A sub-linear time string matching algorithm with Bloom filters acceleration: Design, Implementation and Evaluation

• Main Authors: Yi-Jun Zheng, 鄭伊君
• Other Authors: Yin-Dar Lin
• Format: Others
• Language: zh-TW
• Published: 2006
• Online Access: http://ndltd.ncl.edu.tw/handle/25684296002946637638

碩士 === 國立交通大學 === 資訊科學與工程研究所 === 94 === Many network security applications heavily rely on string matching to detect malicious intrusions, viruses, spam, and so on. A software-based implementation may not meet the performance requirement of high-speed applications due to intensive computation and frequent memory accesses. A hardware solution to take advantage of hardware parallelism is a promising trend to inspect the packet payload at line rate. In this work, we propose an innovative memory-based architecture using Bloom filters to realize a sub-linear time algorithm that can effectively process multiple characters simultaneously. The two key ideas to realize the sub-linear time algorithm in this architecture are (1) replacing the slow table lookup in the external memory with simultaneous queries to several Bloom filters and (2) designing a non-blocking verification interface to keep the worst-case performance in linear time. The proposed architecture is verified in both behavior simulation in C and timing simulation in HDL. The simulation result shows that the throughput is nearly 10 Gbps for Windows executable files and 600 Mbps in the worst case.