mlsanityData to 14 Oct 2026

Bayesian Perspective on Memorization and Reconstruction

arXiv 2505.23658 ↗First posted 29 May 2025cs.CR cs.LGcited by 0

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

Abstract (arXiv)

We introduce a new Bayesian perspective on the concept of data reconstruction, and leverage this viewpoint to propose a new security definition that, in certain settings, provably prevents reconstruction attacks. We use our paradigm to shed new light on one of the most notorious attacks in the privacy and memorization literature - fingerprinting code attacks (FPC). We argue that these attacks are really a form of membership inference attacks, rather than reconstruction attacks. Furthermore, we show that if the goal is solely to prevent reconstruction (but not membership inference), then in some cases the impossibility results derived from FPC no longer apply.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Membership inference attacks, one of 12 directions in LLM safety and security. Builds on 23 of its 38 references that are in the map. First posted on arXiv 29 May 2025.

Builds on

23 of 38 references.

References · 23

  1. 2016Membership Inference Attacks against Machine Learning Modelscited by 2,281
  2. 2020Extracting Training Data from Large Language Modelscited by 2,189
  3. 2018The Secret Sharer: Evaluating and Testing Unintended Memorization in Neural Networkscited by 864
  4. 2022Quantifying Memorization Across Neural Language Modelscited by 763
  5. 2023Extracting Training Data from Diffusion Modelscited by 732
  6. 2020What Neural Networks Memorize and Why: Discovering the Long Tail via Influence Estimationcited by 497
  7. 2019Does Learning Require Memorization? A Short Tale about a Long Tailcited by 460
  8. 2022Reconstructing Training Data from Trained Neural Networkscited by 133
  9. 2020When is Memorization of Irrelevant Training Data Necessary for High-Accuracy Learning?cited by 104
  10. 2013Fingerprinting Codes and the Price of Approximate Differential Privacycited by 82
  11. 2015Between Pure and Approximate Differential Privacycited by 56
  12. 2014Preventing False Discovery in Interactive Data Analysis is Hardcited by 34
11 more
  1. 2015Simultaneous Private Learning of Multiple Conceptscited by 34
  2. 2017Tight Lower Bounds for Differentially Private Selectioncited by 34
  3. 2021The Price of Differential Privacy under Continual Observationcited by 34
  4. 2022New Lower Bounds for Private Estimation and a Generalized Fingerprinting Lemmacited by 33
  5. 2016Make Up Your Mind: The Price of Online Queries in Differential Privacycited by 26
  6. 2023Deconstructing Data Reconstruction: Multiclass, Weight Decay and General Lossescited by 26
  7. 2018The Limits of Post-Selection Generalizationcited by 16
  8. 2021Separating Adaptive Streaming from Oblivious Streamingcited by 9
  9. 2023On Differential Privacy and Adaptive Data Analysis with Bounded Spacecited by 9
  10. 2023Smooth Lower Bounds for Differentially Private Algorithms via Padding-and-Permuting Fingerprinting Codescited by 8
  11. 2024Data Reconstruction: When You See It and When You Don'tcited by 7

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026SoK: Reconstruction Attacks on Synthetic Tabular Data (Insights from Winning the NIST CRC)
  2. 2024Reconstruction Attacks on Machine Unlearning: Simple Models are Vulnerable
  3. 2022Defending against Reconstruction Attacks with R'enyi Differential Privacy
  4. 2026Beyond Indistinguishability: Measuring Extraction Risk in LLM APIs
  5. 2026Understanding Disclosure Risk in Differential Privacy with Applications to Noise Calibration and Auditing (Extended Version)
  6. 2025Ensembling Membership Inference Attacks Against Tabular Generative Models
  7. 2022On the Privacy Effect of Data Enhancement via the Lens of Memorization
  8. 2025Synth-MIA: A Testbed for Auditing Privacy Leakage in Tabular Data Synthesis