Secure, flexible and high-efficient similarity search over encrypted data in multiple clouds

In cloud computing, a cloud user can improve a system’s reliability by utilizing services from multiple clouds, known as the multi-cloud service model. This model, however, causes serious security concerns since multi-clouds increase the chance of being attacked. To maintain data security, a commonly-used strategy is to encrypt data stored in clouds. Yet, issues such as inconsistent data storage (Byzantine faults), dynamic data update, and inefficient data retrieval among multiple clouds, have remained as open challenges. In this paper, we propose a multi-cloud secure similarity search (MC3S) method to effectively and efficiently utilize the encrypted data over multiple clouds. To achieve secure, flexible and high-efficient data search, MC3S utilizes two novel data structures, called gram-filter and prefix-ring. Different from prior works in secure data search, we address the problem of Byzantine faults among multiple clouds, and our algorithm enables dynamic data index insertion and revocation without decryption. We prove that MC3S can resist chosen-keyword attacks and achieve non-adaptive semantic security. Finally, we evaluate the efficiency of MC3S with real world datasets.