Introduces natural identifiers (NIDs) from common training data to support post-hoc differential privacy auditing and dataset inference for LLMs without retraining or private held-out sets.
Blind baselines beat membership inference attacks for foundation models
3 Pith papers cite this work. Polarity classification is still indexing.
3
Pith papers citing it
verdicts
UNVERDICTED 3representative citing papers
Empirical benchmarks show distribution similarity between adaptation and pretraining data increases practical privacy leakage in DP-adapted LLMs at fixed theoretical guarantees, with LoRA providing strongest protection for OOD cases.
CatShift detects training data membership in LLMs by comparing output shifts induced by fine-tuning on member versus non-member data, relying on catastrophic forgetting without requiring logit access.
citing papers explorer
-
Natural Identifiers for Privacy and Data Audits in Large Language Models
Introduces natural identifiers (NIDs) from common training data to support post-hoc differential privacy auditing and dataset inference for LLMs without retraining or private held-out sets.
-
Benchmarking Empirical Privacy Protection for Adaptations of Large Language Models
Empirical benchmarks show distribution similarity between adaptation and pretraining data increases practical privacy leakage in DP-adapted LLMs at fixed theoretical guarantees, with LoRA providing strongest protection for OOD cases.
-
Hey, That's My Data! Token-Only Dataset Inference in Large Language Models
CatShift detects training data membership in LLMs by comparing output shifts induced by fine-tuning on member versus non-member data, relying on catastrophic forgetting without requiring logit access.