Boundary-targeted Membership Inference Attacks on Safety Classifiers

• URL: http://arxiv.org/abs/2605.22373v2
• Date: Fri, 22 May 2026 14:13:36 GMT
• Title: Boundary-targeted Membership Inference Attacks on Safety Classifiers
• Abstract summary: Safety classifiers are trained on sensitive datasets including discussions of self-harm and mental health.<n>We introduce a new boundary-targeted selection strategy that identifies low confidence examples.<n>Experiments show that an adversary can recover 19% of the conversations a safety classifier flagged as indicating user distress, at a 5% false-positive rate.<n>This is $3.5$ times more than attacking using state-of-the-art MIA methods alone.
• Score: 70.20833439671131
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Safety classifiers are essential safeguards within generative AI systems, filtering harmful content or identifying at-risk users when interacting with large language models. Despite their necessity, these models are trained on sensitive datasets including discussions of self-harm and mental health, raising important, yet poorly understood, privacy concerns. Membership inference attacks (MIAs) allow adversaries to infer membership of examples used to train models. In this work, we hypothesize that identifying the examples on which the classifier is least confident are informative for an adversary to infer membership. This reflects a localized failure of generalization, where the model relies on memorization to resolve ambiguity in the training set. To investigate this, we introduce a new boundary-targeted selection strategy that identifies low confidence examples that amplify the signal of an examples membership within a training set. Our experimental results show that an adversary can recover 19% of the conversations a safety classifier flagged as indicating user distress, at a 5% false-positive rate, on a classifier fine-tuned for detecting a user who may require emotional support. This is $3.5$ times more than attacking using state-of-the-art MIA methods alone. Finally, we characterize the boundary laying examples and show that content-based filtering is ineffective for protection, and existing noise strategies can effectively mitigate susceptibility of these examples.

Related papers

• Adversarial Attacks for Good: A Survey of Proactive Protection across the Visual Content Lifecycle [65.6642776933861]
  This survey examines the protective paradigm that has grown around this intervention point.<n>We find that most protections are still validated mainly against static or weakly adaptive adversaries.<n>We close by consolidating cross-stage countermeasures and open problems for robust, composable, and deployable owner-side protection.
  arXiv  Detail & Related papers  (2026-08-05T00:46:50Z)
• Privacy Vulnerabilities of Attention Layers in Tabular Foundation Models and Protection of High-Risk Queries [12.545534835317481]
  In this paper, we demonstrate that predictions generated via the attention mechanism leak sufficient information to enable effective Membership Inference Attacks (MIAs)<n>We introduce an inference-time defence inspired by $k$-anonymity principles to mitigate this risk.
  arXiv  Detail & Related papers  (2026-06-24T16:42:21Z)
• Neighborhood Blending: A Lightweight Inference-Time Defense Against Membership Inference Attacks [5.468130838517792]
  We introduce a novel inference-time defense mechanism called Neighborhood Blending.<n>Our method establishes a consistent confidence pattern, rendering members and non-members indistinguishable to an adversary.<n>It is a model-agnostic approach that offers a practical, lightweight solution that enhances privacy without sacrificing model utility.
  arXiv  Detail & Related papers  (2026-02-13T14:01:21Z)
• Membership Inference Attacks on Tokenizers of Large Language Models [40.2492347972186]
  We present the first study on membership leakage through tokenizers.<n>We explore five attack methods to infer dataset membership.<n>Our findings highlight tokenizers as an overlooked yet critical privacy threat.
  arXiv  Detail & Related papers  (2025-10-07T09:05:40Z)
• VoxGuard: Evaluating User and Attribute Privacy in Speech via Membership Inference Attacks [51.68795949691009]
  We introduce VoxGuard, a framework grounded in differential privacy and membership inference.<n>For attributes, we show that simple transparent attacks recover gender and accent with near-perfect accuracy even after anonymization.<n>Our results demonstrate that EER substantially underestimates leakage, highlighting the need for low-FPR evaluation.
  arXiv  Detail & Related papers  (2025-09-22T20:57:48Z)
• Robust Safety Classifier for Large Language Models: Adversarial Prompt Shield [7.5520641322945785]
  Large Language Models' safety remains a critical concern due to their vulnerability to adversarial attacks. We introduce the Adversarial Prompt Shield (APS), a lightweight model that excels in detection accuracy and demonstrates resilience against adversarial prompts. We also propose novel strategies for autonomously generating adversarial training datasets.
  arXiv  Detail & Related papers  (2023-10-31T22:22:10Z)
• Purifier: Defending Data Inference Attacks via Transforming Confidence Scores [27.330482508047428]
  We propose a method, namely PURIFIER, to defend against membership inference attacks. Experiments show that PURIFIER helps defend membership inference attacks with high effectiveness and efficiency. PURIFIER is also effective in defending adversarial model inversion attacks and attribute inference attacks.
  arXiv  Detail & Related papers  (2022-12-01T16:09:50Z)
• Effective Targeted Attacks for Adversarial Self-Supervised Learning [58.14233572578723]
  unsupervised adversarial training (AT) has been highlighted as a means of achieving robustness in models without any label information. We propose a novel positive mining for targeted adversarial attack to generate effective adversaries for adversarial SSL frameworks. Our method demonstrates significant enhancements in robustness when applied to non-contrastive SSL frameworks, and less but consistent robustness improvements with contrastive SSL frameworks.
  arXiv  Detail & Related papers  (2022-10-19T11:43:39Z)
• Towards A Conceptually Simple Defensive Approach for Few-shot classifiers Against Adversarial Support Samples [107.38834819682315]
  We study a conceptually simple approach to defend few-shot classifiers against adversarial attacks. We propose a simple attack-agnostic detection method, using the concept of self-similarity and filtering. Our evaluation on the miniImagenet (MI) and CUB datasets exhibit good attack detection performance.
  arXiv  Detail & Related papers  (2021-10-24T05:46:03Z)
• Detection of Adversarial Supports in Few-shot Classifiers Using Feature Preserving Autoencoders and Self-Similarity [89.26308254637702]
  We propose a detection strategy to highlight adversarial support sets. We make use of feature preserving autoencoder filtering and also the concept of self-similarity of a support set to perform this detection. Our method is attack-agnostic and also the first to explore detection for few-shot classifiers to the best of our knowledge.
  arXiv  Detail & Related papers  (2020-12-09T14:13:41Z)
• Sampling Attacks: Amplification of Membership Inference Attacks by Repeated Queries [74.59376038272661]
  We introduce sampling attack, a novel membership inference technique that unlike other standard membership adversaries is able to work under severe restriction of no access to scores of the victim model. We show that a victim model that only publishes the labels is still susceptible to sampling attacks and the adversary can recover up to 100% of its performance. For defense, we choose differential privacy in the form of gradient perturbation during the training of the victim model as well as output perturbation at prediction time.
  arXiv  Detail & Related papers  (2020-09-01T12:54:54Z)
• Systematic Evaluation of Privacy Risks of Machine Learning Models [41.017707772150835]
  We show that prior work on membership inference attacks may severely underestimate the privacy risks. We first propose to benchmark membership inference privacy risks by improving existing non-neural network based inference attacks. We then introduce a new approach for fine-grained privacy analysis by formulating and deriving a new metric called the privacy risk score.
  arXiv  Detail & Related papers  (2020-03-24T00:53:53Z)

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.