Privacy-Preserving Multi-user Outsourced Computation for Boolean Circuits

With the prevalence of outsourced computation, such as Machine Learning as a Service, protecting the privacy of sensitive data throughout the whole computation is a critical yet challenging task. The problem becomes even more tricky when multiple sources of input and/or multiple recipients of output are involved, who would encrypt/decrypt data using different keys. Considering many computation tasks demand binary operands and operations but there are only outsourced computation constructions for arithmetic calculations [1], in this paper, the authors propose a privacy-preserving outsourced computation framework for Boolean circuits. The proposed framework can protect sensitive data throughout the whole computation, i.e., input, output and all the intermediate values, ensuring privacy for general outsourced tasks. Moreover, it compresses the ciphertext domain of [1] and attains secure protocols for four logic gates (AND, OR, NOT, and XOR) which are the basic operations in Boolean circuits. With the proposed framework as a building block, a novel Privacy-preserved (encrypted) Bloom Filter and a Multi-keyword Searchable Encryption scheme under the multi-user setting are presented. Security proof and experimental results show that the proposal is reliable and practical.