Lune

ACL Findings2025

Sampling-based Pseudo-Likelihood for Membership Inference Attacks

Masahiro Kaneko, Youmi Ma, Yuki Wata, Naoaki Okazaki

Abstract

Large Language Models (LLMs) are trained on large-scale web data, which makes it difficult to grasp the contribution of each text. This poses the risk of leaking inappropriate data such as benchmarks, personal information, and copyrighted texts in the training data. Membership Inference Attacks (MIA), which determine whether a given text is included in the model's training data, have been attracting attention. Previous studies of MIAs revealed that likelihood-based classification is effective for detecting leaks in LLMs. However, the existing likelihood-based methods cannot be applied to some proprietary models like Chat-GPT or Claude 3 because the likelihood for input text is unavailable to the user. In this study, we propose a Sampling-based Pseudo-Likelihood (SPL) method for MIA (SaMIA) that calculates SPL using only the text generated by an LLM to detect leaks. The SaMIA treats the target text as the reference text and multiple outputs from the LLM as text samples, calculates the degree of n-gram match as SPL, and determines the membership of the text in the training data. Even without likelihoods, SaMIA performed on par with existing likelihood-based methods 1 .

Ask Lune

Know this paper in minutes, not an afternoon.

Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.

6 top-tier papers have cited it since. Know what they found before your reviewers do.

Questions to start from

Lune can draw on

the full paper · the top-tier papers that cite it · related work

Get the answer

Free to start. No credit card required.

Lineage

  • A Statistical and Multi-Perspective Revisiting of the Membership Inference Attack in Large Language Models cites Detecting Pretraining Data from Large Language Models
  • Can We Infer Confidential Properties of Training Data from LLMs? cites Membership Inference Attacks Against Machine Learning Models
  • Can We Infer Confidential Properties of Training Data from LLMs? cites Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling
  • Can We Infer Confidential Properties of Training Data from LLMs? cites BERTScore: Evaluating Text Generation with BERT
  • Investigating How Pre-training Data Leakage Affects Models' Reproduction and Detection Capabilities cites Membership Inference Attacks Against Machine Learning Models
  • Investigating How Pre-training Data Leakage Affects Models' Reproduction and Detection Capabilities cites Detecting Pretraining Data from Large Language Models
  • Investigating How Pre-training Data Leakage Affects Models' Reproduction and Detection Capabilities cites Deduplicating Training Data Mitigates Privacy Risks in Language Models
  • Detecting Pretraining Data from Large Language Models cites Membership Inference Attacks Against Machine Learning Models
  • Detecting Pretraining Data from Large Language Models cites Deduplicating Training Data Mitigates Privacy Risks in Language Models
  • Deduplicating Training Data Mitigates Privacy Risks in Language Models cites Membership Inference Attacks Against Machine Learning Models
Builds onCited byLeft to right by year; the more cited, the higher.

Cited by top-tier papers6

  1. Can We Infer Confidential Properties of Training Data from LLMs?Pengrun Huang, Chhavi Yadav, Kamalika Chaudhuri, Ruihan WuNeurIPS 2025 · 7 citations
  2. Bits Leaked per Query: Information-Theoretic Bounds for Adversarial Attacks on LLMsMasahiro Kaneko, Timothy BaldwinNeurIPS 2025 · 6 citations
  3. Investigating How Pre-training Data Leakage Affects Models' Reproduction and Detection CapabilitiesMasahiro Kaneko, Timothy BaldwinEMNLP 2025 · 2 citations
  4. Reverse-Engineering Model Editing on Language ModelsZhiyu Sun, Minrui Luo, Yu Wang, Zhili Chen et al.ICML 2026 · 1 citation
  5. A Statistical and Multi-Perspective Revisiting of the Membership Inference Attack in Large Language ModelsBowen Chen, Namgi Han, Yusuke MiyaoACL 2025

Builds on15

  1. BERTScore: Evaluating Text Generation with BERTTianyi Zhang, Varsha Kishore, Felix Wu, Kilian Q. Weinberger et al.ICLR 2020 · 8,443 citations
  2. Membership Inference Attacks Against Machine Learning ModelsReza Shokri, Marco Stronati, Congzheng Song, Vitaly ShmatikovS&P 2017 · 5,137 citations
  3. Pythia: A Suite for Analyzing Large Language Models Across Training and ScalingStella Biderman, Hailey Schoelkopf, Quentin Gregory Anthony, Herbie Bradley et al.ICML 2023 · 1,822 citations
  4. Deduplicating Training Data Mitigates Privacy Risks in Language ModelsNikhil Kandpal, Eric Wallace, Colin RaffelICML 2022 · 395 citations
  5. Detecting Pretraining Data from Large Language ModelsWeijia Shi, Anirudh Ajith, Mengzhou Xia, Yangsibo Huang et al.ICLR 2024 · 365 citations

Related papers

  1. Robust Membership Inference for Large Language Models under Adversarial Generative CorruptionYuanhong Huang, Huili Wang, Xueying Bai, Jinrui Wang et al.ACL 2026
  2. Tag&Tab: Pretraining Data Detection in Large Language Models Using Keyword-Based Membership Inference AttackSagiv Antebi, Edan Habler, Asaf Shabtai, Yuval EloviciEMNLP Findings 2025
  3. Decoding Web Memorization: A Semantic Membership Inference Attack on LLMsZhiyao Wu, Zi Liang, Haibo HuWWW 2026
  4. ReCaLL: Membership Inference via Relative Conditional Log-LikelihoodsRoy Xie, Junlin Wang, Ruomin Huang, Minxing Zhang et al.EMNLP 2024 · 8 citations
  5. DPDLLM: A Black-box Framework for Detecting Pre-training Data from Large Language ModelsBaohang Zhou, Zezhong Wang, Lingzhi Wang, Hongru Wang et al.ACL Findings 2024 · 8 citations
Dusk over the sea between two cliffs drawn in fine vertical lines