Differentially Private Sketches for Jaccard Similarity Estimation

TL;DR

This paper introduces two locally-differentially private algorithms that extend MinHash for efficient Jaccard similarity estimation, balancing privacy and utility through innovative modifications and theoretical analysis.

Contribution

It presents novel locally-differential private algorithms for Jaccard similarity estimation using extended MinHash with randomized response and Laplace mechanisms.

Findings

The algorithms achieve a favorable privacy-utility trade-off.

Theoretical bounds on absolute error are established.

Experimental results validate the utility of the proposed methods.

Abstract

This paper describes two locally-differential private algorithms for releasing user vectors such that the Jaccard similarity between these vectors can be efficiently estimated. The basic building block is the well known MinHash method. To achieve a privacy-utility trade-off, MinHash is extended in two ways using variants of Generalized Randomized Response and the Laplace Mechanism. A theoretical analysis provides bounds on the absolute error and experiments show the utility-privacy trade-off on synthetic and real-world data. The paper ends with a critical discussion of related work.