Memory-efficient pattern matching for network intrusion detection.

摘要

In this work we present multiple designs for memory efficient pattern matching. We present several DFA optimization techniques, as well as multiple methods for deriving state signatures to be used with our novel architecture. A methodology for combining these optimizations and signatures is introduced, allowing for the creation of reconfigurable pattern matching engines suitable for matching strings or regular-expressions. Our designs are exceptionally memory-efficient and can achieve memory reductions of over 600 times compared to a straightforward, memory-based DFA approach.