Program 2024

The impact of the advances in generative models on applications and systems

Abstract:  Generative models have achieved unprecedented quality levels across a wide range of data types. This advance often stems from the ever increasing data and compute used to train larger and larger models. One major use case of such synthetic data is in privacy-compliant data sharing. Gartner predicts that synthetic data will reduce by 2025 the need for real data by 70% for analytics and machine learning. We will look at generative models, with a special focus on tabular data, and the issue of democratization of large model training.

Bio: Robert Birke is a tenured assistant professor in the Parallel Computing research group at the University of Torino. He received his Ph.D. in Electronics and Communications Engineering from the Politecnico di Torino, Italy, (2009). Previously he has been a visiting researcher at IBM Research Zurich, Switzerland, and a Principal Scientist at ABB Corporate Research, Switzerland. His research interests are in the broad area of  AI and big-data application optimization, network design, and workload characterization. He has published more than 100 papers at venues related to machine learning, communication, and system performance,  e.g., ECML PKDD, PAKDD, ACML, WWW, SIGCOMM, FAST, INFOCOM, and JSAC.

Security in the evolved wireless system Era

Abstract:  The recent technological evolution in wireless systems is with an unprecedented impact on the daily life of users, by providing new services that would have not been imagined just some few years ago. The always-connected paradigm is enabler of new applications, but with enormous amount of data exchanged among connected devices, it is music to attacker’ ears to realize more sophisticated attacks. Generally, there is a plethora of different and heterogeneous devices, where it is complicated to adopt conventional detection and countermeasure solutions. Moreover, together with technological advances, the cyber-attacks also evolve and become more effective and less detectable.

Some intelligent attacks will be presented, by showing their capability to rely on the vulnerability of the communication technologies and protocols. This type of attacks is demanding a radical paradigm shift on the countermeasure solutions, that need to be more advanced, to counter in an effective way these smart evolved attacks.

Bio: Valeria Loscri is a permanent researcher in the FUN Team at Inria Lille (France) since Oct. 2013. From Dec. 2006 to Sept. 2013, she was Research Fellow in the TITAN Lab of the University of Calabria, Italy. She received her MSc and PhD degrees in Computer Science in 2003 and 2007, respectively, from the University of Calabria and her HDR (Habilitation à diriger des recherches) in 2018 from Université de Lille (France). Her research interests focus on emerging technologies for new communication paradigms such as Visible Light Communication (VLC), mmWave, cooperation and coexistence of wireless heterogeneous devices and cyber security in wireless networks. She is involved in the activity of several European Projects (Horizon Europe MLSysOps, H2020 CyberSANE, FP7 EU project VITAL, etc.). She has been nominated to the 2021 Women Stars in Computer Networking and Communications by the IEEE Communica0on Society. She is in the editorial board of IEEE COMST IEEE Transactions on Information Forensics and Security (TIFS), Elsevier ComNet, ComCom. She is Action Chair and scientific Holder of BEiNG-WISE COST Action (since 2023). She is serving as TPC members in several primary international conferences, such as IEEE ESORICS, IEEE CNS, IEEE INFOCOM, IEEE PerCom. Since 2019, she is Scien0fic International Delegate for Inria Lille.

Reproducible Research in Computer Science 

Abstract: Open science and reproducible research are becoming the new golden standard in most domains of science and especially in computer science. But what are we talking about exactly ? Why is this happening and what are the tools and methods that researchers are now expected to master ? Beyond these tools, are there any specific needs in distributed computing/networking/operating system research ? Hopefully, you'll have a better understanding at the end of this talk.

Bio: Arnaud Legrand is a senior CNRS researcher at University Grenoble-Alpes since 2004. His research interests encompass the study of large scale distributed computing systems, their optimization (scheduling, combinatorial optimization, and game theory), and performance evaluation (trace analysis, performance prediction, statistical learning), in particular through simulation. He is one of the core developers of the SimGrid project and he is involved in the promotion of better research practices and methods, in particular through the MOOC "Reproducible Research: methodological principles for a transparent science".

Design your processor and build it at home

Abstract:  The talk will be divided into two parts, both about processor designs.

The first part concerns classic processors. For a computer scientist, his/her machine is a mysterious piece of technology, a black box he/she prefers not to open. However, it is better to know what is hidden under the hood. First because as scientists, we are curious. Second, because it is the only way to keep the hardware efficient. Third, because today, we have tools to go deep in the understanding of the processor design. This part of the talk will present such tools, with the RISC-V Instruction Set Architecture, hardware design languages, high level synthesis, synthesizors and eventually, boards and programmable chips to really implement your own processor at home.

The second part concerns my new activity at Cortus as an engineer. The job of a researcher goes from the idea to the paper and the job of the engineer starts with a design (often imagined from published papers) and ends with a product. I will present a Cortus project aiming to build an accelerator designed to run a YOLO-like application, i.e. a Deep Neural Network to recognize patterns in a video stream. The resulting chip should be cheaper and more energy effcient than a from-the-shelf GPU.

Bio: Bernard Goossens is an Emeritus Professor at the University of Perpignan in France, member of the LIRMM laboratory. He is also an R&D engineer at Cortus, a French company designing, implementing and selling processors, micro-controllers and SoCs. He has more than 30 years of experience in teaching computer architecture and processor microarchitecture. He is the author of many journal and conference publications on microarchitectural parallel designs. He is also the author of “Architecture et microarchitecture des processeurs” (in French) and "Guide to Computer Processor Architecture A RISC-V Approach, with High-Level Synthesis", both published by Springer.

Why software is eating the planet

Abstract:  Given the proliferation of online services that invade our daily lives, the question of sustainable development is more relevant than ever for software engineering. Indeed, the massive deployment of digital services has contributed to the explosion of energy consumption in this field over the past years, with forecasts targeting 20% of global consumption by 2025. Without overlooking the numerous challenges posed by this evolution in usage, reducing the environmental footprint of software services remains a particularly critical challenge given the limited availability of our planet's resources. The interdisciplinary nature of this research theme requires coordinated efforts in many areas to achieve significant gains. During this presentation, I will strive to illustrate the various dimensions of this research theme through the work that our team has undertaken in recent years.

Bio: Romain Rouvoy is Full Professor at the Faculty of Science and Technology of the University of Lille. He is also a member of the Inria Spirals project-team. His research focuses on the intersection of software engineering and distributed systems, addressing issues such as sustainable software development and privacy preservation.

Algorithmic audits of AIs: What do we know (we don't know)?

Abstract:  Many of the most impactful online algorithms (recommendation, rating, ranking and classification) are now being hosted by third-party providers, without users having the slightest idea of how they operate and produce decisions from their data. Such a black-box configuration challenges scientists to understand what is feasible in terms of auditing these (by querying these algorithms, observing responses and inferring properties), both theoretically and in terms of practical effectiveness. This presentation will review the state of knowledge that can be mobilized to build realistic audits for some of these algorithms.

Bio: Erwan Le Merrer worked as a Senior Scientist at Technicolor from 2008 to 2019. He is currently a Researcher at Inria, France. He is also leading the Scientific Council of the Société Informatique de France. He is involved in the algorithmic audit of AI-based decision-making algorithms.