Locally private Jaccard similarity estimation

Abstract

Summary Jaccard Similarity has been widely used to measure the distance between two sets (or preference profiles) owned by two different users. Yet, in the private data collection scenario, it requires the untrusted curator could only estimate an approximately accurate Jaccard similarity of the involved users but without being allowed to access their preference profiles. This paper aims to address the above requirements by considering the local differential privacy model. To achieve this, we initially focused on a particular hash technique, MinHash, which was originally invented to estimate the Jaccard similarity efficiently. We designed the PrivMin algorithm to achieve the perturbation of MinHash signature by adopting Exponential mechanism and build the Locally Differentially Private Jaccard Similarity Estimation (LDP‐JSE) protocol for allowing the untrusted curator to approximately estimate Jaccard similarity. Theoretical and empirical results demonstrate that the proposed protocol can retain a highly acceptable utility of the estimated similarity as well as preserving privacy.

Differentially Private Sketches for Jaccard Similarity Estimation

“…The paper by Yan et al [17] is closest to our approach. It discusses an LDP approach based on MinHash by selecting certain hash values in a differentially private manner using the exponential mechanism.…”

“…The response of the user is the K elements chosen in this way. Given two responses x and ŷ, [17] returns the value |x ∩ ŷ|/K.…”

“…This paper proposes two randomized mechanisms when users have a collection of items and are interested in finding their similarity with other users under the Jaccard similarity in a private manner. The proposed algorithms build upon the papers [11,4,17] and a precise account of the relation will be given in the related work section at the end of this paper. In a nutshell, each user starts by applying MinHash as introduced by Broder [1] with the range compression of Li and König [12] (Section 3.1) to produce a sketch of their data.…”

PPDTS: Predicting potential drug–target interactions based on network similarity

“…In the actual drug–target interactions’ network, a drug can bind on multiple targets, and the same target can be bound by multiple drugs. When the degree of nodes in the DTI network is larger, Zhou [ 26 ] shows that the accuracy of similarity calculation methods such as Adamic‐Adar index [ 31 ], cosine similarity [ 32 ] and Jaccard similarity [ 33 ] coefficient are lower. In addition, it takes a long time and a large amount of memory to calculate the similarity based on the global nodes of the network.…”

Foreword to the Special Issue of the 8th International Conference on Applications and Techniques for Information Security (ATIS 2017)

“…There is a new need for designing differentially private protocols in which the untrusted curator could only estimate an approximately accurate Jaccard similarity of the involved preference profiles but without being allowed to access them directly. The paper “Locally private Jaccard similarity estimation” addressed the above requirements by considering the local differential privacy model . To achieve this, the authors initially focused on a particular hash technique, MinHash, and designed the PrivMin algorithm by the Exponential mechanism to build the LDP‐JSE protocol.…”