Permission-based Index Clustering for Secure Multi-User Search

Secure keyword search in shared infrastructures prevents stored documents from leaking sensitive information to unauthorized users. A shared index provides confidentiality if it is exclusively used by users authorized to search all the indexed documents. We introduce the Lethe indexing workflow to improve query and update efficiency in secure keyword search. The Lethe workflow clusters together documents with similar sets of authorized users, and creates shared indices for configurable document subsets accessible by the same users. We examine different datasets based on the empirical statistics of a document sharing system and alternative theoretical distributions. We apply Lethe to generate indexing organizations of different tradeoffs between the search and update cost. We show the robustness of our method by obtaining configurations of similar low costs from application of prototype-based and density-based clustering algorithms. With measurements over an open-source distributed search engine, we experimentally confirm the improved search and update performance of the particular indexing configurations that we introduce.