Title: Rache: Radix-additive caching for homomorphic encryption

Abstract: One of the biggest concerns for many applications in cloud computing lies in data privacy. A potential solution to this problem is homomorphic encryption (HE), which supports certain operations directly over the ciphertexts. Conventional HE schemes, however, exhibit significant performance overhead and are hardly applicable to real-world applications. This paper presents Rache, a caching optimization for accelerating the performance of HE schemes. The key insights of Rache include (i) caching some homomorphic ciphertexts before encrypting the large volume of plaintexts; (ii) expanding the plaintexts into a summation of powers of radixes; and (iii) constructing the ciphertexts with only homomorphic addition. The extensive evaluation shows that Rache exhibits almost linear scalability and outperforms Paillier by orders of magnitude.