A Secure Computation Scheme of Inner Product Based on Fully Homomorphic Encryption

Among many approaches to solve the problem of top-k retrieval over encrypted cloud data, we focus on an approach with homomorphic encryption, which is public key encryption supporting some operations on encrypted data. In top-k retrieval of encrypted data, the inner product is often used as a metric to compute the similarity between the file feature vector and the query vector. In this paper, we propose an efficient scheme to compute the inner product on encrypted data using the homomorphic encryption based on the learning with errors over ring (RLWE) problem, in which batch and packing techniques are adopted to achieve lower computation and communication cost.