Bogdan Kulynych

/bogh-duhn koo-lin-ich/

I am a researcher studying privacy, security, reliability, and broader social implications of algorithmic systems. I am currently a research scientist at the Lausanne University Hospital (Switzerland) in the Clinical Data Science group. I hold a Ph.D. in Computer Science from EPFL (Switzerland) where I was advised by Carmela Troncoso, and a B.Sc. in Applied Mathematics from Kyiv Mohyla Academy (Ukraine). Before that, I had the opportunity to be a visiting fellow at Harvard University with Flavio du Pin Calmon, and intern at Google and CERN.

Bluesky: @bogdankulynych
Twitter: @hiddenmarkov (not that active anymore)
Email: [first name] at kulyny.ch

Latest and Upcoming Talks

• Feb 24, 2025 Interpretable and Operational Data Privacy Protection in Machine Learning
  Nokia Bell Labs Responsible AI Seminar.
  
• Dec 17, 2024 Attack-Aware Noise Calibration for Differential Privacy
  Harvard Privacy Tools Seminar (Harvard SEAS @ 11am).
  
• Dec 12, 2024 Attack-Aware Noise Calibration for Differential Privacy
  NeurIPS. Poster presentation (Evening West session).
  
• Nov 18, 2024 Attack-Aware Noise Calibration for Differential Privacy
  Imperial College London (Yves-Alexandre de Montjoye's group).
  
• Sep 20, 2024 Fundamental Limits of Least-Privilege Learning
  University of Toronto (Nicolas Papernot's group).
  

Selected Publications

(ε,δ) Considered Harmful: Best Practices for Reporting Differential Privacy Guarantees
J. Gomez, B. Kulynych, G. Kaissis, J. Hayes, B. Balle, A. Honkela. Preprint, 2025
[arxiv]

A Scoping Review of Privacy and Utility Metrics in Medical Synthetic Data
B. Kaabachi, J. Despraz, T. Meurers, K. Otte, M. Halilovic, B. Kulynych, F. Prasser, J. Raisaro. npj Digital Medicine, 2025
[medRxiv]

Attack-Aware Noise Calibration for Differential Privacy
B. Kulynych*, J. Gomez*, G. Kaissis, F. Calmon, C. Troncoso. NeurIPS, 2024
[arxiv] [twitter thread]

The Fundamental Limits of Least-Privilege Learning
T. Stadler*, B. Kulynych*, N. Papernot, M. Gastpar, C. Troncoso. ICML, 2024
[arxiv]

Prediction without Preclusion: Recourse Verification with Reachable Sets
A. Kothari*, B. Kulynych*, T. Weng, B. Ustun. ICLR (spotlight), 2024
[arxiv]

Arbitrary Decisions are a Hidden Cost of Differentially Private Training
B. Kulynych, H. Hsu, C. Troncoso, F. Calmon. FAccT, 2023
[arxiv] [twitter thread]

Adversarial Robustness for Tabular Data through Cost and Utility Awareness
K. Kireev*, B. Kulynych*, C. Troncoso. NDSS, 2023
[arxiv] [twitter thread]

What You See is What You Get: Principled Deep Learning via Distributional Generalization
B. Kulynych*, Yao-Yuan Yang*, Y. Yu, J. Blasiok, P. Nakkiran. NeurIPS, 2022
[arxiv] [video] [twitter thread]

Disparate Vulnerability to Membership Inference Attacks
B. Kulynych, M. Yaghini, G. Cherubin, M. Veale, C. Troncoso. PETS, 2022
[arxiv] [twitter thread]

Media Coverage and Comments

• Twitter algorithmic bias challenge: The Guardian, Wired, The Verge, CNET
• Premise app safety concerns in the beginning of 2022 Russia's invasion of Ukraine: The Wall Street Journal