Authenticating kNN Query Results in Data Publishing

Abstract

In data publishing model, data owners engage third-party data publishers to manage their data and process queries on their behalf. As the publishers may be untrusted or susceptible to attacks, it could produce incorrect query results. In this paper, we extend the signature-based mechanism for users to verify that their answers for k nearest neighbors queries on a multidimensional dataset are complete (i.e. no qualifying data points are omitted), authentic (i.e. no answer points are tampered) and minimal (i.e. no non-answer points are returned in the plain). Essentially, our scheme returns k answer points in the plain, and a set of ($\tilde{p}, q$)-pairs, where $\tilde{p}$ is the digest of a non-answer point p in the dataset used to facilitate the signature chaining mechanism to verify the authenticity of the answer points, and q is a reference point (not in the dataset) used to verify that p is indeed further away from the query point than the kth nearest point. We study two instantiations of the approach - one based on the native data space using space partitioning method (a.k.a. R-tree) and the other based on the metric space using iDistance. We conducted an experimental study, and report our findings here.