Imitative Membership Inference Attack

Yuntao Du, Yuetian Chen, Hanshen Xiao, Bruno Ribeiro, Ninghui Li
9/8/2025
cs.CRcs.LG

No official code found. Best implementation: zealscott/IMIA (6 stars, last commit 2026-01-09).

Abstract

A Membership Inference Attack (MIA) assesses how much a target machine learning model reveals about its training data by determining whether specific query instances were part of the training set. State-of-the-art MIAs rely on training hundreds of shadow models that are independent of the target model, leading to significant computational overhead. In this paper, we introduce Imitative Membership Inference Attack (IMIA), which employs a novel imitative training technique to strategically construct a small number of target-informed imitative models that closely replicate the target model's behavior for inference. Extensive experimental results demonstrate that IMIA substantially outperforms existing MIAs in various attack settings while only requiring less than 5% of the computational cost of state-of-the-art approaches.

View on arXivView PDF

Code Implementations(1)

zealscott/IMIACommunity65%

Code for Imitative Membership Inference Attack. USENIX Security 2026.

61PythonAug 5, 20259 months ago
distillationmachine-learningmembership-inference-attackprivacysecurity+1 more

Similar Papers

CLMIA: Membership Inference Attacks via Unsupervised Contrastive Learning

88% match2024

MIA-BAD: An Approach for Enhancing Membership Inference Attack and its Mitigation with Federated Learning

87% match1 repo2023

Membership Inference Attacks for Unseen Classes

87% match1 repo2025

Detecting Training Data of Large Language Models via Expectation Maximization

87% match1 repo72024

Membership Inference Attacks Beyond Overfitting

86% match1 repo12025

Confidence Is All You Need for MI Attacks

86% match1 repo92023

Cite this paper

@article{du2025imitative,
  title  = {Imitative Membership Inference Attack},
  author = {Yuntao Du and Yuetian Chen and Hanshen Xiao and Bruno Ribeiro and Ninghui Li},
  year   = {2025},
  eprint = {2509.06796},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2509.06796v3}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2509.06796.svg)](https://codeofpaper.com/paper/2509.06796)

Discussion