Efficient Integer Vector Homomorphic Encryption

Homomorphic Encryption yields deliberately malleable ciphertext, allowing operations on encrypted data. We outline and implement a homomorphic encryption scheme as suggested by Zhou and Wornell [1] that encrypts integer vectors to allow computation of arbitrary polynomials in the encrypted domain with a bounded degree. This finds applications particularly in cloud computation, when one is interested in learning low dimensional representations of the stored encrypted data. With our optimizations to the scheme, we find that it is orders of magnitude faster than HELib, an implementation of fully homomorphic encryption, but in turn a few orders of magnitude slower than operations on unencrypted data. This makes the scheme nearly viable for practical use, which we give basic implementations of such applications as search and classification on encrypted data.