Pith. sign in

REVIEW 3 cited by

Privacy Auditing of Large Language Models

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2503.06808 v1 pith:CJA6FM66 submitted 2025-03-09 cs.CR cs.AIcs.LG

Privacy Auditing of Large Language Models

classification cs.CR cs.AIcs.LG
keywords privacycanariesmodelsauditingllmsmodelpriorapproach
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Current techniques for privacy auditing of large language models (LLMs) have limited efficacy -- they rely on basic approaches to generate canaries which leads to weak membership inference attacks that in turn give loose lower bounds on the empirical privacy leakage. We develop canaries that are far more effective than those used in prior work under threat models that cover a range of realistic settings. We demonstrate through extensive experiments on multiple families of fine-tuned LLMs that our approach sets a new standard for detection of privacy leakage. For measuring the memorization rate of non-privately trained LLMs, our designed canaries surpass prior approaches. For example, on the Qwen2.5-0.5B model, our designed canaries achieve $49.6\%$ TPR at $1\%$ FPR, vastly surpassing the prior approach's $4.2\%$ TPR at $1\%$ FPR. Our method can be used to provide a privacy audit of $\varepsilon \approx 1$ for a model trained with theoretical $\varepsilon$ of 4. To the best of our knowledge, this is the first time that a privacy audit of LLM training has achieved nontrivial auditing success in the setting where the attacker cannot train shadow models, insert gradient canaries, or access the model at every iteration.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 3 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Advancing the State-of-the-Art in Empirical Privacy Auditing

    cs.LG 2026-06 unverdicted novelty 6.0

    Proposes high-temperature synthetic canaries and auxiliary-model auditing to improve empirical privacy measurement for LLM fine-tuning and synthetic data generation.

  2. Agents That Know Too Much: A Data-Centric Survey of Privacy in LLM Agents

    cs.CR 2026-06 unverdicted novelty 5.0

    A data-centric survey finds that only information-flow control covers compositional and cross-session leakage in LLM agents and that no single benchmark tests an agent across all its data surfaces under one policy.

  3. Barriers to Evidence in AI-Related Cases and the Privatization of Proof

    cs.CY 2026-05 unverdicted novelty 5.0

    The paper identifies seven asymmetries in access to AI evidence and proposes a three-part test for courts to resolve disclosure disputes using proportionality and reasonable alternatives.