# Low-Cost High-Power Membership Inference by Boosting Relativity

**Authors:** Sajjad Zarifzadeh, Philippe Liu, Reza Shokri  
**arXiv:** [2312.03262](https://arxiv.org/abs/2312.03262)  
**Date:** 2023-12-06  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/2312.03262v1/

## Key points

- Authors Zarifzadeh, Liu, and Shokri introduce a robust membership inference attack (RMIA) that enhances differentiation between population data and training data on target models.
- The algorithm uses reference models and data in a likelihood ratio test for superior true-positive rates compared to previous methods.
- RMIA method shows remarkably low false-positive error rates as low as 0.
- The approach is effective under computation constraints, requiring only a limited number of reference models (as few as 1).
- Their technique outperforms prior attacks that resort to random guessing in similar scenarios.
- The work lays a foundation for cost-effective yet robust privacy risk assessment in machine learning.
- Emphasis on leveraging both reference models and data sets showcases a novel approach to enhancing security and privacy considerations within machine learning systems.

## Summary

In their paper titled "Low-Cost High-Power Membership Inference by Boosting Relativity," authors Sajjad Zarifzadeh, Philippe Liu, and Reza Shokri introduce a robust membership inference attack (RMIA) that enhances the differentiation between population data and training data on any target model. The algorithm utilizes both reference models and reference data in a likelihood ratio test to demonstrate superior test power in terms of true-positive rate compared to previous methods. This results in remarkably low false-positive error rates as low as 0. One key strength of their approach is its effectiveness under computation constraints, where only a limited number of reference models (as few as 1) are available. The RMIA method outperforms prior attacks that tend to resort to random guessing in such scenarios. This makes their technique not only powerful but also practical for privacy risk analysis of machine learning algorithms. Overall, the work by Zarifzadeh, Liu, and Shokri lays a solid foundation for cost-effective yet robust privacy risk assessment in the realm of machine learning. Their emphasis on leveraging both reference models and data sets showcases a novel approach to enhancing the security and privacy considerations within machine learning systems.