Preserving Privacy of Co-occurring Keywords over Encrypted Data

Abstract

The indexes of ranked searchable encryption contain encrypted keywords and their encrypted relevance scores. The encryption scheme of relevance scores must preserve the plaintext order after encryption so as to enable the cloud server to determine ranks of the documents directly from the encrypted keywords' scores for a given trapdoor. Existing schemes such as Order Preserving Encryption (OPE) and One-to-Many OPE preserve the plaintext order. However, they leak the distribution information, i.e., the frequency of ciphertext values, due to the insufficient randomness employed in these schemes. The cloud server uses frequency analysis attack to infer plaintext keywords of the indexes based on the frequency leakage. In this paper, an Enhanced One-to-Many OPE scheme is proposed to minimize the frequency leakage. The proposed scheme reduces not only the frequency leakage of individual keywords but also the co-occurring keywords of the phrases like "computer network", and "communication network".