Program 2025

Low Power Global Area Networks in the context of the Thingsat project

Abstract:  TBA

Bio: Didier Donsez is a full professor in the ERODS research group at the Université Grenoble Alpes (UGA). He received his Ph.D in Computer Sciences from the University of Paris VI: Pierre et Marie Curie (1994) and his HDR in Computer Sciences from Université Joseph Fourier (2006). Previously, he has been assistant professor in Computer Sciences at University of Valenciennes, France (1996-2001), then at Université Joseph Fourier (2001-2007). His research interests are in the broad areas of Internet of Things (IoT) and related network (e.g. LoRa).

Fog Computing Challenges in Natural Environment Observatories

Abstract: Natural environment observatories allow a wide range of scientists such as biologists, botanists and hydrologists, to observe a zone of particular interest from the points of view of their different scientific disciplines. They follow a data-driven approach based on a variety of sensors and/or actuators deployed in the natural environment, coupled with different techniques to report the produced data to a public or private cloud for further analysis. The constraints which stem from the specificities of such observatories however deviate from traditional IoT use-cases deployed in urban environments. Observatories are often created in remote locations, where energy supply, cellular networking coverage and human maintenance are more challenging than usual. Based on a survey of current practice in 34 observatories in France and abroad, I will make a case for the introduction of fog technologies in these systems, and outline a research agenda to adapt existing fog computing platforms to their specific requirements.

Bio: Guillaume Pierre is a Professor in Computer Science at the University of Rennes. Prior to this he spent 13 years at the VU University Amsterdam. His main interests are Fog computing, Cloud computing, and all other forms of large-scale distributed systems.  He took part in several European and EIT Digital projects and acted as the lead designer of the ConPaaS platform-as-a-Service environment and the coordinator of the FogGuru H2020 Maria-Sklodowska project. He is the leader of the Magellan research team at INRIA/IRISA.

Architectural Security for Critical Distributed Systems

Abstract: Distributed computing systems are found across the entire range of, e.g., critical infrastructures, telecommunications , and control systems. Distributed computing has ubiquitously penetrated our lives. However, securing these systems is incredibly difficult as we are dealing with heterogeneous hardware, a wide variety of software stacks, and almost infinite possibilities of interactions. Considerations for security and privacy, the development of threat models and mitigation strategies, must all be an integral part of the design and development process of any such system. A powerful class of security mechanisms are Trusted Execution Environments (TEEs), which involve processor extensions and software designs that provide the means to implement confidential computing in distributed systems. In theory this enables the processing of secret information on public infrastructure, e.g., across public data centres, and allows for notions of end-to-end security between embedded systems in the IoT or in critical control systems, and in the cloud. In practice, however, there are challenges to effectively use TEEs and to properly understand the resulting security guarantees.

In this lecture, I will give a general introduction to security extensions of modern processors. I will discuss how working with these processors is to be integrated in the software engineering life-cycle, and how distributed systems benefit from security related hardware support to implement security and privacy objectives. I will highlight challenging aspects regarding heterogeneity, the security of and on public infrastructures, and availability guarantees, with a bit of an emphasis on TEEs.

Bio: Jan Tobias Muehlberg works as a professor at Université libre de Bruxelles, École polytechnique (BE). He researches topics involving privacy, safety and security of information and communication systems, with particular interest in dependable embedded systems and secure critical ICT infrastructures, and in interdisciplinary research on questions around the responsible and sustainable development and use of ICTs. Jan Tobias is specifically interested in societal aspects of security and privacy in dependable systems, in trusted execution environments, and in security architectures for safety-critical embedded systems. Before joining ULB, Jan Tobias worked as a research manager at KU Leuven (BE), as a postdoc researcher at the University of Bamberg (DE), obtained a Ph.D. from the University of York (UK), and worked as a researcher at the University of Applied Sciences in Brandenburg (DE), where he also acquired his Master’s degree in Computer Science.

Trustworthy AI Systems

Abstract: Generative artificial intelligence (GAI) models, such as language models and diffusion models, are widely used for generating texts, images, tables, and graphs. There are increasing risks of abusing GAI to produce incorrect and adversarial contents. Watermarking GAI content is one of the essential solutions to govern the GAI applications and guardrail their misuse and harm to the society, even requested by the governmental policies. In this talk, I will discuss our ongoing work on post watermarking the content of different modalities from language models and diffusion models, i.e., synthetic text, tables, and graphs. We aim to design watermarking schemes to achieve objectives of having minimal degradation of the generated content quality, imperceptible to humans for avoiding alteration, detectable by machines for rigorous auditing, and robust against post editing attacks. I will highlight the key requirements and technical challenges with respect to the GAI models and the data modalities, through our preliminary results. 

Bio: Lydia Y. Chen is a Professor in the Department of Computer Science at the University of Neuchatel in Switzerland and Delft University of Technology in the Netherlands. Prior to joining TU Delft, she was a research staff member at the IBM Research Zurich Lab from 2007 to 2018. She holds a PhD from Pennsylvania State University and a BA from National Taiwan University. Her research interests are federated machine learning, generative AI, and their robustness and privacy management. She has published papers in peer-reviewed journals and serves on the technical program committees of system and AI conferences and the editorial boards on multiple IEEE Transactions journals.

