such as healthcare, finance, and social dynamics.

The tutorial is organized into three 15-minute presentations, followed by a 15-minute discussion.

Automated Clinical Coding models

In this work we present a methodology involving Natural Language Processing (NLP) and Artificial Intelligence models which automatically assign standardized medical codes to the given clinical reports, both in English and Italian. The objective is to provide a useful pipeline leveraging external knowledge as the standardized medical annotations in the ICD (International Classification of Diseases) from the World Health Organization. On doing this, the overall accuracy and efficiency in medical billing, reporting, and data analysis is significantly improved. In addition, they are essential for modern healthcare systems aiming to reduce manual workload and errors.

Speaker: Irene Siragusa (with Marco Speciale, Roberto Pirrone, Antonella Plaia). Irene Siragusa is a postdoctoral fellow at the department of Engineering of the University of Palermo. Her PhD in Information and Communication Technologies focused on NLP techniques towards Information Retrieval involving generative Large Language Models and architectures based on the Retrieval Augmented Generation approach in domain-specific applications, such as the biomedical one.

Composite Indicators of Global Unrest: A Quantitative Approach to Instability

In this contribution, we propose to quantify global instability by designing and implementing a comprehensive framework of indicators that systematically track and capture unrest-related events across countries and over time. The primary objective is to distill the complexity, intensity, and social impact of such events into a set of clear, up-to-date, and interpretable measures. These indicators are intended to serve as practical tools for policymakers, analysts, and stakeholders, helping them anticipate potential crises, formulate effective preventive strategies, and make more informed decisions in response to evolving geopolitical challenges. To build this framework, we develop three composite indicators, each combining a variety of data sources to ensure robustness and breadth. Specifically, we integrate large-scale, time-sensitive data from the ACLED (Armed Conflict Location and Event Data) project—known for its detailed records of conflict and political violence—with a dedicated dataset curated by our research team. This custom dataset documents a wide spectrum of unrest events, including wars, armed conflicts, civil wars, violent demonstrations, and peaceful protests, covering the period from 2010 to the present. Through this integrated approach, we aim to provide a nuanced and dynamic view of global instability that can support both academic research and real-world decision-making. A dashboard, defined by the authors, will be present to the participants.,/p>

Speakers: Marco Tagliapietra and Luca Macis (with Elena Siletti and Paola Pisano). Marco Tagliapietra and Luca Macis are PhD students in Economics at the Department of Economics and Statistics 'Cognetti de Martiis', University of Turin, and co-founders of Deeplomacy, a startup platform for geopolitical risk forecasting through AI and predictive analytics. They collaborate with the HighESt laboratory at the Department of Economics and Statistics 'Cognetti de Martiis', University of Turin on applied research in machine learning and early warning systems. Especially, Marco has been involved in international research competitions and interdisciplinary projects of high scientific impact, with a focus on innovation in data integration and AI for peacebuilding. Luca is an expert in artificial intelligence applied to geopolitical analysis and has authored several scientific publications in the field. Moreover, he collaborates with academic and institutional partners on research projects that combine theoretical insights with applied methodologies, contributing to high-impact solutions for early warning and international security.

LLM-Based Document Segmentation for Enhanced RAG Performance

High-frequency trading (HFT) represents a rapidly evolving area in financial markets, where decision making and execution occur at high speeds. However, exploiting high-frequency data (HFD) requires innovative methods to handle inherent complexities, including noise and nonlinear dynamics. Deep learning (DL) models have emerged in HFT, outperforming traditional statistical methods in tasks such as forecasting and classification. In this study, we focus on convolutional neural networks (CNNs), particularly 2D CNNs, to evaluate different preprocessing and feature engineering approaches, with the goal of identifying which approach can lead to better accuracy. We will compare the results with each other and also with more traditional models such as Logistic Regression and Random Forest thanks to the innovative S.A.F.E. methodology, which allows us to compare even very different models. The use of 2D CNNs involves transforming datasets into image representations, allowing an alternative pattern recognition technique that uncovers relationships and structures not readily apparent in raw datasets.

Speaker: Jacopo Chiapparino (with Paola Cerchiello). Jacopo Chiapparino is a highly skilled economist and data scientist with expertise in Python, R, machine learning, and financial econometrics, currently pursuing a PhD in Artificial Intelligence applied to finance. Experienced in AI engineering, risk management, and quantitative modeling, with a strong academic background and ongoing research in high-frequency trading and pattern recognition. Recognized for innovative contributions, including publications and awards in financial and technical analysis fields.

