Paper 2016/847 Original: eprint.iacr.org/2016/847 ↗

On the smallest ratio problem of lattice bases

Abstract

Let $(\mathbf{b}_1, \ldots, \mathbf{b}_{n})$ be a lattice basis with Gram-Schmidt orthogonalization $(\mathbf{b}_1^{\ast}, \ldots, \mathbf{b}_{n}^{\ast})$, the quantities $\|\mathbf{b}_{1}\|/\|\mathbf{b}_{i}^{\ast}\|$ for $i = 1, \ldots, n$ play important roles in analyzing lattice reduction algorithms and lattice enumeration algorithms. In this paper, we study the problem of minimizing the quantity $\|\mathbf{b}_{1}\|/\|\mathbf{b}_{n}^{\ast}\|$ over all bases $(\mathbf{b}_{1}, \ldots, \mathbf{b}_{n})$ of a given $n$-dimensional lattice. We first prove that there exists a basis $(\mathbf{b}_{1}, \ldots, \mathbf{b}_{n})$ for any lattice $L$ of dimension $n$ such that $\|\mathbf{b}_1\| = \min_{\mathbf{v} \in L\backslash\{\mathbf{0}\}} \|\mathbf{v}\|$, $\|\mathbf{b}_{1}\|/\|\mathbf{b}_{i}^{\ast}\| \leq i$ and $\|\mathbf{b}_{i}\|/\|\mathbf{b}_{i}^{\ast}\| \leq i^{1.5}$ for $1 \leq i \leq n$. This leads us to introduce a new NP-hard computational problem, that is, the smallest ratio problem (SRP): given an $n$-dimensional lattice $L$, find a basis $(\mathbf{b}_{1}, \ldots, \mathbf{b}_{n})$ of $L$ such that $\|\mathbf{b}_{1}\|/\|\mathbf{b}_{n}^{\ast}\|$ is minimal. The problem inspires the new lattice invariant $\mu_{n}(L) = \min\{\|\mathbf{b}_1\|/\|\mathbf{b}_n^{\ast}\|: (\mathbf{b}_1, \ldots, \mathbf{b}_n) \textrm{ is a basis of } L\}$ and new lattice constant $\mu_{n} = \max \mu_{n}(L)$ over all $n$-dimensional lattices $L$: both the minimum and maximum are justified. The properties of $\mu_{n}(L)$ and $\mu_{n}$ are discussed. We also present an exact algorithm and an approximation algorithm for SRP. This is the first sound study of SRP. Our work is a tiny step towards solving an open problem proposed by Dadush-Regev-Stephens-Davidowitz (CCC '14) for tackling the closest vector problem with preprocessing, that is, whether there exists a basis $(\mathbf{b}_{1}, \ldots, \mathbf{b}_{n})$ for any $n$-rank lattice such that $\max_{1 \le i \le j \le n} \|\vec{b}_{i}^{\ast}\|/\vec{b}_{j}^{\ast}\| \le \textrm{poly}(n)$.