KV-Auditor: Auditing Local Differential Privacy for Correlated Key-Value Estimation

2025-08-15

AI Summary

Existing Local Differential Privacy (LDP) auditing methods primarily target centralized DP or simple discrete frequency estimation, lacking systematic support for key-value (KV) data—where both key (discrete) frequencies and value (continuous) means must be simultaneously protected. Method: We propose KV-Auditor, the first LDP auditing framework tailored for KV data, supporting both interactive and non-interactive mechanisms. It introduces an empirical lower bound estimation method based on unbounded output distributions to precisely quantify privacy leakage in continuous-value settings, and employs a dual horizontal/vertical architecture adaptable to varying domain sizes, integrating sample distribution analysis and segmentation strategy modeling for iterative privacy loss accumulation. Results: Experiments demonstrate KV-Auditor’s accuracy in estimating privacy lower bounds across diverse LDP mechanisms, establishing the first practical auditing tool and theoretical foundation for designing, evaluating, and optimizing KV-type LDP protocols.

Technology Category

• Machine Learning: Privacy
• Data Mining & Knowledge Management: Representing, Reasoning, and Using Provenance, Trust
• Natural Language Processing: Ethics — Bias, Fairness, Transparency & Privacy

Application Category

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

Abstract

To protect privacy for data-collection-based services, local differential privacy (LDP) is widely adopted due to its rigorous theoretical bound on privacy loss. However, mistakes in complex theoretical analysis or subtle implementation errors may undermine its practical guarantee. To address this, auditing is crucial to confirm that LDP protocols truly protect user data. However, existing auditing methods, though, mainly target machine learning and federated learning tasks based on centralized differentially privacy (DP), with limited attention to LDP. Moreover, the few studies on LDP auditing focus solely on simple frequency estimation task for discrete data, leaving correlated key-value data - which requires both discrete frequency estimation for keys and continuous mean estimation for values - unexplored. To bridge this gap, we propose KV-Auditor, a framework for auditing LDP-based key-value estimation mechanisms by estimating their empirical privacy lower bounds. Rather than traditional LDP auditing methods that relies on binary output predictions, KV-Auditor estimates this lower bound by analyzing unbounded output distributions, supporting continuous data. Specifically, we classify state-of-the-art LDP key-value mechanisms into interactive and non-interactive types. For non-interactive mechanisms, we propose horizontal KV-Auditor for small domains with sufficient samples and vertical KV-Auditor for large domains with limited samples. For interactive mechanisms, we design a segmentation strategy to capture incremental privacy leakage across iterations. Finally, we perform extensive experiments to validate the effectiveness of our approach, offering insights for optimizing LDP-based key-value estimators.

Problem

• Auditing LDP for correlated key-value data privacy
• Addressing gaps in LDP auditing for continuous data
• Proposing KV-Auditor to estimate empirical privacy bounds

Innovation

• Estimates empirical privacy lower bounds
• Classifies LDP mechanisms into interactive types
• Proposes horizontal and vertical KV-Auditor