Submission link -> Create new submission -> Workshop: Scientific HPC in the pre-Exascale era

Valentina Cesare
valentina.cesare@inaf.it

Valentina Cesare is a fixed-term technologist at INAF - IRA (starting date 15/05/2025), where she is about to begin a work concerning GPU porting of scientific applications related to NGCroce project. From 01/12/2020 to 14/05/2025, she worked at INAF - OACT, firstly as fellowship student, then as research associate, and at last as fixed-term technologist, within a project about GPU porting of scientific applications related to the Gaia space mission, within the framework of the ICSC - National Center for Research in HPC, Big Data, and Quantum; Computing (PNRR - Future Computing initiative). A future involvement in the Euclid Consortium is planned. She received her Ph.D. in Physics and Astrophysics in March 2021 from the Physics Department of the University of Turin, with a thesis focused on galaxy dynamics in the framework of the modified gravity theory Refracted Gravity.

Alberto Vecchiato
vecchiat@oato.inaf.it

Alberto Vecchiato is working in software development as responsible of the AVU-GSR pipeline within the Gaia mission at INAF-Astrophysical Observatory of Torino, where he has held a permanent position since 2007. Generally, he is mainly working in the fields of astrometry, physics of gravitation, and tests of gravity physics theories. Since 2012, he has developed an interest for archaeoastronomy and the history of astronomy. He got his master thesis in Physics in 1996 and his PhD in Physics in 2001 at the University of Padova. A future involvement in the Euclid Consortium is planned.

Gianluca Mittone is a postdoctoral researcher in computer science at the University of Turin, and his research is focused on the convergence between High-Performance Computing (HPC) and Artificial Intelligence (AI) techniques. In less than 5 years of research activity, he achieved 16 scientific publications and an H-index of 9 (source Google Scholar). His works are mainly related to the use of AI in medicine and Federated Learning (FL). Specifically, he is currently investigating the deployment of cross-HPC FL workloads through workflow-based approaches; and the use of FL as a tool to allow AI-based computation to scale efficiently for HPC benchmarking purposes. He is currently co-principal investigator in a joint research effort between the University of Turin and Telecom Italia (TIM) to develop an FL-as-a-Service platform for the "TIM Edge & Cloud Continuum" IPCEI European Project. His achievements rewarded him with an HPC-Europa3 scholarship and an EuroPar foundation studentship, together with the 'Best PhD Symposium Award' during the 2023 edition of the conference. His PRAISE Score, an AI-based diagnostic tool, has been awarded as an officially recommended diagnostic software" by the European Society of Cardiology in their 2023 guidelines.

Bruno Casella
bruno.casella@unito.it

Bruno Casella is a research associate at the Computer Science Department of the University of Turin. He got the Ph.D. in Modeling and Data Science in June 2025 at the same department, financed by Leonardo Company. He graduated in Computer Engineering in 2020 with a thesis on the performances of AlphaZero, an artificial intelligence method based on reinforcement learning for the game of chess, that is able to win against the human world champion, in different scenarios. He also received the Master's Degree in Data Science for management in 2021 with a thesis on Federated Transfer Learning.

Secure Data Architectures and Post-Quantum Strategies: Technologies for Resilient ICT Systems

Abstract

Ensuring data security in current and future ICT systems requires coordinated efforts across cryptographic research, software engineering, and institutional support. This workshop presents integrated strategies that address emerging threats, such as those posed by quantum computing, and practical challenges in secure software deployment. Experimental results on post-quantum TLS and digital signatures show the performance trade-offs of adopting quantum-resistant algorithms in real-world settings. These findings complement efforts to improve container security in CI/CD pipelines through automated threat analysis and enforcement mechanisms. The workshop also highlights the role of public funding and regional initiatives—such as those led by CERICT—in enabling collaborative research and innovation. This also draws attention to challenges at the network edge—where distributed, resource-constrained systems expand the attack surface and demand tailored security for edge-cloud environments. Together, the contributions provide a structured view of how secure architectures can be designed and deployed in scalable, future-ready environments.

Workshop goals, objectives, outcomes

The purpose of this workshop is to connect researchers, practitioners, and institutional actors working on secure digital infrastructures with a focus on post-quantum cryptography, software supply chain security, and edge-cloud systems. The workshop provides a space where academic results, experimental validations, and technology transfer models can be shared and discussed. Experts from research institutions and industry will present their ongoing work on post-quantum communication protocols, container security mechanisms, and proactive defense strategies for distributed systems. Representatives from public initiatives will outline how funding programs can support collaborative development and enable the integration of these technologies in production environments. The workshop is intended to support dialogue across sectors and to promote practical adoption of secure and scalable ICT solutions.

