Acceleration Techniques for Fully-Homomorphic Encryption Schemes

It is known that cloud computing opened new opportunities for many organizations regarding the cost-savings and the availability rate. Homomorphic encryption schemes are well known for solving the privacy preserving issue, but these schemes bring deficiency in execution times, therefore accelerating these encryption schemes using different technologies is the next step towards the wide use of homomorphic primitives. The main contribution is an implementation of a homomorphic encryption library using multiple GPU's, more precisely an adaptation of the HElib library. While other single GPU accelerations of HElib do exist, the proposed implementation uses multiple graphic processors units to distribute computational tasks.