Efficient arithmetic on ARM‐NEON and its application for high‐speed RSA implementation. Issue 18 (24th January 2017)

• Record Type: Journal Article
• Authors: Seo, Hwajeong; Liu, Zhe; Großschädl, Johann; Kim, Howon
• Is Part Of: Security and communication networks. Volume 9:Issue 18(2016)
• Page Start: 5401
• Page End: 5411
• Publication Date: 2017-01-24
• Subjects: public‐key cryptography -- modular arithmetic -- SIMD‐level parallelism -- vector instructions -- ARM‐NEON -- RSA; Computer networks -- Security measures -- Periodicals; Computer security -- Periodicals; Cryptography -- Periodicals; 005.805
• DOI: 10.1002/sec.1706
• Languages: English
• ISSNs: 1939-0114

Abstract

Abstract: Advanced modern processors support single instruction, multiple data instructions (e.g., Intel‐AVX and ARM‐NEON) and a massive body of research on vector‐parallel implementations of modular arithmetic, which are crucial components for modern public‐key cryptography ranging from Rivest, Shamir, and Adleman (RSA), ElGamal, Digital Signature Algorithm (DSA), and elliptic curve cryptography, have been conducted. In this paper, we introduce a novel double operand scanning method to speed up multi‐precision squaring with non‐redundant representations on single instruction, multiple data architecture where the part of the operands are doubled to compute the squaring operation without read‐after‐write dependencies between source and destination variables. Afterwards, Karatsuba algorithm is applied to both multiplication and squaring operations. For modular multiplication, separated Montgomery algorithm is chosen. Finally, the Rivest, Shamir, and Adleman (RSA) implementations outperform the best‐known results on the ARM‐NEON platforms. Copyright © 2017 John Wiley & Sons, Ltd. Abstract : In this paper, we introduce a novel double operand scanning method to speed up multi‐precision squaring with non‐redundant representations on single instruction, multiple data architecture where the part of the operands is doubled to compute the squaring operation. For modular multiplication, separated Montgomery algorithm is chosen. Finally, the Rivest, Shamir, and Adleman implementations outperform the best known results on the ARM‐NEON platforms.