THE QUANTUM INTELLIGENCER

Vectorized SQIsign: Speeding Up Isogeny Signatures with AVX-512 In Plain English: Digital signatures keep online communications secure, but current ones might not survive quantum computers. SQIsign is a new type of signature that could resist quantum attacks, but it’s too slow to use widely. This team made it much faster by optimizing it to run on modern CPUs that can process multiple calculations at once. They achieved more than double the speed for signing and verifying. This matters because it brings a promising quantum-safe tool closer to real-world use, especially as hardware support for these optimizations is confirmed to continue in future processors. Summary: SQIsign is the only isogeny-based digital signature scheme in the NIST Post-Quantum Cryptography standardization process. It offers compact keys and signatures, but its signing process is computationally heavy, limiting practical use. This paper introduces the first fully vectorized implementation of SQIsign using AVX-512's Integer Fused Multiply-Add (IFMA) instruction set. The authors redesigned multiple layers of the computational stack—prime and extension field arithmetic, elliptic curve operations like batched point doubling and scalar multiplication, and pairing computations using cubical arithmetic and 2D isogeny evaluations—to exploit SIMD (Single Instruction, Multiple Data) parallelism. The optimized implementation achieves a 2.69x speedup in signing and 3.18x in verification at NIST security level I when combined with Qlapoti technology. These gains are significant because they reduce one of the main barriers to deploying SQIsign in real systems. The work demonstrates that hardware-aware optimization can substantially improve performance even without changes to the underlying algorithm. The authors also address concerns about AVX-512's future, noting that Intel's upcoming AVX10 architecture (revision 10.2, expected late 2026) will standardize AVX-512 capabilities—including IFMA—across both performance and efficiency cores. This ensures long-term viability for such optimizations. The vectorization strategies developed are not limited to SQIsign and can serve as a foundation for accelerating other isogeny-based cryptographic schemes. The paper concludes that SIMD vectorization is a crucial, underutilized dimension for enhancing post-quantum cryptography performance. Key Points: - SQIsign is the only isogeny-based signature in NIST's post-quantum cryptography competition, known for small key sizes but slow signing. - This work presents the first full vectorization of SQIsign using AVX-512 IFMA instructions, optimizing multiple mathematical layers. - Performance improves by 2.69x for signing and 3.18x for verification at NIST level I when combined with Qlapoti. - Optimizations include vectorized prime/extension field arithmetic, elliptic curve operations, and isogeny evaluations. - Contrary to assumptions, AVX-512 (including IFMA) will be standardized in Intel’s AVX10 architecture by late 2026. - The vectorization approach is generalizable to other isogeny-based cryptographic systems. - SIMD acceleration is presented as a critical, underexplored path for improving post-quantum scheme efficiency. - The work makes SQIsign more viable for real-world deployment despite its algorithmic complexity. Notable Quotes: - "Our implementation attains a $2.69\times$ speedup for signing and a $3.18\times$ improvement for verification at NIST security level I." - "Contrary to misconceptions regarding the obsolescence of AVX-512, we emphasize that Intel's AVX10 instruction set architecture... will standardize AVX-512 capabilities." - "SIMD vectorization represents a critical yet underexplored optimization dimension for post-quantum isogeny-based schemes." - "We achieve substantial performance improvements over the reference implementation." - "Our vectorization strategies are architecture-agnostic and provide a methodological foundation applicable to broader isogeny-based cryptographic constructions." Data Points: - 2.69x speedup in signing performance (with Qlapoti). - 3.18x speedup in verification performance (with Qlapoti). - Optimization targets NIST post-quantum security level I. - AVX-512 IFMA instructions used for vectorized arithmetic. - Intel AVX10 revision 10.2 scheduled for late 2026. - First comprehensive vectorized implementation of SQIsign. - Speed improvements achieved across prime-field, extension-field, and elliptic curve operations. - Pairing computations optimized via cubical arithmetic and 2D isogeny evaluations. - Vectorization applied to batched point doubling and scalar multiplication. - Methodology includes redesign of the entire computational stack for SIMD parallelism. Controversial Claims: - The claim that AVX-512 is not obsolete and will be standardized in AVX10 may be contested given Intel's previous de-prioritization of AVX-512 in client CPUs. - The assertion that SIMD vectorization is a 'critical yet underexplored' dimension may downplay prior work in vectorizing lattice-based or hash-based PQC schemes. - The integration with Qlapoti technology is mentioned as contributing to speedups, but the paper does not clarify how much of the gain comes from vectorization versus Qlapoti itself. - The generalizability of the vectorization strategy to other isogeny schemes is claimed but not demonstrated with additional implementations. Technical Terms: - SQIsign: A post-quantum digital signature scheme based on supersingular isogeny problems. - AVX-512: Advanced Vector Extensions, a SIMD instruction set for x86 processors enabling parallel computation. - IFMA (Integer Fused Multiply-Add): AVX-512 instructions that perform integer multiply-add operations in one step, accelerating modular arithmetic. - Isogeny: A special type of mapping between elliptic curves used in cryptographic constructions. - Endomorphism ring problem: A hard mathematical problem underlying the security of isogeny-based cryptography. - SIMD (Single Instruction, Multiple Data): A parallel computing paradigm where one instruction operates on multiple data points simultaneously. - Cubical arithmetic: A method for efficiently computing pairings on elliptic curves. - NIST security level I: A defined level of cryptographic strength, equivalent to 128-bit symmetric security. - Vectorization: The process of rewriting code to exploit SIMD capabilities for performance gains. - Qlapoti: A technology integrated with the implementation to enhance performance, though not detailed in the abstract. —Inspector Grey Dispatch from The Prepared E0