TC: Small: Collaborative Research: Privacy-Constrained Searching
Miami University, Oxford OH
Investigators
Abstract
The goal of this project is to make possible database searching in a privacy-constrained manner: A private database provider allows only properly authorized searches by clients, in a manner that does not reveal the search criteria yet enforces the requirement that the client learns only what is authorized by the search. The initial focus will be on techniques for the case of exact matches, later extended to the much more difficult case of approximate matching. If multiple matches are found, either all of them are returned, or a subset of the "best" of them, under appropriately defined notions of quality, is returned; in approximate matching there is a natural notion of quality, namely, having smaller distance to the target of the search as specified by the query. The technical challenges include verification of the validity of a search request, and then carrying out the search, in manner that enforces the search's authorized criteria without revealing them. The project holds the promise of leading to substantial improvements in the highly unsatisfactory current "state of the practice" for searches carried out on private and sensitive databases, that unnecessarily reveal too much information and prevent useful collaborations from taking place due to concerns over the leakage of sensitive information. The minimal-disclosure feature of the protocols will also make possible a de facto "defense in depth", in that a compromised server will no longer automatically imply the compromise of all the clients' interactions with that server.
View original record on NSF Award Search →