Valentin Dorseuil

I am a second-year PhD student in the DATA Team at the Computer Science Department (DI ENS) of École Normale Supérieure (ENS) - PSL, under the supervision of Olivier Cappé and Jamal Atif. My research focuses on Machine Learning, specifically Differential Privacy and Membership Inference Attacks.

In Summer 2026, I was a visiting PhD student at the Vector Institute and the University of Toronto, hosted by Nicolas Papernot in the CleverHans Lab.

Before joining ENS, I graduated from École Polytechnique (Cycle Ingénieur, X2021), where I majored in Applied Mathematics with a minor in Computer Science. I also hold a Master’s degree in Machine Learning from Université Paris-Saclay.

I am always happy to discuss research or potential collaborations, so feel free to reach out!

Email: • GitHub • Scholar • LinkedIn

Selected publications

Preprint

Reliable Estimation of Differential Privacy under Selection Bias

Valentin Dorseuil, Meenatchi Sundaram Muthu Selva Annamalai, Jamal Atif, Olivier Cappé, and Nicolas Papernot

arXiv preprint, 2026

arXiv Bib PDF

@article{dorseuil2026ReliableEstimation,
  title = {Reliable Estimation of Differential Privacy under Selection Bias},
  author = {Dorseuil, Valentin and Annamalai, Meenatchi Sundaram Muthu Selva and Atif, Jamal and Cappé, Olivier and Papernot, Nicolas},
  year = {2026},
  archiveprefix = {arXiv},
  journal = {arXiv preprint},
  url = {https://arxiv.org/abs/2610.27021},
}

NeurIPS

Assessing Per-Sample Membership Inference Vulnerability without Retraining

Valentin Dorseuil, Jamal Atif, and Olivier Cappé

In Advances in Neural Information Processing Systems (NeurIPS), 2026

arXiv Bib PDF Poster

@inproceedings{dorseuil2026AssessingPerSample,
  title = {Assessing Per-Sample Membership Inference Vulnerability without Retraining},
  author = {Dorseuil, Valentin and Atif, Jamal and Cappé, Olivier},
  year = {2026},
  booktitle = {Advances in Neural Information Processing Systems (NeurIPS)},
  url = {https://arxiv.org/abs/2602.15919},
}

View all publications →

Latest news

Oct 19, 2026	[Preprint]: We released our latest preprint, Reliable Estimation of Differential Privacy under Selection Bias, on arXiv! In this work, we identify and fix a selection bias found in standard privacy auditing protocols. Read the full paper here! 📄
Oct 16, 2026	[Preprint]: Our new preprint, Detectability without Interference: Efficient Canary Design for Privacy Auditing in One Run, is live on arXiv. It frames canary optimization as a bi-level problem and introduces IBIS, a framework that efficiently crafts multiple canaries for privacy audits in one run. You can check it out here!📄
Oct 09, 2026	[Paper]: Our paper, Assessing Per-Sample Membership Inference Vulnerability without Retraining, has been accepted at NeurIPS 2026. It’s available here. See you in Paris! 🎉