Fast Evaluation of Homomorphic Encryption Schemes Based on Ring-LWE

Abstract

When evaluating Homomorphic Encryption (HE) schemes, only one set of input parameters is usually considered. Evaluation reports HE scheme time execution and memory consumption since these are the main challenges of HE. The original version of PAnTHErS enables to evaluate HE schemes without executing the scheme, hence with a very low processing time (< 10 sec). Results are provided in terms of computational complexity and memory cost. This allows to evaluate a scheme for numerous sets of input parameters. In this paper, PAnTHErS is improved by a calibration phase, and four HE schemes based on Ring-LWE are analyzed and compared using the proposed tool. Experimentation results show the approach allows analyses of HE scheme with an average of 6% error for a given implementation with a speedup up to 15× compared to actual scheme executions.