Intersections of Fairness and Privacy: A Local Differential Privacy Perspective

Abstract:  This keynote explores the interplay between Differential Privacy (DP) and fairness in machine learning, with a particular focus on distributed systems through the lens of Local DP (LDP). On the one hand, DP provides a mathematical framework to protect individual data by introducing calibrated noise, ensuring privacy in centralized and distributed settings. On the other hand, fairness addresses biases/discrimination in machine learning models, ensuring equitable outcomes across sensitive demographic groups. In distributed systems, LDP offers a decentralized approach to privacy by perturbing data locally at the user level before aggregation, eliminating the need for a trusted curator. While this distributed paradigm enhances privacy guarantees, it presents unique challenges and opportunities for fairness and data utility. This talk examines how LDP can be leveraged to mitigate biases while maintaining utility, highlighting strategies for optimizing the trade-offs between privacy, fairness, and utility in distributed machine learning systems.

Bio: Héber H. Arcolezi is a full-time researcher at Inria Grenoble in France. He completed his Ph.D. in Computer Science at the University Bourgogne Franche-Comté in 2022, followed by a postdoctoral research position at Inria Saclay & École Polytechnique. His current research interests are on differential privacy and the ethical aspects of machine learning, including fairness and privacy issues.

Concurrence relâchée, mais concurrence prouvée

Abstract: Cette présentation portera sur plusieurs contributions dans le domaine des systèmes d’exploitation, avec un accent particulier sur la détection de bugs dans des environnements concurrents et l’optimisation des performances sur des infrastructures modernes.

Dans un premier volet, nous verrons une nouvelle approche pour inférer les interactions concurrentes dans des programmes multithreadés sans verrou. Ce travail repose sur l’appariement des barrières mémoire, permettant de détecter des fonctions potentiellement concurrentes dans des contextes dépourvus de verrous explicites. En appliquant cette méthode au noyau Linux, nous avons découvert 12 bugs critiques qui menaçaient la stabilité et l’intégrité des données, et dont les correctifs ont été intégrés depuis la version 5.15.

Dans un second volet, nous explorerons les limites des bases de données Key-Value (KV) sur les systèmes de stockage NVMe SSD modernes. Nous avons conçu KVell, un nouveau KV exploitant pleinement les capacités de ces disques, éliminant les limitations dues au CPU qui affectent les KVs existants. KVell, basé sur une architecture sans partage, permet des performances de lecture et d’écriture significativement plus rapides, surpassant l’état de l’art avec des gains de 2× en lecture et 5× en écriture.

Bio: Baptiste Lepers est titulaire d'une chaire de professeur junior à l'Inria (tenure track vers un poste de Directeur de Recherche). Son travail se concentre sur la preuve et l'amélioration des performances des systèmes concurrents. Il travaille à détecter des bugs dans Linux, à optimiser les performances du stockage, de la mémoire, des moteurs de graphes, des ordonnanceurs et des systèmes distribués. 

Graph geometrisation: How to monitor dynamic large scale graphs

Abstract: La surveillance de grands réseaux dynamiques est un défi majeur pour un large éventail d'applications. La complexité provient du fait que de légères modifications locales peuvent entraîner des variations importantee  des propriétés globales ; par exemple, dans certaines conditions, une simple coupure de lien peut entraîner la partition du graphe. En outre, il est souvent difficile de déterminer si un changement se propagera globalement ou restera local. Les mesures traditionnelles de la théorie des graphes, telles que la centralité ou l'assortativité du graphe, ne sont pas satisfaisantes pour caractériser les propriétés globales du graphe. Dans présentation, nous abordons le problème de la surveillance en temps réel des graphes dynamiques à grande échelle en développant une approche géométrique qui s'appuie sur les notions de courbure géométrique et sur les récents développements en géométrisation des graphes. Nous illustrons l’application  de ces méthodes à la surveillance des variations dynamiques de l'Internet mondial à l'aide d'informations sur les changements de topologie fournies par la combinaison de plusieurs flux BGP. En particulier, cette méthode permet détecter les événements et les changements majeurs par le biais de la géométrie de l'intégration du graphe.

Bio: Kavé Salamatian est professeur d'informatique à l'Université de Savoie. Ses principaux domaines de recherche sont la mesure et la modélisation de l'Internet, la sécurité des réseaux et la théorie de l'information sur les réseaux. Il était auparavant lecteur à l'université de Lancaster (Royaume-Uni) et professeur associé à l'université Pierre et Marie Curie. Kavé a obtenu son diplôme en 1998 à l'université Paris SUD-Orsay où il a travaillé sur le codage de canal à source commune appliqué à la transmission multimédia sur l'internet pour son doctorat. Dans une vie antérieure, il a obtenu un MBA, a travaillé sur le marché en tant qu'analyste de risque et a apprécié d'être un modélisateur de trafic urbain pendant quelques années. Il a été professeur invité à l'Académie chinoise des sciences. Il est titulaire d'un prix présidentiel de l'Académie chinoise des sciences pour 2018.  Ces jours-ci, il s'efforce de déterminer si la mise en réseau est une science ou un simple passe-temps et, si c'est une science, quels en sont les fondements. Il a publié plus de 170 articles.