mlsanityData to 13 Oct 2026

LUMIA: Linear probing for Unimodal and MultiModal Membership Inference Attacks leveraging internal LLM states

arXiv 2411.19876 ↗First posted 29 Nov 2024cs.CR cs.AIcited by 10

On the map: LLM safety and security → Membership inference attacks

Abstract (arXiv)

Large Language Models (LLMs) are increasingly used in a variety of applications, but concerns around membership inference have grown in parallel. Previous efforts focus on black-to-grey-box models, thus neglecting the potential benefit from internal LLM information. To address this, we propose the use of Linear Probes (LPs) as a method to detect Membership Inference Attacks (MIAs) by examining internal activations of LLMs. Our approach, dubbed LUMIA, applies LPs layer-by-layer to get fine-grained data on the model inner workings. We test this method across several model architectures, sizes and datasets, including unimodal and multimodal tasks. In unimodal MIA, LUMIA achieves an average gain of 15.71 % in Area Under the Curve (AUC) over previous techniques. Remarkably, LUMIA reaches AUC>60% in 65.33% of cases -- an increment of 46.80% against the state of the art. Furthermore, our approach reveals key insights, such as the model layers where MIAs are most detectable. In multimodal models, LPs indicate that visual inputs can significantly contribute to detect MIAs -- AUC>60% is reached in 85.90% of experiments.

Key facts. Cited by 10 papers in the map (data through 13 Oct 2026). Placed in Membership inference attacks, one of 12 directions in LLM safety and security. Builds on 39 of its 56 references that are in the map. First posted on arXiv 29 Nov 2024.

Builds on

39 of 56 references.

Closest ancestors · 6

Same direction, newest first

  1. 2024Detecting Training Data of Large Language Models via Expectation Maximization10 Oct
  2. 2024Membership Inference Attacks Cannot Prove that a Model Was Trained On Your Data29 Sep
  3. 2024Blind Baselines Beat Membership Inference Attacks for Foundation Models23 Jun
  4. 2024Do Membership Inference Attacks Work on Large Language Models?12 Feb
  5. 2023Did the Neurons Read your Book? Document-level Membership Inference for Large Language Models23 Oct
  6. 2021Membership Inference Attacks From First Principles7 Dec

Other references · 33

  1. 2017Attention Is All You Needcited by 60,106
  2. 2015Deep Residual Learning for Image Recognitioncited by 51,950
  3. 2014Adam: A Method for Stochastic Optimizationcited by 37,093
  4. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 30,043
  5. 2023GPT-4 Technical Reportcited by 18,278
  6. 2024The Llama 3 Herd of Modelscited by 14,486
  7. 2024LLaVA-OneVision: Easy Visual Task Transfercited by 2,692
  8. 2020The Pile: An 800GB Dataset of Diverse Text for Language Modelingcited by 2,379
  9. 2020Extracting Training Data from Large Language Modelscited by 2,181
  10. 2022Learn to Explain: Multimodal Reasoning via Thought Chains for Science Question Answeringcited by 1,896
  11. 2023Pythia: A Suite for Analyzing Large Language Models Across Training and Scalingcited by 1,761
  12. 2022ChartQA: A Benchmark for Question Answering about Charts with Visual and Logical Reasoningcited by 1,461
21 more
  1. 2016Understanding intermediate layers using linear classifier probescited by 1,313
  2. 2024How Far Are We to GPT-4V? Closing the Gap to Commercial Multimodal Models with Open-Source Suitescited by 1,041
  3. 2019Compressive Transformers for Long-Range Sequence Modellingcited by 707
  4. 2022A-OKVQA: A Benchmark for Visual Question Answering using World Knowledgecited by 706
  5. 2024PaliGemma: A versatile 3B VLM for transfercited by 686
  6. 2024Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Trainingcited by 508
  7. 2020TextCaps: a Dataset for Image Captioning with Reading Comprehensioncited by 398
  8. 2023Detecting Pretraining Data from Large Language Modelscited by 383
  9. 2023What Makes Good Data for Alignment? A Comprehensive Study of Automatic Data Selection in Instruction Tuningcited by 357
  10. 2024Magpie: Alignment Data Synthesis from Scratch by Prompting Aligned LLMs with Nothingcited by 342
  11. 2019White-box vs Black-box: Bayes Optimal Strategies for Membership Inferencecited by 270
  12. 2021IconQA: A New Benchmark for Abstract Diagram Understanding and Visual Language Reasoningcited by 264
  13. 2024VITA: Towards Open-Source Interactive Omni Multimodal LLMcited by 172
  14. 2024Math-LLaVA: Bootstrapping Mathematical Reasoning for Multimodal Large Language Modelscited by 133
  15. 2024Black-Box Access is Insufficient for Rigorous AI Auditscited by 125
  16. 2024On Protecting the Data Privacy of Large Language Models (LLMs): A Surveycited by 94
  17. 2024Get my drift? Catching LLM Task Drift with Activation Deltascited by 71
  18. 2024Do LLMs Know about Hallucination? An Empirical Investigation of LLM's Hidden Statescited by 52
  19. 2024Probing Language Models for Pre-training Data Detectioncited by 24
  20. 2024Cross-Modal Projection in Multimodal LLMs Doesn't Really Project Visual Attributes to Textual Spacecited by 21
  21. 2023Energy and Carbon Considerations of Fine-Tuning BERTcited by 11

Built on by

Cited by year (2026 to 13 Oct)

Citing papers per year: 2025: 5; 2026: 5.

10 papers.

  1. 2025Membership Inference Attacks on Large-Scale Models: A Surveycited by 17
  2. 2026G-Drift MIA: Membership Inference via Gradient-Induced Feature Drift in LLMscited by 7
  3. 2025Black-Box Membership Inference Attack for LVLMs via Prior Knowledge-Calibrated Memory Probingcited by 4
  4. 2025SMA: Who Said That? Auditing Membership Leakage in Semi-Black-box RAG Controllingcited by 3
  5. 2025Data Provenance Auditing of Fine-Tuned Large Language Models with a Text-Preserving Techniquecited by 3
  6. 2025MrM: Black-Box Membership Inference Attacks against Multimodal RAG Systemscited by 2
  7. 2026What Hard Tokens Reveal: Exploiting Low-confidence Tokens for Membership Inference Attacks against Large Language Modelscited by 0
  8. 2026SERSEM: Selective Entropy-Weighted Scoring for Membership Inference in Code Language Modelscited by 0
  9. 2026DistractMIA: Black-Box Membership Inference on Vision-Language Models via Semantic Distractioncited by 0
  10. 2026MemCatalyst: Amplifying Data Auditing on Vision-Language Models via Data Poisoningcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026Membership Inference on LLMs in the Wild
  2. 2025Exploring the limits of strong membership inference attacks on large language models
  3. 2025Automatic Calibration for Membership Inference Attack on Large Language Models
  4. 2024A Statistical and Multi-Perspective Revisiting of the Membership Inference Attack in Large Language Models
  5. 2026AttenMIA: LLM Membership Inference Attack through Attention Signals
  6. 2024Scaling Up Membership Inference: When and How Attacks Succeed on Large Language Models
  7. 2025In-Context Probing for Membership Inference in Fine-Tuned Language Models
  8. 2025FiMMIA: scaling semantic perturbation-based membership inference across modalities