What type of study is this?

This is a Quantitative Study study.

October 12, 2025Open Access

Imitative Membership Inference Attack

Key Points

Imitative Membership Inference Attack achieves significant improvements in inference accuracy with reduced computational costs.
Results show that IMIA requires less than 5% of the computational resources compared to traditional membership inference attacks.
The approach relies on a small number of imitative models built from the target model's behavior, rather than hundreds of shadow models.
Experimental findings indicate that IMIA outperforms existing methods across various attack scenarios, demonstrating its efficiency.

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.

Imitative Membership Inference Attack

Key Points

Abstract

Cite This Study