Acceleration of a Classic McEliece Postquantum Cryptosystem With Cache Processing
Abstract
The National Institute of Standards and Technology's postquantum cryptography standardization process is in its fourth round, with a first key encapsulation mechanism standard based on learning with errors and three candidates based on error-correcting codes. These primitives' implementation are designed to be optimal on classical hardware architecture targets. However, emerging architectures with processing in memory (PIM), made to be multipurpose, contrary to cryptographic coprocessors, have proven their efficiency in multiple use cases and show better overall computational speed. In this article, we show that classic McEliece performance can be improved on PIM architectures. Notably, the public-key-generation benefits of a 12.6x speedup on architectures with bit-line operations. We also describe an open source RISC-V simulator specifically developed for our experiments, including both in-cache and vectored operations. We discuss how these architecture changes may open the possibility of redesigning primitives or parameter sets for better efficiency.
