Armed with Faster Crypto: Optimizing Elliptic Curve Cryptography for ARM Processors.

Area of Science:

• Computer Science
• Cryptography
• Embedded Systems

Background:

• Elliptic curve cryptography (ECC) is fundamental for secure digital communications, underpinning key agreement protocols, digital signatures, and zero-knowledge proofs.
• High-performance ECC relies on micro-architecture-specific modular arithmetic optimizations.
• ARM processors, prevalent in smartphones, are a key target for performance enhancements.

Purpose of the Study:

• To evaluate and optimize the Curve25519 elliptic curve for ARM processors.
• To leverage the ARM NEON Single Instruction, Multiple Data (SIMD) architecture for accelerated cryptographic operations.

Main Methods:

• Introduced a novel 128-bit NEON SIMD vector representation optimized for parallelization.
• Implemented an extended twisted Edwards curve Curve25519 backend in the Rust library 'curve25519-dalek'.
• Conducted extensive evaluations on multiple ARM devices using cryptographic benchmarks and Signal protocol benchmarks.

Main Results:

• Achieved a backend speed-up of at least 20% for ARM NEON.
• Demonstrated a notable speed improvement of at least 15% for benchmarked Signal functions.
• The novel SIMD vector representation significantly accelerates elliptic curve operations.

Conclusions:

• The optimized Curve25519 implementation offers substantial performance gains on ARM NEON architectures.
• This work enhances the efficiency of secure communication protocols on mobile devices.
• The findings pave the way for faster and more efficient cryptographic operations in embedded systems.