Séminaire ouvert à toutes les équipes de recherche IRISA et du centre Inria de l'Université du Rennes.
Dans le cadre du programme iVisit, Maël Le Treust accueillera Tobias Oechtering, Prof KTH Stockholm et Mengyuan Zhao, PhD à KTH Stockholm au sein de l'équipe ERMINE
Seminar #1: Tobias Oechtering, KTH Royal Institute of Technology, Stockholm, Sweden
10:00 - 11:00 Thursday 22th 2025 - room Rennes -Salle Lipari F202
Title: Rethinking Privacy: Challenges, Lessons, and Results from Developing the Pointwise Maximal Leakage Framework
Abstract: Differential privacy has become a widely adopted standard for ensuring data privacy, but its limitations—such as its impact model performance in learning applications and missing interpretability of its privacy parameter—spark interest into the exploration of alternative privacy frameworks. To this end, information-theoretic privacy measures often offer advantages with respect to utility and interpretability of guarantees. In particular, the pointwise maximal leakage (PML) framework has recently been shown to combine many of the advantages of differential privacy and the framework of quantitative information flow. In this talk, we will discuss some key goals, learnings and challenges encountered when developing the PML framework. Besides introducing the privacy measure and some main results, we will also focus on open challenges, and what is needed to enable its use in practical applications. These challenges offer insights into the fundamental issues and tradeoffs faced by privacy researchers and practitioners.
Bio: Tobias Oechtering is Professor at KTH Royal Institute of Sweden in the Division of Information Science and Engineering. He received his Dipl-Ing degree in Electrical Engineering and Information Technology in 2002 from RWTH Aachen University, Germany, his Dr-Ing degree in Electrical Engineering in 2007 from the Technische Universität Berlin, Germany. In 2009, he received the ``Förderpreis 2009” from the Vodafone Foundation. Dr. Oechtering is currently Senior Editor of IEEE Transactions on Information Forensic and Security since May 2020 and served previously as Associate Editor for the same journal since June 2016, and IEEE Communications Letters during 2012-2015. He has served on numerous technical program committees for IEEE sponsored conferences, and he was general co-chair for IEEE ITW 2019. His research interests include privacy and physical layer security, statistical learning and signal processing, communication and information theory, as well as communications for networked control.
Seminar #2: Mengyuan Zhao, KTH Royal Institute of Technology, Stockholm, Sweden
11:00 - 12:00 Thursday 22th 2025 - room Rennes-Salle Lipari F202
Title: Coordination coding for decentralized decision-making problems: the Witsenhausen counterexample
Abstract: We investigate the long-standing open problem of decentralized control posed by Witsenhausen in 1968. This famous counterexample illustrates the suboptimality of affine strategies in the Linear Quadratic Gaussian (LQG) settings when the information available at the decision-makers is asymmetric. Recent advancements in studying the Witsenhausen counterexamples have been driven by information theoretic methods for its vector-valued extensions. This enables the adoption of advanced coordination coding techniques that allow to characterize the set of achievable Witsenhausen power costs and estimation costs. In particular, we propose a new scheme that has zero estimation cost.
Bio: Mengyaun Zhao is Ph.D. student under the supervision of Prof. Tobias Oechtering and the co-supervision of Dr. Maël Le Treust on the topic of information theory of stochastic decision problems since June 2023. Before joining KTH, she received the master's degree of applied mathematics at ETH Zürich. She is interested in classical information theory using various mathematical tools, e.g. statistics, analysis, graph theory, etc.. At present, her research focuses on deriving fundamental limits and designing coordination-coding schemes for a distributed decision-making problem that was initially formulated by Hans Witsenhausen as a famous counterexample in 1968