Full description

Securing modern digital infrastructures requires a multidisciplinary approach that connects cryptographic innovation, secure software engineering, and institutional support. As technology evolves—from the emergence of quantum computing to the widespread use of microservice architectures—security strategies must adapt at both the algorithmic and system levels. At the same time, public and regional initiatives play a critical role in sustaining applied research and enabling real-world deployment of advanced solutions. This presentation brings together different key efforts that reflect this intersection: the implementation of post-quantum secure communication protocols, the integration of security mechanisms into containerized software pipelines, and the role of regional funding initiatives in enabling scalable and collaborative innovation. Together, they outline a framework for designing secure and resilient digital ecosystems, capable of withstanding emerging threats and supporting sustainable development.

Namirial

Namirial has performed tests in different areas impacted by the quantum threat; we have investigated how a quantum-safe version of TLS differs from the current version and have also taken into account a "hybrid" scenario. Moreover, we have investigated the difference in performances between new, quantum-safe signature algorithms, such as Dilithium and Falcon, and the well known and widely adopted RSA-based signatures.

CERICT

Public and regional funding plays a key role in supporting research, innovation, and technology transfer. The Competence Center on ICT of the Campania Region (CERICT) has been actively involved in leveraging national and regional funding programs to promote initiatives in the field of Information and Communication Technologies. This presentation highlights CERICT’s approach to identifying and exploiting available funding tools to support collaborative research and industrial innovation. Through targeted calls and structured partnerships, the center has contributed to the development of projects involving universities, research institutions, and local enterprises. Several notable initiatives will be presented, with a focus on their objectives, involved stakeholders, and achieved results. Particular attention will be given to how funding mechanisms have been used not only to support technical development, but also to strengthen regional cooperation and the growth of a sustainable innovation ecosystem.

SeCCo-OC

Modern applications increasingly rely on containers to support the microservices development model. Containers simplify deployment and integration, especially when used in CI/CD pipelines, where developers focus mainly on automating delivery workflows. However, security tasks are often neglected in this process, raising the risk of introducing vulnerabilities into the application, the platform, or the underlying framework. SecCo-OC addresses this gap by designing a container security architecture that can be integrated directly into the CI/CD workflow. The goal is to automate the identification and mitigation of security threats at both development time (static and dynamic analysis) and runtime (through enforcement mechanisms), ensuring that containers are securely built and maintained before being deployed. The solution also focuses on extending the capabilities of container technology. It explores virtualization techniques, controlled access to specific hardware resources, and the embedding of security services and policies inside the container. These elements contribute to a container model that balances security with functionality and performance. To enable adoption in different deployment scenarios, including edge and pervasive computing, the SecCo-OC architecture is built to support scalability, flexibility, and reliability. It leverages cloud and edge infrastructures to extend DevOps practices to security enforcement in distributed environments.

DEFEDGE

The DEFEDGE project aims to define a set of techniques for the development of secure and resilient edge-cloud systems and for their assessment based on a threat-driven approach. The main idea is to leverage the results of a guided threat modeling process to derive both the security controls and mechanisms to enforce as a mitigation for these threats and the security tests to perform in order to verify the effectiveness of controls in place. In particular, security controls selection and enforcement will follow Moving Target Defense principles, according to which the attack surface of a system is continually and proactively changed to reduce attack success probability. Security testing will exploit existing threat intelligence and attack patterns knowledge bases to derive a set of general-purpose attack procedures that can be suitably customized to test a target system. For the generation of attack procedures and their customization, the project will also explore machine learning techniques to infer new attack patterns and scenarios, in order to improve overall testing effectiveness.

