High-Speed Implementation of Lattice Enumeration with Discrete Pruning for Solving the SVP

The emergence of quantum computing poses a significant threat to contemporary mainstream cryptographic systems, such as RSA, Diffie-Hellman and ECC. In this context, lattice-based cryptography has garnered widespread interest among researchers due to its potential resistance to quantum attacks. Lattice cryptography's security is underpinned by a set of hard problems, with the Shortest Vector Problem (SVP) being one of the most fundamental and crucial. Discrete pruning enumeration is one of the key algorithms used to solve the SVP. In this paper, we present a parallel version of the discrete pruning enumeration and optimize it for CPU-GPU heterogeneous systems. The experimental results show that the parallel implementation using both CPU and GPU can achieve up to a 54x speed-up in the 100-dimensional SVP challenge, and up to a 77x speed-up in the 95-dimensional instances.