Generalized Leverage Score for Scalable Assessment of Privacy Vulnerability

2026-02-17

AI Summary

This work proposes an efficient method for quantifying the privacy risk of individual data points under membership inference attacks without requiring model retraining or shadow model construction. Through theoretical analysis, it establishes—for the first time—a formal connection between an individual’s influence on a model and their vulnerability to privacy leakage. In linear models, this risk is shown to correspond precisely to the leverage score, a well-known statistical measure. The approach is then extended to deep learning settings by introducing a computationally efficient generalized leverage score as a proxy for privacy risk. Experimental results demonstrate a strong correlation between this metric and the success rate of membership inference attacks, offering a practical and scalable tool for individual-level privacy assessment.

Technology Category

• Machine Learning: Privacy
• Reasoning under Uncertainty: Other Foundations of Reasoning under Uncertainty
• Computer Vision: Adversarial Attacks & Robustness

Application Category

• Security and Privacy: Large-scale security measurements
• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems
• Responsible Web: Data and user privacy-enhancing technologies for the Web

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.

Problem

• privacy vulnerability
• membership inference attack
• leverage score
• individual privacy risk
• data point influence

Innovation

• leverage score
• membership inference attack
• privacy vulnerability
• scalable assessment
• deep learning