Abstract
Matrix inversion is one of the most significant operations on a matrix. For any non-singular matrix A∈Zn×n, the inverse of this matrix may contain countless numbers of non-integer entries. These entries could be endless floating-point numbers. Storing, transmitting, or operating such an inverse could be cumbersome, especially when the size n is large. The only square integer matrix that is guaranteed to have an integer matrix as its inverse is a unimodular matrix U∈Zn×n. With the property that det(U)=±1, then U−1∈Zn×n is guaranteed such that UU−1=I, where I∈Zn×n is an identity matrix. In this paper, we propose a new integer matrix G˜∈Zn×n, which is referred to as an almost-unimodular matrix. With det(G˜)≠±1, the inverse of this matrix, G˜−1∈Rn×n, is proven to consist of only a single non-integer entry. The almost-unimodular matrix could be useful in various areas, such as lattice-based cryptography, computer graphics, lattice-based computational problems, or any area where the inversion of a large integer matrix is necessary, especially when the determinant of the matrix is required not to equal ±1. Therefore, the almost-unimodular matrix could be an alternative to the unimodular matrix.
Funder
Fundamental Research Grant Scheme, Ministry of Higher Education Malaysia.
Subject
General Mathematics,Engineering (miscellaneous),Computer Science (miscellaneous)
Reference16 articles.
1. On Lovász’ lattice reduction and the nearest lattice point problem
2. Lattices and Cryptography;Hoffstein,2008
3. Public-key Cryptosystems from Lattice Reduction Problems;Goldreich,1997
4. A Security Upgrade on the GGH Lattice-based Cryptosystem
5. Efficient Key Generation for Rainbow BT;Petzoldt,2020