PrivatePond: Outsourced Management of Web Corpuses

With the rise of cloud computing, it is increasingly attractive for end-users (organizations and individuals) to outsource the management of their data to a small number of largescale service providers. In this paper, we consider a user who wants to outsource storage and search for a corpus of web documents (e.g., an intranet). At the same time, the corpus may contain confidential documents that the organization does not want to reveal to the service provider. While past work has considered the problems of secure keyword search and secure indexing, all of the proposed tools require significant modifications to existing search engines and infrastructure. In this paper, we propose a system called PrivatePond, which allows confidential outsourced web search using an unmodified search engine. The system is built around the central idea of a secure indexable representation, which is attached to each document in the corpus, and constructed with the goal of balancing confidentiality and searchability. In addition, a secure local proxy is used to provide transparency to the end-user. While the idea of a secure indexable representation is very general, we propose a preliminary instantiation of this idea, which provides practical confidentiality. In addition, an experimental evaluation indicates that this indexable representation can provide high-quality search and ranking, similar to what is available using the unmodified corpus.