16ème Atelier sur la Protection de la Vie Privée (APVP 2026)

7:30 - 9:00 (1h30)

Petit déjeuner

9:00 - 10:00 (1h)

Keynote 1: Building Effective Privacy-Preserving Data Analytics with PrivaGym

Apostolos Pyrgelis (RISE Research Institutes of Sweden)

https://apvp2026.sciencesconf.org/data/program/apostolos_pyrgelis_keynote.txt

10:00 - 10:40 (40min)

Session A1 -- Privacy vulnerabilities: data and models

Salle 1

› A Dual Perspective on Synthetic Trajectory Generators: Utility Framework and Privacy Vulnerabilities - Aya CHERIGUI, Franche-Comté Électronique Mécanique, Thermique et Optique - Sciences et Technologies (UMR 6174), Orange Research 10:00-10:20 (20min)

› Beyond the mask : The illusion of privacy in defaced brain MRI - Imen Bajar, Centre Inria de Lyon - Mohamed Maouche, Centre Inria de Lyon 10:20-10:40 (20min)

10:40 - 11:10 (30min)

Pause café

11:10 - 12:50 (1h40)

Session A2 -- Privacy vulnerabilities: data and models

› Uncovering Memorization in Timeseries Imputation models: LBRM Membership Inference and its link to attribute Leakage - Braham TALEB, EDF Labs, Services répartis, Architectures, MOdélisation, Validation, Administration des Réseaux 11:10-11:30 (20min)

› Mobility Traces under Attack: How LLMs Automate Privacy and Social Engineering Attacks - Marwa EL KAMIL, Inria, Ecole de Technologie Supérieure [Montréal] 11:30-11:50 (20min)

› Practical and Ready-to-Use Methodology to Assess the re-identification Risk in Anonymized Datasets - Louis Philippe Sondeck, COACHMESEC Consulting (Clever Identity) - maryline laurent, SAMOVAR, Télécom SudParis, Institut Polytechnique de Paris 11:50-12:10 (20min)

› Inferring Sensitive Attributes from Knowledge Graph Embeddings: Attack and Defense Strategies - Yasmine Hayder, Institut National des Sciences Appliquées - Centre Val de Loire, LIFO, Concevoir des technologies d'amélioration de la vie privée explicables et efficaces 12:10-12:30 (20min)

› Cryptographic Traceability and Verifiability of Sensitive Security Data Before Generative AI Processing - Yulliwas AMEUR, CEDRIC. Réseaux et Objets Connectés, EFREI Paris 12:30-12:50 (20min)

12:50 - 14:00 (1h10)

Déjeuner

14:00 - 15:00 (1h)

Table Ronde

15:00 - 15:40 (40min)

Session B1 -- Privacy-preserving machine learning

› Improving Unlearning with Model Updates Probably Aligned with Gradients - Virgile Dine, Centre Inria de l'Université de Rennes 15:00-15:20 (20min)

› Privacy-Utility Trade-off under Rate Constraint in Federated Learning - Mamour Ndiaye, LISTIC 15:20-15:40 (20min)

15:40 - 16:10 (30min)

Pause café

16:10 - 17:10 (1h)

Session B2 -- Privacy-preserving machine learning

› Learning with Locally Private Examples by Inverse Weierstrass Private Stochastic Gradient Descent - Jean Dufraiche, MAGNET (Machine Learning in Information Networks) 16:10-16:30 (20min)

› Unified Privacy Guarantees for Decentralized Learning via Matrix Factorization - Dimitri Lerévérend, Centre Inria de l'Université de Rennes, the World Is Distributed Exploring the tension between scale and coordination 16:30-16:50 (20min)

› Titre : Google, l'apprentissage fédéré et la protection de la vie privée, ou comment redéfinir un problème pour mieux l'invisibiliser. Sous-titre : Comprendre l'institution d'un cadrage par l'analyse de controverse. - Marie Garin, Centre Internet et Société 16:50-17:10 (20min)

17:15 - 18:15 (1h)

Session C -- Applied cryptography and privacy-preserving distributed algorithms

› Opening Pandora's Box: White-Box Attacks on Microsoft's PhotoDNA Perceptual Hash Function - Diane Leblanc-Albarel, COSIC, KU Leuven 17:15-17:35 (20min)

› POPPY: Scalable and Secure Spectral Centrality for Distributed Graphs via Homomorphic Encryption - Claire Guichemerre, Université de Rennes 17:35-17:55 (20min)

› Giskard : Scalable Byzantine-Robust and Privacy-Preserving Decentralized Learning - Ousmane Touat, CNRS 17:55-18:15 (20min)

19:30 - 22:00 (2h30)

Dîner

Programme