A Pattern Group Partitioning for Parallel String Matching using a Pattern Grouping Metric

Abstract

Considering the increasing number of target patterns for the intrusion detection systems (IDS), memory requirements should be minimized for reducing hardware overhead. This paper proposes an algorithm that partitions a set of target patterns into multiple subgroups for homogeneous string matchers. Using a pattern grouping metric, the proposed pattern partitioning makes the average length of the mapped target patterns onto a string matcher approximately equal to the average length of total target patterns. Therefore, the variety of target pattern lengths can be mitigated because the number of mapped target patterns onto each string matcher is balanced.