Saved in:
Bibliographic Details
Main Authors: Even, Mathieu, Berenfeld, Clément, Bleistein, Linus, Cebere, Tudor, Josse, Julie, Bellet, Aurélien
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2602.02819
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913177361448960
author Even, Mathieu
Berenfeld, Clément
Bleistein, Linus
Cebere, Tudor
Josse, Julie
Bellet, Aurélien
author_facet Even, Mathieu
Berenfeld, Clément
Bleistein, Linus
Cebere, Tudor
Josse, Julie
Bellet, Aurélien
contents Membership Inference Attacks (MIAs) aim to distinguish training points (members) from unseen data (non-members), and are widely used to quantify memorization and assess privacy risks. Standard MIA evaluation requires repeated retraining, which is computationally costly for large models. One-run (single training with randomized data inclusion) and zero-run (post hoc evaluation) methods are often used instead, but their statistical validity remains unclear. We address this gap by framing MIA evaluation as a causal inference problem, defining \emph{memorization as the causal effect of including a data point in the training set}. This novel formulation reveals and formalizes key sources of bias in existing protocols: one-run methods suffer from interference between jointly included points, while zero-run evaluations are additionally confounded by distribution shift between member and non-member evaluation data. We derive causal analogues of standard MIA metrics and propose practical estimators for multi-run, one-run, and zero-run regimes with non-asymptotic consistency guarantees. We validate our approach in several settings, including pretrained and fine-tuned LLMs, showing that it enables reliable measurement of MIA performance without retraining and under distribution shift. Overall, our framework provides a principled foundation for privacy evaluation in modern AI systems.
format Preprint
id arxiv_https___arxiv_org_abs_2602_02819
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Causal Evaluation of Membership Inference Attacks
Even, Mathieu
Berenfeld, Clément
Bleistein, Linus
Cebere, Tudor
Josse, Julie
Bellet, Aurélien
Machine Learning
Membership Inference Attacks (MIAs) aim to distinguish training points (members) from unseen data (non-members), and are widely used to quantify memorization and assess privacy risks. Standard MIA evaluation requires repeated retraining, which is computationally costly for large models. One-run (single training with randomized data inclusion) and zero-run (post hoc evaluation) methods are often used instead, but their statistical validity remains unclear. We address this gap by framing MIA evaluation as a causal inference problem, defining \emph{memorization as the causal effect of including a data point in the training set}. This novel formulation reveals and formalizes key sources of bias in existing protocols: one-run methods suffer from interference between jointly included points, while zero-run evaluations are additionally confounded by distribution shift between member and non-member evaluation data. We derive causal analogues of standard MIA metrics and propose practical estimators for multi-run, one-run, and zero-run regimes with non-asymptotic consistency guarantees. We validate our approach in several settings, including pretrained and fine-tuned LLMs, showing that it enables reliable measurement of MIA performance without retraining and under distribution shift. Overall, our framework provides a principled foundation for privacy evaluation in modern AI systems.
title Causal Evaluation of Membership Inference Attacks
topic Machine Learning
url https://arxiv.org/abs/2602.02819