PrivUn: Unveiling Latent Ripple Effects and Shallow Forgetting in Privacy Unlearning

• URL: http://arxiv.org/abs/2604.22076v1
• Date: Thu, 23 Apr 2026 21:01:53 GMT
• Title: PrivUn: Unveiling Latent Ripple Effects and Shallow Forgetting in Privacy Unlearning
• Abstract summary: Large language models (LLMs) often memorize private information during training, raising serious privacy concerns.<n>While machine unlearning has emerged as a promising solution, its true effectiveness against privacy attacks remains unclear.
• Score: 21.354585495856345
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Large language models (LLMs) often memorize private information during training, raising serious privacy concerns. While machine unlearning has emerged as a promising solution, its true effectiveness against privacy attacks remains unclear. To address this, we propose PrivUn, a new evaluation framework that systematically assesses unlearning robustness through three-tier attack scenarios: direct retrieval, in-context learning recovery, and fine-tuning restoration; combined with quantitative analysis using forgetting scores, association metrics, and forgetting depth assessment. Our study exposes significant weaknesses in current unlearning methods, revealing two key findings: 1) unlearning exhibits gradient-driven ripple effects: unlike traditional forgetting which follows semantic relations (e.g., knowledge graphs), privacy unlearning propagates across latent gradient-based associations; and 2) most methods suffer from shallow forgetting, failing to remove private information distributed across multiple deep model layers. To validate these insights, we explore two strategies: association-aware core-set selection that leverages gradient similarity, and multi-layer deep intervention through representational constraints. These strategies represent a paradigm shift from shallow forgetting to deep forgetting.

Related papers

• Erase or Hide? Suppressing Spurious Unlearning Neurons for Robust Unlearning [31.829376135133554]
  Large language models trained on web-scale data can memorize private or sensitive knowledge, raising significant privacy risks.<n>Some unlearning methods remain vulnerable to "relearning" during subsequent training, allowing a substantial portion of forgotten knowledge to resurface.<n>We introduce Ssiuu, a new class of unlearning methods that employs attribution-guided regularization to prevent spurious negative influence and faithfully remove target knowledge.
  arXiv  Detail & Related papers  (2025-09-26T12:27:23Z)
• FEDLAD: Federated Evaluation of Deep Leakage Attacks and Defenses [50.921333548391345]
  Federated Learning is a privacy preserving decentralized machine learning paradigm.<n>Recent research has revealed that private ground truth data can be recovered through a gradient technique known as Deep Leakage.<n>This paper introduces the FEDLAD Framework (Federated Evaluation of Deep Leakage Attacks and Defenses), a comprehensive benchmark for evaluating Deep Leakage attacks and defenses.
  arXiv  Detail & Related papers  (2024-11-05T11:42:26Z)
• Towards Effective Evaluations and Comparisons for LLM Unlearning Methods [97.2995389188179]
  This paper seeks to refine the evaluation of machine unlearning for large language models.<n>It addresses two key challenges -- the robustness of evaluation metrics and the trade-offs between competing goals.
  arXiv  Detail & Related papers  (2024-06-13T14:41:00Z)
• Exploring Federated Unlearning: Review, Comparison, and Insights [101.64910079905566]
  federated unlearning enables the selective removal of data from models trained in federated systems.<n>This paper examines existing federated unlearning approaches, examining their algorithmic efficiency, impact on model accuracy, and effectiveness in preserving privacy.<n>We propose the OpenFederatedUnlearning framework, a unified benchmark for evaluating federated unlearning methods.
  arXiv  Detail & Related papers  (2023-10-30T01:34:33Z)
• Protecting Split Learning by Potential Energy Loss [70.81375125791979]
  We focus on the privacy leakage from the forward embeddings of split learning. We propose the potential energy loss to make the forward embeddings become more 'complicated'
  arXiv  Detail & Related papers  (2022-10-18T06:21:11Z)
• On the Privacy Effect of Data Enhancement via the Lens of Memorization [20.63044895680223]
  We propose to investigate privacy from a new perspective called memorization. Through the lens of memorization, we find that previously deployed MIAs produce misleading results as they are less likely to identify samples with higher privacy risks. We demonstrate that the generalization gap and privacy leakage are less correlated than those of the previous results.
  arXiv  Detail & Related papers  (2022-08-17T13:02:17Z)
• Semantics-Preserved Distortion for Personal Privacy Protection in Information Management [65.08939490413037]
  This paper suggests a linguistically-grounded approach to distort texts while maintaining semantic integrity. We present two distinct frameworks for semantic-preserving distortion: a generative approach and a substitutive approach. We also explore privacy protection in a specific medical information management scenario, showing our method effectively limits sensitive data memorization.
  arXiv  Detail & Related papers  (2022-01-04T04:01:05Z)
• Where Did You Learn That From? Surprising Effectiveness of Membership Inference Attacks Against Temporally Correlated Data in Deep Reinforcement Learning [114.9857000195174]
  A major challenge to widespread industrial adoption of deep reinforcement learning is the potential vulnerability to privacy breaches. We propose an adversarial attack framework tailored for testing the vulnerability of deep reinforcement learning algorithms to membership inference attacks.
  arXiv  Detail & Related papers  (2021-09-08T23:44:57Z)
• Gradient Masking and the Underestimated Robustness Threats of Differential Privacy in Deep Learning [0.0]
  This paper experimentally evaluates the impact of training with Differential Privacy (DP) on model vulnerability against a broad range of adversarial attacks. The results suggest that private models are less robust than their non-private counterparts, and that adversarial examples transfer better among DP models than between non-private and private ones.
  arXiv  Detail & Related papers  (2021-05-17T16:10:54Z)
• Privacy-Preserving Federated Learning on Partitioned Attributes [6.661716208346423]
  Federated learning empowers collaborative training without exposing local data or models. We introduce an adversarial learning based procedure which tunes a local model to release privacy-preserving intermediate representations. To alleviate the accuracy decline, we propose a defense method based on the forward-backward splitting algorithm.
  arXiv  Detail & Related papers  (2021-04-29T14:49:14Z)
• SPEED: Secure, PrivatE, and Efficient Deep learning [2.283665431721732]
  We introduce a deep learning framework able to deal with strong privacy constraints. Based on collaborative learning, differential privacy and homomorphic encryption, the proposed approach advances state-of-the-art.
  arXiv  Detail & Related papers  (2020-06-16T19:31:52Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.