Obfuscation via Information Density Estimation

TL;DR

This paper introduces a data-driven framework for identifying and obfuscating features that leak sensitive information using a novel information density estimator, with proven leakage guarantees and practical implementation on real datasets.

Contribution

We propose a new framework utilizing information density estimation to identify and obfuscate leaking features, including a novel estimator called TIDE with theoretical guarantees.

Findings

Effective identification of information-leaking features.

Successful implementation of obfuscation with leakage guarantees.

Validation on three real-world datasets.

Abstract

Identifying features that leak information about sensitive attributes is a key challenge in the design of information obfuscation mechanisms. In this paper, we propose a framework to identify information-leaking features via information density estimation. Here, features whose information densities exceed a pre-defined threshold are deemed information-leaking features. Once these features are identified, we sequentially pass them through a targeted obfuscation mechanism with a provable leakage guarantee in terms of $\mathsf{E}_\gamma$-divergence. The core of this mechanism relies on a data-driven estimate of the trimmed information density for which we propose a novel estimator, named the trimmed information density estimator (TIDE). We then use TIDE to implement our mechanism on three real-world datasets. Our approach can be used as a data-driven pipeline for designing obfuscation mechanisms targeting specific features.