Towards a Very Fast Multiple String Matching Algorithm for Short Patterns

Multiple exact string matching is one of the fundamental problems in computer science and finds applications in many other fields, among which computational biology and intrusion detection. It turns out that short patterns appear in many instances of such problems and, in most cases, sensibly affect the performances of the algorithms. Recent solutions in the field of string matching try to exploit the power of the word RAM model to speed-up the performances of classical algorithms. In this model an algorithm operates on words of length w, grouping blocks of characters, and arithmetic and logic operations on the words take one unit of time. This study presents a first preliminary attempt to develop a filter based exact multiple string matching algorithm for searching set of short patterns by taking benefit from Intel's SSE (streaming SIMD extensions) technology. Our experimental results on small, medium, and large alphabet text files show that the proposed algorithm is competitive in the case of short patterns against other efficient solutions, which are known to be among the fastest in practice.