International Association
for Cryptologic Research

Recently, a new framework, called secure server-designation public key encryption with keyword search (SPEKS), was introduced to improve the security of dPEKS (which suffers from the on-line keyword guessing attack) by defining a new security model ‘original ciphertext indistinguishability’. In this paper, we note that off-line keyword guessing attack can be launched by a malicious server to find the keyword used for generating the trapdoor, which was not considered in the related work. SPEKS can suffer from this kind of attack. Moreover, the security model defined for TD-IND in SPEKS is incomplete. Owing to the shown weaknesses, the existing security models are enhanced for trapdoor indistinguishability by defining two new security models. Finally, we propose a new framework.