ML Security PapersImitative Membership Inference Attack
Yuntao Du 1, Yuetian Chen 1, Hanshen Xiao 1,2, Bruno Ribeiro 1, Ninghui Li 1
Published on arXiv
2509.06796
Membership Inference Attack
OWASP ML Top 10 — ML04
Key Finding
IMIA substantially outperforms state-of-the-art membership inference attacks while requiring less than 5% of their computational cost by using target-informed imitative models instead of independent shadow models.
IMIA (Imitative Membership Inference Attack)
Novel technique introduced
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.
Key Contributions
- Introduces IMIA, a membership inference attack that uses imitative training to construct a small number of target-informed shadow models that closely replicate the target model's behavior
- Reduces computational cost to less than 5% of state-of-the-art MIAs that require training hundreds of independent shadow models
- Demonstrates that IMIA substantially outperforms existing MIAs across various attack settings
🛡️ Threat Analysis
The paper's sole contribution is a new membership inference attack — IMIA — that determines whether specific query instances were part of a target model's training set, directly matching ML04's core definition.