mlsanityData to 7 Oct 2026

User-Level Membership Inference Attack against Metric Embedding Learning

arXiv 2203.02077 ↗First posted 4 Mar 2022cs.LG cs.AI cs.CRcited by 22

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

Abstract (arXiv)

Membership inference (MI) determines if a sample was part of a victim model training set. Recent development of MI attacks focus on record-level membership inference which limits their application in many real-world scenarios. For example, in the person re-identification task, the attacker (or investigator) is interested in determining if a user's images have been used during training or not. However, the exact training images might not be accessible to the attacker. In this paper, we develop a user-level MI attack where the goal is to find if any sample from the target user has been used during training even when no exact training sample is available to the attacker. We focus on metric embedding learning due to its dominance in person re-identification, where user-level MI attack is more sensible. We conduct an extensive evaluation on several datasets and show that our approach achieves high accuracy on user-level MI task.

Key facts. Cited by 22 papers in the map (data through 7 Oct 2026). Placed in Membership inference attacks, one of 12 directions in LLM safety and security. Builds on 5 of its 9 references that are in the map. First posted on arXiv 4 Mar 2022.

Builds on

5 of 9 references.

Closest ancestors · 1

Same direction, newest first

  1. 2021EncoderMI: Membership Inference against Pre-trained Encoders in Contrastive Learning25 Aug

Other references · 4

  1. 2016Membership Inference Attacks against Machine Learning Modelscited by 2,221
  2. 2017In Defense of the Triplet Loss for Person Re-Identificationcited by 784
  3. 2018ML-Leaks: Model and Data Independent Membership Inference Attacks and Defenses on Machine Learning Modelscited by 475
  4. 2019The Audio Auditor: User-Level Membership Inference in Internet of Things Voice Servicescited by 25

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2022: 4; 2023: 7; 2024: 8; 2025: 2; 2026: 1.

22 papers.

Sort:

By area:

  1. 2023User Inference Attacks on Large Language Modelscited by 36
  2. 2024Fine-Tuning Large Language Models with User-Level Differential Privacycited by 28
  3. 2022Does CLIP Know My Face?cited by 19
  4. 2023FACE-AUDITOR: Data Auditing in Facial Recognition Systemscited by 15
  5. 2023SLMIA-SR: Speaker-Level Membership Inference Attacks against Speaker Recognition Systemscited by 14
  6. 2024SoK: Dataset Copyright Auditing in Machine Learning Systemscited by 14
  7. 2024Is My Data in Your AI? Membership Inference Test (MINT) applied to Face Biometricscited by 12
  8. 2022Data Isotopes for Data Provenance in DNNscited by 11
  9. 2022Similarity Distribution based Membership Inference Attack on Person Re-identificationcited by 8
  10. 2025Cascading and Proxy Membership Inference Attackscited by 6
  11. 2023Balancing Transparency and Risk: The Security and Privacy Risks of Open-Source Machine Learning Modelscited by 4
  12. 2023Defending Our Privacy With Backdoorscited by 4
10 more
  1. 2024Analyzing Inference Privacy Risks Through Gradients in Machine Learningcited by 4
  2. 2022On the Discredibility of Membership Inference Attackscited by 3
  3. 2023SoK: Comparing Different Membership Inference Attacks with a Comprehensive Benchmarkcited by 3
  4. 2023Privacy Threats in Stable Diffusion Modelscited by 3
  5. 2024Inference Attacks: A Taxonomy, Survey, and Promising Directionscited by 2
  6. 2024Fingerprint Membership and Identity Inference Against Generative Adversarial Networkscited by 2
  7. 2024Inference Attacks Against Face Recognition Model without Classification Layerscited by 1
  8. 2024CLMIA: Membership Inference Attacks via Unsupervised Contrastive Learningcited by 1
  9. 2025VICTOR: Dataset Copyright Auditing in Video Recognition Systemscited by 0
  10. 2026Quantifying Training Membership Information in the Hyperspherical Embedding Geometry of Face Recognition Modelscited by 0