18 Nov 2026, Wed

Researchers Push Post-Quantum Cryptography Forward with Vectorized AVX-512 Speedups for SQIsign

A team of international cryptographers has unveiled a major performance breakthrough for post-quantum cryptography, introducing a high-performance implementation of higher-dimensional isogenies utilizing AVX-512 vector instructions. The research, detailed in Paper 2026/2773 on the Cryptology ePrint Archive, focuses on vectorized (2, 2)-isogenies within the theta model and delivers dramatic speedups for SQIsign, one of the leading candidates in post-quantum signature schemes.

The collaborative effort brings together researchers Marius A. Aardal and Diego F. Aranha from Aarhus University, Patrick Longa from Microsoft in the United States, and Giacomo Pope, representing both the University of Bristol and NCC Group. Their work addresses one of the most persistent bottlenecks in modern isogeny-based cryptography: the heavy computational cost associated with higher-dimensional isogeny evaluations.

In recent years, higher-dimensional isogenies have risen to prominence as a powerful structural foundation for advanced cryptographic protocols. Among them, SQIsign has attracted substantial attention from the global cryptographic community due to its exceptional security guarantees and remarkably compact public and signature sizes. These properties make it an intensely attractive candidate for standardization and real-world deployment in an era anticipating the advent of cryptographically relevant quantum computers.

However, this theoretical elegance has historically come at a steep operational price. Despite offering compelling security and data footprint profiles, schemes relying on these mathematical structures remain computationally expensive. Bridging the gap between their theoretical promise and practical, real-world utility requires aggressive performance optimization across both algorithmic design and underlying hardware utilization.

To tackle this challenge, the research team developed a specialized code generator designed to produce highly efficient AVX-512 implementations of arithmetic over finite fields. Specifically, the generated code handles operations over prime fields and their quadratic extensions for a specific, highly relevant family of Montgomery-friendly primes. These primes take the standard cryptographic form where $p$ is equal to $c$ multiplied by two raised to the power of $e$, minus one, with $e$ closely approximating the base-2 logarithm of $p$. This particular mathematical structure underpins several protocols dependent on two-dimensional isogenies, placing SQIsign squarely in its crosshairs.

Building directly upon this optimized foundational arithmetic, the authors successfully vectorized the complex (2, 2)-isogeny evaluation formulas recently introduced by Dartois and collaborators at ASIACRYPT 2024. By effectively exploiting the data-level parallelism natively provided by modern AVX-512 vector registers, the new implementation achieves profound performance gains. These speedups far outperform corresponding highly optimized scalar implementations written in traditional x86-64 assembly language, demonstrating the untapped potential of modern CPU vector extensions for advanced algebraic geometry operations.

The practical implications of these algorithmic and architectural improvements are substantial when applied to cryptographic signature protocols. The research team integrated their vectorized framework into the Round 2 specifications of SQIsign, recording dramatic verification speedups across multiple security tiers. Specifically, the implementation yields a 2.5-fold performance increase at NIST security level I. The gains become even more pronounced at higher security brackets, registering 3.2-fold and 3.3-fold speedups at NIST security levels III and V, respectively.

Furthermore, the team’s automated code generator demonstrates high flexibility and adaptability by readily supporting the larger prime configurations introduced in the Round 3 specifications of SQIsign. For these more demanding parameter sets, the researchers observed verification speedups ranging between 3.1-fold and 3.4-fold, proving the scalability of their vectorization strategy across evolving protocol standards.

By significantly lowering the computational overhead of signature verification, this work marks an important step toward making resource-intensive post-quantum protocols viable for everyday deployment in constrained environments, network servers, and embedded systems alike. The complete details of the methodology, mathematical formulas, and integration strategies are documented in the paper now available to the academic and industrial cryptographic community via the Cryptology ePrint Archive.

Leave a Reply

Your email address will not be published. Required fields are marked *