Probabilistic Generation Of Trapdoors: Reducing Information Leakage Of Searchable Symmetric Encryption

Searchable symmetric encryption (SSE) enables a user to outsource a collection of encrypted documents in the cloud and to perform keyword searching without revealing information about the contents of the documents and queries. On the other hand, the information (called search pattern) whether or not the same keyword is searched in each query is always leaked in almost all previous schemes whose trapdoors are generated deterministically. Therefore, reducing the search pattern leakage is outside the scope of almost all previous works. In this paper, we tackle to the leakage problem of search pattern, and study methodology to reduce this leakage. Especially, we discuss that it might be possible to reduce the search pattern leakage in cases where a trapdoor does not match any encrypted document. We also point out that the same search pattern is leaked regardless of probabilistic or deterministic generation of trapdoors when the user searches using a keyword which has already searched and matched a certain encrypted document. Thus, we further aim to construct SSE schemes with fast "re-search" process, in addition to reducing the search pattern leakage. In order to achieve the above, we introduce a new technique "trapdoor locked encryption" which can extract a deterministic trapdoor from a probabilistic trapdoor, and then propose a new SSE scheme which can generate trapdoors probabilistically and reduce the search pattern leakage. Our scheme is constructed by applying our technique to the well-known and influential scheme SSE-2 (ACM CCS 2006) and can be proved secure in the standard model.