The ring-learning with errors (R-LWR) problem is utilized to build many ciphers resisting quantum-computing attacks and fully homomorphic encryption that allows computations to be carried out on encrypted data. Modular multiplication of long polynomials with large coefficients is the most critical operation in these schemes. The polynomial multiplication complexity can be reduced by the Karatsuba formula. In this paper, a new method is proposed to integrate the modular reduction into the Karatsuba polynomial multiplication. Modular reduction is applied to intermediate segment products instead of the final product. As a result, additional substructure sharing is enabled and the number of coefficient additions needed for assembling the segment products to get the final result is substantially reduced. For polynomial multiplications with decomposition factors 2, 3, and 4, the proposed scheme reduces the number of additions by 13-17%.
|Original language||English (US)|
|Number of pages||5|
|Journal||ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings|
|State||Published - Jun 6 2021|
|Event||2021 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2021 - Virtual, Toronto, Canada|
Duration: Jun 6 2021 → Jun 11 2021
Bibliographical noteFunding Information:
This work is supported in part by Semiconductor Research Corporation under contract number 2020-HW-2988.
© 2021 IEEE
- Fully homomorphic encryption
- Karatsuba multiplication
- Modular polynomial multiplication
- Ring-learning with errors (R-LWE)
- Substructure sharing