NAMESAKES: Probing Identity Memorization in Text-to-Image Models
This work addresses the privacy concern of identity memorization in T2I models for users and regulators, providing a practical tool for auditing without model access.
The authors introduce a black-box behavioral probe to distinguish whether a generated face is memorized or fabricated in text-to-image models, without requiring ground-truth photos or training data. They present the NAMESAKES dataset and show their probe substantially predicts identity memorization across state-of-the-art T2I models.
Text-to-image (T2I) models generate realistic likenesses of some individuals when prompted with their names, raising privacy concerns. However, distinguishing whether a generated face is memorized or fabricated currently requires ground-truth photos, access to training data, or white-box access to model internals, limiting applicability. We introduce a fully black-box behavioral probe that distinguishes between these regimes while requiring no reference photos or prior knowledge of training data. To benchmark this task, we present the NAMESAKES dataset of over one thousand names and faces of public figures spanning a wide range of fame levels, along with perturbed, less famous names. Experiments on state-of-the-art T2I models show that our probe substantially predicts identity memorization and separates memorized from unrecognized names, with further insights into differences across model families.