An Efficient and Parallel R-LWE Cryptoprocessor
Yuqing Zhang, Chenghua Wang, Dur‐e‐Shahwar Kundi, Ayesha Khalid, Máire O’Neill, Weiqiang Liu
Abstract
Lattice-based cryptography (LBC) is a promising and efficient public key cryptography scheme whose theoretical foundation usually lies in Learning with Error (LWE) problem and its variant such as Ring-LWE (R-LWE) is the most studied cryptosystem which allows for more efficient implementation while maintaining the hardness of an original problem. Polynomial multiplication is the bottleneck of R-LWE, that can either be done using Number Theoretic Transform (NTT) or schoolbook polynomial multiplication (SPM) algorithm. The use of SPM is wider and possible for all parameters of R-LWE schemes. This brief proposes an efficient and parallel strategy for SPM in R-LWE; by successfully reducing its time complexity from n <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> to n <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> /4 (making it 1.8× faster and 1.4× hardware efficient). Furthermore, by adjusting the bit width for the error terms, the polynomial multiplication and addition blocks are reused for both encryption and decryption modules resulting in 14% reduced area and 1.7× better throughput in comparison to state-of-art SPM based R-LWE designs.