Profile picture for Igor Shilov

Google Scholar / Website / LinkedIn / Twitter

Igor Shilov

Email: i.shilov23@imperial.ac.uk

Igor is a PhD student at the AI Security and Privacy Lab having started in October 2023. His research interests include Differential Privacy, Privacy Attacks against ML models and the impact of privacy-preserving technologies on privacy regulations.

Prior to joining the lab, Igor obtained his undergraduate degree in Computer Science in 2013 and has been working as a Software Engineer since, most recently at Meta AI. Working in a Research Engineer capacity in Ilya Mironov’s group focusing on privacy-preserving technologies, Igor has picked up a passion for research, privacy and open-source.

During his time at Meta AI Igor has lead the development of Opacus, a PyTorch Library for training models with Differential Privacy (DP-SGD), designed the architecture of StopNCII.org - a privacy-preserving platform helping combat non-consensual intimate image sharing, and contributed to various R&D projects on Differential Privacy, Federated Learning and Privacy Attacks.

Igor’s prior experience also includes work on Trust & Safety, Recommender Systems and Information Retrieval.

News

Publications

  • Hayes, J., Shumailov, I., Choquette-Choo, C. A., Jagielski, M., Kaissis, G., Nasr, M., Ghalebikesabi, S., Muthu Selva Annamalai, M. S., Mireshghallah, N., Shilov, I., Meeus, M., de Montjoye Y. A., K. Lee, F. Boenisch, A. Dziedic and A. F. Cooper Exploring the limits of strong membership inference attacks on large language models. The Thirty-ninth Annual Conference on Neural Information Processing Systems (NeurIPS) (2025).

  • Dodd, E., Krčo, N., Shilov, I. and de Montjoye Y. A. The Tail Tells All: Estimating Model-Level Membership Inference Vulnerability Without Reference Models. ArXiv preprint (2025).

  • Meeus, M., Shilov, I., Kaissis, G. and de Montjoye Y. A. Counterfactual Influence as a Distributional Quantity. Workshop on The Impact of Memorization on Trustworthy Foundation Models (MemFM) @ ICML 2025 (2025).

  • Wicker, M., Sosnin, P., Shilov, I., Janik, A., Müller, M. N., de Montjoye, Y. A., Weller, A. and Tsay, C. Certification for Differentially Private Prediction in Gradient-Based Training. 42nd International Conference on Machine Learning (ICML) (2025).

  • Meeus, M., Shilov, I., Jain, S., Faysse, M., Rei, M. and de Montjoye Y. A. SoK: Membership Inference Attacks on LLMs are Rushing Nowhere (and How to Fix It). IEEE Conference on Secure and Trustworthy Machine Learning (SaTML) (2025).

  • Pollock, J., Shilov, I., Dodd, E. and de Montjoye Y. A. Free Record-Level Privacy Risk Evaluation Through Artifact-Based Methodsa. 34th USENIX Security Symposium (USENIX Security 2025) (2025).

  • Meeus, M., Shilov, I., Faysse, M. and de Montjoye Y. A. Copyright Traps for Large Language Models. 41st International Conference on Machine Learning (ICML 2024) (2024).
    Selected Press: MIT Technology Review, Nature News

  • Meeus, M., Shilov, I., and de Montjoye Y. A. Mosaic Memory: Fuzzy Duplication in Copyright Traps for Large Language Models. ArXiv preprint (2024).