Generalized Leverage Score for Scalable Assessment of Privacy Vulnerability

TL;DR

This paper introduces a scalable method to assess individual data point privacy vulnerability by generalizing leverage scores, linking them to membership inference attack risks without retraining models.

Contribution

It formalizes the connection between leverage scores and privacy risk, and extends this concept to deep learning for efficient vulnerability assessment.

Findings

Strong correlation between the proposed score and MIA success

Efficient computation of privacy vulnerability in deep models

Theoretical foundation linking influence and privacy risk

Abstract

Can the privacy vulnerability of individual data points be assessed without retraining models or explicitly simulating attacks? We answer affirmatively by showing that exposure to membership inference attack (MIA) is fundamentally governed by a data point's influence on the learned model. We formalize this in the linear setting by establishing a theoretical correspondence between individual MIA risk and the leverage score, identifying it as a principled metric for vulnerability. This characterization explains how data-dependent sensitivity translates into exposure, without the computational burden of training shadow models. Building on this, we propose a computationally efficient generalization of the leverage score for deep learning. Empirical evaluations confirm a strong correlation between the proposed score and MIA success, validating this metric as a practical surrogate for individual privacy risk assessment.