arXiv:2611.15578 · v1
Constructing and Validating Memorisation Audits for EHR Foundation Models
摘要
Autoregressive electronic health record (EHR) foundation models may disclose training information at inference, yet clinical predictability makes patient-specific memorisation difficult to distinguish from generalisation. We evaluate fixed-window token reconstruction, sensitive-attribute disclosure, and random patient-level canaries in GPT-2-style models trained on MIMIC-IV, across early, best-validation, and final checkpoints. Reconstruction improves with additional context, but training--held-out differences remain small. Trajectory-unique prompts containing limited clinical information produce no excess sensitive disclosure for training patients. In a matched model with random binary patient-level canaries, training-patient label-recovery AUROC increases from 0.512 to 0.549 while held-out recovery remains at chance. An overfit 1,000-patient positive control achieves near-perfect recovery, confirming sensitivity to strong memorisation. These results reveal weak cohort-level memorisation but no reliable patient-level disclosure under the evaluated attacks. They motivate clinically grounded inference-time audits that can detect patient-specific memorisation in existing models without retraining with canaries.
作者备注：Accepted to the NeurIPS 2026 Workshop on Beyond Private Training: The New Landscape of AI Privacy (InfPriv) and the Machine Learning for Health (ML4H) 2026 Symposium, Findings Track