BoostReduce - A Framework For Strong Lattice Basis Reduction

In this paper, we propose a new generic reduction framework BoostReduce for strong lattice basis reduction. At the core of our new framework is an iterative method which uses a newly-developed algorithm for finding short lattice vectors and integrating them efficiently into an improved lattice basis. We present BoostBKZ as an instance of BoostReduce using the Block-Korkine-Zolotarev (BKZ) reduction. BoostBKZ is tailored to make effective use of modern computer architectures in that it takes advantage of multiple threads. Experimental results of BoostBKZ show a significant reduction in running time while maintaining the quality of the reduced lattice basis in comparison to the traditional BKZ reduction algorithm.