Communication Dans Un Congrès Année : 2025

Evaluating the Confidentiality of Synthetic Clinical Texts Generated by Language Models

Évaluation de la confidentialité des textes cliniques synthétiques générés par des modèles de langue

Résumé

Large Language Models (LLMs) can be used to produce synthetic documents that mimic real documents when these are not available due to confidentiality or copyright restrictions. Herein, we investigate potential privacy breaches in automatically generated documents. We use synthetic texts generated from a pre-trained model fine-tuned on French clinical cases to evaluate potential privacy breaches according to three directions: (1) similarity between real, training corpus and synthetic corpus (2) strong correlations between clinical features in training and synthetic corpus and (3) Membership Inference Attack (MIA) using a fined tuned model on the synthetic corpus. We identify clinical feature associations that suggest strategies for filtering training corpus that could contribute to privacy preservation. Membership attacks were not conclusive.

Fichier principal
Vignette du fichier
AIME_Confidentiality_of_Synthetic_Clinical_Corpora.pdf (1.22 Mo) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)
Licence

Dates et versions

hal-05046326 , version 1 (25-04-2025)
hal-05046326 , version 2 (13-05-2025)

Licence

Identifiants

Citer

Foucauld Estignard, Sahar Ghannay, Julien Girard-Satabin, Nicolas Hiebel, Aurélie Névéol. Evaluating the Confidentiality of Synthetic Clinical Texts Generated by Language Models. 23rd International Conference on Artificial Intelligence in Medicine (AIME), Jun 2025, Pavie, Italy. ⟨10.1007/978-3-031-95838-0_13⟩. ⟨hal-05046326v1⟩
574 Consultations
338 Téléchargements

Altmetric

Partager

  • More