Fast Implementation of LSH With SIMD

In this paper, we propose a method of efficient software implementation for the cryptographic hash function LSH with single instruction multiple data (SIMD). The method is based on word-wise permutations of LSH. Using the modified functions
$Step'_{j}=P \circ Step_{j} \circ P^{-1}$
and
$MsgExp'$
instead of the original step function
$Step_{j}$
and message expansion function
$MsgExp$
, where
$P$
is a permutation and
$P^{-1}$
is the inverse permutation of
$P$
, we show that the number of the SIMD instructions for implementing LSH is reduced. For efficient implementation of LSH in other environments (e.g., MIMD), various types of word permutations are listed.