Chairs

Daniele Granata

Angela Bonifati - Distinguished Professor at Lyon 1 University

Topic: Property Graphs to the Rescue for Declarative Causal Analysis

Short Bio:

Angela Bonifati is a Distinguished Professor of Computer Science at Lyon 1 University and at the CNRS Liris research lab, where she leads the Database Group. She is also an Adjunct Professor at the University of Waterloo in Canada from 2020 and a Senior member of the French University Institute (IUF) from 2023. Her current research interests are on several aspects of data management, including graph databases, knowledge graphs, data integration and their applications to data science and artificial intelligence. She has co-authored more than 200 publications in top venues of the data management field, including five Best Paper Awards, two books and an invited paper in ACM Sigmod Record 2018. She is a recipient of an ERC Advanced Grant 2024 dedicated to leading researchers in Europe. She is the youngest recipient of the prestigious IEEE TCDE Impact Award 2023 and a co-recipient of an ACM Research Highlights Award 2023. She is the General Chair of VLDB 2026 and was the Program Chair of IEEE ICDE 2025, ACM Sigmod 2022 and EDBT 2020. She is currently an Associate Editor for the Proceedings of VLDB Vol. 19 and for IEEE TKDE and ACM TODS. She is the Chair of the ACM Sigmod Executive Committee (2025-2029) and was the President of the EDBT Executive Board and Association (2020-2024). She is a member of the IEEE Technical Committee on Data Engineering (2024-2029) and a member of the PVLDB Board of Trustees (2024-2029).

Abstract

Graphs are powerful abstractions for modeling relationships across data and enabling complex data science tasks. In this talk, I will highlight powerful declarative graph operations and delineate their counterparts in causal inference, to support data-driven, personalized decision-making across several scientific domains. I will present our work to align causal analysis with property graphs—the foundation of modern graph databases—by rethinking graph models to incorporate hypernodes, structural equations, and causality-aware query semantics. By unifying graph databases with causal reasoning, causal tasks such as interventions and counterfactuals are mapped to property graph manipulation and transformation, combining expressiveness with computational efficiency.

Horst Simon - Director, ADIA Lab, Abu Dhabi

Topic: ADIA Lab and Three Types of Climate Data Networks

Short Bio:

Dr. Horst D. Simon is an internationally recognized expert in high-performance computing and computational science, with over four decades of experience in parallel algorithms and large-scale numerical methods. After completing his Ph.D. in Mathematics at UC Berkeley, he held leadership roles across academia (Stony Brook, UC Berkeley), industry (Boeing, SGI), and national research labs (NASA Ames, Lawrence Berkeley Lab. At Berkeley Lab, he directed NERSC and served as Deputy Lab Director, earning two prestigious Gordon Bell Prizes and co-editing the biannual TOP500 list of supercomputers. Since 2023, Dr. Simon has been the founding Director of ADIA Lab in Abu Dhabi, spearheading cutting-edge research in computational and data science. He continues to advance scalable algorithms that address complex scientific and societal challenges.

Abstract

This presentation begins with a brief introduction to ADIA Lab, an independent research institute based in Abu Dhabi. ADIA Lab's mission is to advance fundamental and applied research in computational and data science, with a focus on addressing complex real-world challenges across domains such as climate, finance, and health. By fostering interdisciplinary collaborations and developing scalable algorithms and models, the lab seeks to drive innovation at the intersection of theory and practice. As an example of the projects underway at ADIA Lab, we then present our research on climate networks, structured around three complementary paradigms: (a) networks of data, where connections between geographic nodes are derived from statistical relationships, commonly referred to as Tsonis networks; (b) climate data over networks, where climate variables are defined on fixed topologies such as river basins or atmospheric grids—termed geophysical networks; and (c) networks for data, which leverages machine learning and statistical models grounded in network theory to analyze and interpret climate information. Special emphasis is placed on the first two frameworks. We examine how these network types are constructed, the insights they offer for understanding climate variability and model output, and their implications for climate governance. Finally, we discuss how integrating these perspectives can inform more robust analytical tools and policy strategies in the face of climate change.

Emilio Porcu - Professor at Khalifa University, Abu Dhabi

Topic: ADIA Lab and Three Types of Climate Data Networks

Short Bio:

Professor Emilio Porcu is a distinguished expert in spatial-temporal statistics and data science. He earned his Ph.D. in statistics in 2005 and became a full professor by 2012 while holding Chair positions at Newcastle University and Trinity College Dublin. Since August 2020, he has served as Professor of Statistics & Data Science at Khalifa University in Abu Dhabi, and he continues as an adjunct professor at Trinity College Dublin. Dr. Porcu leads a research group specializing in space-time covariance modeling, producing nearly 180 papers and advancing theory and application in areas such as Gaussian random fields on non-Euclidean domains, covariance functions on spheres and networks, and scalable kernel-based methods. He is also a Research Fellow at ADIA Lab, contributing to innovative data science initiatives in climate, complex systems, and computational statistics.

Program

Updated information will be published soon