Module lattices and their shortest vectors
Nihar Gargava, Vlad Serban, Maryna Viazovska, Ilaria Viglino
TL;DR
This work investigates shortest-vector phenomena for module lattices over number fields, focusing on cyclotomic fields, by extending Rogers-style second-moment analysis to the algebraic setting. The authors develop height-based controls and unit-count bounds to bound the second moment $\\mathbb{E}[\\rho_V(\\Lambda)^2]$ of the number of nonzero lattice vectors in a ball of volume $V$, obtaining the leading term $V^2+\\omega_K V$ with exponentially small errors. They specialize to cyclotomic fields to derive explicit bounds and, in favorable regimes, Poisson-type convergence for the vector counts with mean $V/\\omega_K$, enabling probabilistic estimates for the shortest vector through the volume-minimum $\\mathcal{V}_1(\\Lambda)$. The results yield practically sharp probabilistic guarantees for SVP across various random module lattice models, including discrete-lattice constructions from codes, and provide a concrete workflow for fixed fields (e.g., a 256-dimensional case) to assess cryptographic-security-relevant parameters.
Abstract
We study the shortest vector lengths in module lattices over arbitrary number fields, with an emphasis on cyclotomic fields. In particular, we sharpen the techniques of arXiv:2308.15275v2 to establish improved results for the variance of the number of lattice vectors of bounded Euclidean norm in a random module lattice. We then derive tight probabilistic bounds for the shortest vector lengths for several notions of random module lattice.
