mlsanityData to 14 Oct 2026

Machine Unlearning as Private Retroactive Algorithms

arXiv 2609.05329 ↗First posted 4 Sep 2026cs.CR cs.DScited by 0

On the map: LLM safety and security → Machine unlearning

Abstract (arXiv)

Machine unlearning typically aims to emulate retraining from scratch: upon a deletion request, the unlearning algorithm should produce an outcome that would have been obtained had the deleted point never been included. Recent work has shown that this emulation requirement carries no meaningful privacy semantics against an adversary who observes a sequence of releases. Machine unlearning is thus not a privacy question per se, but rather a data maintenance question, which is precisely the subject of retroactive algorithms. These are algorithms supporting modifications of past operations, guaranteeing that all subsequent answers reflect the revised history as if it had always been in force. We put forward a definition of private retroactive algorithms, combining the retroactivity requirement with differential privacy under continual observation. We present constructions achieving both privacy and retroactivity at no asymptotic cost over privacy alone for linear statistics, clustering, and histograms, alongside impossibility results.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Machine unlearning, one of 12 directions in LLM safety and security. Builds on 20 of its 34 references that are in the map. First posted on arXiv 4 Sep 2026.

Builds on

20 of 34 references.

Closest ancestors · 4

Same direction, newest first

  1. 2026Revisiting Privacy Leakage in Machine Unlearning: Membership Inference Beyond the Forgotten Set1 May
  2. 2026Protecting the Undeleted in Machine Unlearning18 Feb
  3. 2021Adaptive Machine Unlearning8 Jun
  4. 2021Remember What You Want to Forget: Algorithms for Machine Unlearning4 Mar

Other references · 16

  1. 2019Machine Unlearningcited by 1,061
  2. 2019Certified Data Removal from Machine Learning Modelscited by 535
  3. 2019Making AI Forget You: Data Deletion in Machine Learningcited by 465
  4. 2020Descent-to-Delete: Gradient-Based Methods for Machine Unlearningcited by 311
  5. 2020Approximate Data Deletion from Machine Learning Modelscited by 255
  6. 2018Privacy Amplification by Subsampling: Tight Analyses via Couplings and Divergencescited by 242
  7. 2020When Machine Unlearning Jeopardizes Privacycited by 198
  8. 2022The Privacy Onion Effect: Memorization is Relativecited by 136
  9. 2015Differentially Private Release and Learning of Threshold Functionscited by 60
  10. 2021The Price of Differential Privacy under Continual Observationcited by 34
  11. 2022Almost Tight Error Bounds on Differentially Private Continual Countingcited by 26
  12. 2022~Optimal Differentially Private Learning of Thresholds and Quasi-Concave Optimizationcited by 18
4 more
  1. 2024Lower Bounds for Differential Privacy Under Continual Observation and Online Threshold Queriescited by 12
  2. 2023Counting Distinct Elements in the Turnstile Model with Differential Privacy under Continual Observationcited by 5
  3. 2023Differential Privacy for Clustering Under Continual Observationcited by 2
  4. 2026The Binary Tree Mechanism is Optimal for Differentially Private Continual Countingcited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025Reminiscence Attack on Residuals: Exploiting Approximate Machine Unlearning for Privacy
  2. 2022Forget Unlearning: Towards True Data-Deletion in Machine Learning
  3. 2023Learn to Unlearn: A Survey on Machine Unlearning
  4. 2024The Utility and Complexity of in- and out-of-Distribution Machine Unlearning
  5. 2025Leveraging Per-Instance Privacy for Machine Unlearning
  6. 2026On the Recoverability of Private Information Unlearning in Large Language Models
  7. 2026Auditing Machine Unlearning: A Systematic Research on Whether Models Truly Forget
  8. 2025Rectifying Privacy and Efficacy Measurements in Machine Unlearning: A New Inference Attack Perspective