Pattern Matching is a computationally intensive task used in many research fields and real world applications. Due to the ever-growing volume of data to be processed, and increasing link speeds, the number of patterns to be matched has risen significantly. In this paper we explore the parallel capabilities of modern General Purpose Graphics Processing Units (GPGPU) applications for high speed pattern matching. A highly compressed failure-less Aho-Corasick algorithm is presented for Intrusion Detection Systems on off-the-shelf hardware. This approach maximises the bandwidth for data transfers between the host and the Graphics Processing Unit (GPU). Experiments are performed on multiple alphabet sizes, demonstrating the capabilities of the library to be used in different research fields, while sustaining an adequate throughput for intrusion detection systems or DNA sequencing. The work also explores the performance impact of adequate prefix matching for alphabet sizes and varying pattern numbers achieving speeds up to 8Gbps and low memory consumption for intrusion detection systems.
@article{arxiv.1704.02272,
title = {A Highly-Efficient Memory-Compression Scheme for GPU-Accelerated Intrusion Detection Systems},
author = {Xavier Bellekens and Christos Tachtatzis and Robert Atkinson and Craig Renfrew and Tony Kirkham},
journal= {arXiv preprint arXiv:1704.02272},
year = {2017}
}
Comments
Published in The 7th International Conference of Security of Information and Networks, SIN 2014, Glasgow, UK, September, 2014