ChinaCom2008-Network and Information Security Symposium

Research Article

A Multiple Regular Expressions Matching Architecture for Network Intrusion Detection System

  • @INPROCEEDINGS{10.1109/CHINACOM.2008.4685118,
        author={Wei Zhang and Tian Song and Dongsheng Wang},
        title={A Multiple Regular Expressions Matching Architecture for Network Intrusion Detection System},
        proceedings={ChinaCom2008-Network and Information Security Symposium},
        publisher={IEEE},
        proceedings_a={CHINACOM2008-NIS},
        year={2008},
        month={11},
        keywords={pattern matching; regular expression matching; intrusion detection},
        doi={10.1109/CHINACOM.2008.4685118}
    }
    
  • Wei Zhang
    Tian Song
    Dongsheng Wang
    Year: 2008
    A Multiple Regular Expressions Matching Architecture for Network Intrusion Detection System
    CHINACOM2008-NIS
    IEEE
    DOI: 10.1109/CHINACOM.2008.4685118
Wei Zhang1,*, Tian Song2,*, Dongsheng Wang3,*
  • 1: Dept. of Computer Sci. & Tech. Tsinghua University Beijing, 100084, P.R.China
  • 2: School of Computer Sci. & Tech. Beijing Institute of Technology Beijing, 100081, P.R.China
  • 3: Microprocessor and SoC Tech. R&D Tsinghua University Beijing, 100084, P.R.China
*Contact email: zhwei02@mails.tsinghua.edu.cn, songtian@bit.edu.cn, wds@tsinghua.edu.cn

Abstract

Regular expressions are increasingly used in network security applications. Multiple regular expressions matching is one of the most important performance bottlenecks in those systems. This paper proposes a new hardware-based multiple regular-expressions matching architecture, called MRM, for network intrusion detection system. It shows that traditional algorithm, such as AC, has to face the serious spatial explosion problem when simultaneously detecting a large number of regular expressions because of constrained repetitions. MRM utilizes hardware RAM modules to share matching signals and exploits hardware register counting to implement constrained repetitions. This paper also proposes a software compiler to construct the hardware architecture and generate information in MRM’s RAMs for the given regular expressions. Experiments in actual snort and bro regular expression sets show that MRM can achieve the high throughput of 2.1Gbps and 2.8Gbps on Virtex2 and Virtex4 devices respectively.