Draft · v0.2 · Unreviewed
← Index/Project · Nadarasa ReductionDraft · v0.2 · Selene Emulator
← Overview

Lattice / LWE

Where every NIST post-quantum standard rests, and where the last serious quantum attack visibly failed at extraction, not preparation.

State of the art

Best known quantum sieve for SVP is 2^(0.2570 n + o(n)) (Chailloux–Loyer 2021 via quantum random walks on Johnson graphs), tightened by Bonnetain–Chailloux–Schrottenloher–Shen 2023 (reusable quantum walks) and pushed slightly further into 3-tuple sieving by Engelberts et al. 2025. Classical baseline (BDGL 2016) is 2^(0.2920 n) — a modest ~10% exponent gap. Cho et al. 2024 prove the gap requires poly(n) QRAM; in a circuit-only model, no super-constant quantum advantage over classical sieve currently survives. Regev's 2005 quantum reduction runs in the hardness direction (GapSVP →_Q LWE) and is not an attack. Chen's April 2024 attempt (ePrint 2024/555) at a polynomial-time quantum LWE algorithm was withdrawn within a week; the surgical post-mortem (Hongxun Wu + Thomas Vidick, April 18 2024) located the bug in Step 9 / Section 3.5.9: the state exiting Step 8 is globally entangled, breaking the product-state assumption Step 9 needs. Steps 1–8 (complex Gaussian prep + windowed QFT) were not refuted and may contain salvageable primitives.

Known dead ends (read first)

  • ×Chen 2024 Step 9 extraction loop — the state exiting Step 8 is globally entangled across the LWE secret coordinates; the iterative single-bit measurement collapses the joint state and destroys the bootstrap. Chen ePrint 2024/555 §3.5.9 (WITHDRAWN); post-mortem Hongxun Wu & Thomas Vidick, 2024-04-18
  • ×Eldar–Shor 2016 polynomial-time quantum lattice algorithm — flaw found by Regev within days, paper abandoned. Direct precedent for Chen 2024. arXiv:1611.06999 (Eldar–Shor 2016, withdrawn/abandoned)
  • ×DHSP route to polynomial-time LWE — 20+ years of effort hit the 2^O(√n) sieve wall; complexity evidence suggests DHSP is not in BQP. Kuperberg quant-ph/0302112; Regev quant-ph/0406151
  • ×Pure Grover-over-enumeration — gives only quadratic speedup, easily absorbed by parameter increase. Laarhoven 2016 thesis; Aono–Nguyen–Shen 1709.01411
  • ×Variational / QAOA approaches to SVP/CVP — NISQ heuristics with no proven exponential speedup and no convergence guarantees in high dimensions. arXiv:2503.08403; arXiv:2505.08386
  • ×Quantum sieve speedup without QRAM — proven to require poly(n)-width QRAM; in a circuit-only model the 0.265 vs 0.292 advantage vanishes. Cho, Hhan, Kim, Lee, Shen arXiv:2410.15565

Open sub-problems

Rescue a narrow case of Chen's Step 9 for structured moduli (power-of-two cyclotomic / q = p^k with small p): formalize what algebraic condition on the post-Step-8 state would permit ring-aware single-coordinate measurement; check Ring-LWE.

Tractability
2/5

QRAM-free quantum speedup for the dual attack via quantum rejection sampling on the lattice Gaussian sampling bottleneck — adapt sequential-query rejection sampling to the Klein/GPV sampler.

Tractability
3/5

Tight quantum reduction from GapSVP_{√n·polylog(n)} to LWE — sharper Gaussian tail analysis inside Regev's dual-lattice sampling, shaving a polynomial factor and closing the dark zone around Kyber/Dilithium parameters.

Tractability
3/5

Paired donor analogies

Key papers

On lattices, learning with errors, random linear codes, and cryptography
Regev · 2005
2401.03703accepted

The keystone. Quantum worst-case-to-average-case reduction GapSVP/SIVP →_Q LWE; a security argument, not an attack — flows in the wrong direction for the attacker.

Quantum algorithms for lattice problems
Chen · 2024
ia.cr/2024/555withdrawn

WITHDRAWN. Flaw at Step 9 / §3.5.9 — entangled state breaks the product-state assumption needed for iterative bit extraction. Hongxun Wu and Thomas Vidick identified the bug on 2024-04-18.

Lattice sieving via quantum random walks
Chailloux, Loyer · 2021
2105.05608accepted

Quantum walk on the Johnson graph over sieve lists. Current asymptotic frontier: 2^(0.2570 n) for SVP. Requires QRAM.

Finding many collisions via reusable quantum walks — application to lattice sieving
Bonnetain, Chailloux, Schrottenloher, Shen · 2023
2205.14023accepted

Reusable quantum walks amortize setup across many near-neighbor pairs. Matches / slightly improves the 0.265 barrier. EUROCRYPT 2023.

An improved quantum algorithm for 3-tuple lattice sieving
Engelberts, Chen, Gilani, van Hoof, Jeffery, de Wolf · 2025
2510.08473preprint

Extends reusable quantum walks to 3-tuple sieving — current frontier of quantum SVP algorithms, pushes slightly below 0.2570 n.

Does quantum lattice sieving require quantum RAM?
Cho, Hhan, Kim, Lee, Shen · 2024
2410.15565preprint

Proves the leading quantum sieve speedups require poly(n) QRAM; without QRAM the quantum advantage over classical sieve vanishes. Critical caveat for real-world security estimates.

Quantum augmented dual attack
Albrecht, Shen · 2022
2205.13983accepted

Quantum speedup for the dual attack via QRACM; can outperform pure primal-sieve in some NIST PQC parameter regimes. CRYPTO 2022.

An efficient quantum factoring algorithm
Regev · 2023
2308.06572preprint

Cuts Shor's gate count to Õ(n^{3/2}) by offloading CVP post-processing classically. Directly links lattice hardness to quantum factoring cost.

A decade of lattice cryptography
Peikert · 2016
ia.cr/2015/939accepted

Canonical survey. Read before claiming any 'obvious' quantum attack — most have been tried.