A robust approach for matching mixed case-sensitive and case-insensitive patterns

Abstract

As one of the key methods as well as a bottleneck for Network Intrusion Detection Systems (NIDSes) to detect and eliminate malicious traffic, pattern matching is increasingly gaining popularity while also faces threats from hackers' overloading attempts. The support of mixed case-sensitive and case-insensitive patterns, which is essential for NIDSes to detect possible attacks targeting different applications and operating systems, is currently a potential vulnerability since the widely used Convert-Search-Verify (CSV) approach encounters severe performance degradation in the worst-case scenarios. This paper firstly gives a thorough analysis on the reasons causing jams in the worst case, and then boosts up the performance by leveraging a novel mechanism named Convert-Search-incrementally-Verify (CSiV). CSiV differs from CSV in that it first merges possible case-sensitive matches to suspicious segments in the "Search" phase, and then leverages an Aho-Corasick like algorithm to verify them. The infeasibility of the simple Double Search (DS) approach is also explained by analyzing its low average-case throughput. Extensive experiments based on real pattern sets along with both collected and artificial traffic traces show that, the performance of the proposed approach outperforms the DS approach by a factor of 2 in the ordinary cases, and is better than the CSV approach up to 5 times under the worst-case scenario, indicating both its feasibility and robustness for a worst-case safe NIDS. © 2007 IEEE.