中文摘要：

传统的深度包检测算法通常存在频率带宽瓶颈、不能精确匹配、不切实际的存储要求等其中之一或数个缺点.本文基于哈希与Bloom Filter提出一种新型精确匹配结构：Bloom Filter分类器,首先基于哈希对特征串分组,再用多组Bloom Filter对输入串分类,在每长度定位到唯一可能的匹配串并对比验证.对Snort、ClamAV集合进行了存储实验评估,以约1.22（字节/字符）的低存储代价实现对万条字符串集的精确匹配.该结构具有精确匹配、多字节匹配扩展简单、不存在带宽瓶颈等优点.

英文摘要：

Traditional Deep Packet Inspection algorithm suffers from throughput bottleneck,fuzzy matching or unfeasible requirement of hardware memory.This paper proposes a novel exact string matching scheme called Bloom Filter Classifier,which uses hash and multiple bloom filters to classify the input string and locates the only probable signature on each length,and then conducts verification.Evaluation with Snort and ClamAV string sets shows that BFC costs only 1.22 Byte/Character memory and provides exact matching result without throughput bottleneck.