The paper evaluates membership inference attacks (MIAs) on NLP text classifiers using the GLUE SST‑2 sentiment dataset. It compares a TF‑IDF + Logistic Regression pipeline with a fine‑tuned DistilBERT model under a loss‑threshold MIA, finding that both models leak membership signals despite high accuracy. The study also tests mitigations, showing that stronger regularization reduces leakage for Logistic Regression at a utility cost, while fine‑tuning DistilBERT for fewer epochs lowers leakage with minimal accuracy loss.
arXiv:2606. 17464v1 Announce Type: new Abstract: Membership inference attacks (MIAs) are a canonical way to assess a machine learning model's privacy properties.
arXiv:2607. 28862v1 Announce Type: cross Abstract: The rapid development of Large Language Models (LLMs) has led to significant advances across a wide range of language tasks, while simultaneously raising growing concerns about unauthorized data exploitation and privacy leakage.
arXiv:2410. 06814v2 Announce Type: replace Abstract: Over-parameterized models are typically vulnerable to membership inference attacks, which aim to determine whether a specific sample is included in the training of a given model.
Conformal Privacy Auditing (CPA) is a distribution‑free framework that calibrates re‑identification risk for each released document against large language model (LLM)‑empowered adversaries. It outputs a conformal ambiguity set of candidate identities that is guaranteed to contain the true identity with a user‑chosen confidence level under exchangeability, along with an interpretable leakage proxy derived from the set size. CPA supports both logit‑access and sampling‑only attackers, enabling audits of both open‑source and proprietary models, and demonstrates calibrated coverage across various benchmarks and attacker configurations.
arXiv:2606. 09401v1 Announce Type: new Abstract: Recent work has applied differential privacy (DP) to adapt large language models (LLMs) for sensitive applications, offering theoretical guarantees.
The paper investigates whether inexpensive spectral metrics from the heavy‑tailed self‑regularisation framework can predict membership inference attack (MIA) vulnerability, offering a scalable alternative to costly shadow‑model attacks. Experiments on image and tabular classification tasks show that stable rank correlates positively with overall MIA success, while Log alpha‑Norm correlates negatively with MIA risk in low false‑positive regimes, outperforming conventional generalisation gap measures. These findings suggest that neural network spectra contain privacy leakage signals not captured by traditional overfitting metrics, pointing to spectral analysis as a promising direction for privacy auditing.
The paper introduces Word-level Probability MIA (WPMIA), a black-box membership inference attack that estimates word-level generation probabilities via Monte Carlo sampling and local kernel smoothing, then aggregates them into a sequence-level likelihood estimator. By conditioning on different prefixes, WPMIA amplifies distributional differences between member and non-member texts, outperforming existing black-box baselines on open-source LLMs and achieving an average TPR@5%FPR of 42.0 on proprietary models such as GPT‑5‑Chat, Gemini‑2.5‑Flash, and Claude‑4.5‑Haiku.
QuanText is a training‑free, large‑language‑model‑agnostic mechanism for releasing textual datasets that protects dataset‑level secrets such as the proportion of records with a particular diagnosis or gender. It perturbs both the secret distribution and correlated attribute distributions by selecting candidate release distributions close to the private empirical distribution and rewriting each text sample to match the chosen distribution using attribute‑related snippets. The method is inspired by the Statistic Maximal Leakage framework and, under idealized conditions, satisfies an SML guarantee, while empirical evaluations show a superior privacy‑utility trade‑off compared to existing data generation baselines.
The paper introduces Pairwise Likelihood MIA (PL‑MIA), a unified membership inference attack that combines a Gaussian likelihood‑ratio statistic with population calibration and the Cauchy combination test. PL‑MIA generates p‑values from pairwise comparisons between a query point and reference points, then aggregates these continuous signals using the Cauchy test to preserve evidence strength. Experiments show that PL‑MIA surpasses strong baselines, boosting true positive rates by over 25% in low‑false‑positive settings, thereby validating the theoretical advantages of the proposed statistical framework.
arXiv:2608. 00144v2 Announce Type: replace Abstract: Membership inference (MIA) on language models is usually summarised by aggregate ROC-AUC, but such evaluations are confounded: model-free blind baselines can separate members from non-members using surface text alone.
arXiv:2310. 16152v5 Announce Type: replace-cross Abstract: Federated learning (FL) has become a key component in various language modeling applications such as machine translation, next-word prediction, and medical record analysis.
arXiv:2607. 16620v1 Announce Type: cross Abstract: Differential privacy (DP) is increasingly deployed to limit membership inference risk in machine-learning systems.