Exipnos: An Efficient Verifiable Dynamic Symmetric Searchable Encryption Scheme with Forward and Backward Privacy

Abstract

Multiple approaches have been developed to address data privacy concerns, as cloud services increasingly gain traction. One of these methods is Searchable Encryption ($\mathsf {SE}$), which enables a user to search over encrypted data. When applied to a dynamic dataset, it is important that $\mathsf {SE}$ achieves two essential properties upon updating a dynamic dataset: (1) Forward Privacy, which guarantees that an updated document would not be linked to previous searches and (2) Backward Privacy, which prevents information leakage from deleted data.

In this paper, we propose an efficient Verifiable Dynamic Symmetric Searchable Encryption ($\mathsf {VDSSE}$) scheme, achieving forward and backward privacy. The scheme is designed based on the principle of additive secret sharing, where each keyword is assigned a secret and each document containing the keyword is assigned a share of the secret to hide its entry. To support a dynamic update, the last secret share, which is stored only on the client-side, is recursively shared. Each secret share is applied to reconstruct the secret. If the secret is reconstructed correctly, a search result is considered correct. We formally prove the security of the proposed $\mathsf {VDSSE}$ scheme and show its practicality by conducting a large number of experiments over a publicly available dataset 20 Newsgroups. Experimental results show that it takes less than 1 microsecond ($\upmu$s) - on average - to retrieve a document from an encrypted dataset.