Secure Similarity Search

Abstract

One of the most substantial ways to protect users' sensitive information is encryption. This paper is about the keyword index search system on encrypted documents. It has been thought that the search with errors over encrypted data is impossible because 1 bit difference over plaintexts may reduce to enormous bits difference over cyphertexts. We propose a novel idea to deal with the search with errors over encrypted data. We develop two similarity search schemes, implement the prototypes and provide substantial analysis. We define security requirements for the similarity search over encrypted data. The first scheme can achieve perfect privacy in similarity search but the second scheme is more efficient.