当前位置:网站首页>QCF for deep packet inspection论文总结
QCF for deep packet inspection论文总结
2022-07-26 14:43:00 【桐青冰蝶Kiyotaka】
QCF for deep packet inspection论文总结
Abstract

作者提出了一种新的成员查询数据结构变体,称为基于商的 Cuckoo 过滤器 (QCF),它反映了 QF 和 CF 之间的合并过程。
Introduction
- QCF use only two hash functions rather than three in the CF.
- Consuming the same memory requirements just such as CF.
- Employing only two hash functions rather than three in CF making it faster than BF, QF, and CF.
- Having same false-positive rate of CF which is smaller than the false-positive rate of BF and QF making it more accurate.
- Simpler to implement and consume less computational complexity than BF, QF, and CF.
- Supports delete operation.
- The ability to make dynamic signatures update and doing that quickly.
- Higher throughput than using BF, QF, and CF.
Related works

- 近似成员查询数据结构而不是使用有限自动机进行正则表达式匹配
- 实现简单、节省内存、对进行动态更新的快速响应、无需特定自动机
Deep PI
DPI 代表了 NIDS 通过检查网络流量内容来发现攻击和异常的主要任务。
区分良好 DPI 系统的主要两个特征:
- 执行强签名检测的能力(the ability to perform strong signature detection)
- 以高吞吐量和最小延迟生成结果的能力(produce the results with high throughput and minimum latency)
Probabilistic data structures
Bloom filter

false-positive rate f:
Quotient filter

详细介绍见这篇文章: Fast 2D filter with low false positive for network packet inspection论文总结
Cuckoo filter

Insertion process

Lookup process

Deletion process

Proposed QCF


- 第一个哈希函数中检索到的值被用作第一个候选桶的索引
- 第一个哈希函数检索到的值应用商法,以获得用作元素指纹的剩余部分
- 第二个哈希函数用于对元素的指纹进行哈希处理,获得第二个候选桶

Main steps of retrieving element’s fingerprint and the two candidate buckets in QCF:
Main steps of retrieving element’s fingerprint and the two candidate buckets in QCF and an example for QCF that has four buckets (three entries per bucket)
- 元素x将被发送到第一个哈希函数hash1,返回的值将用作第一个候选桶的索引
- 同时,对该返回值应用一个商技术,以从r最低有效位中检索X的指纹f
- 异或计算第二个候选存储桶的索引
- 指纹f将被插入到这两个候选桶之一
Insertion process of QCF

计算两个可能的候选桶的索引。接下来,根据计算的指纹f,检查这两个可能的候选桶,如果其中任何一个持有这个指纹,查找/删除
Query process of QCF

Deletion process of QCF

Proposed DPI system architecture
所有给定的(即预定义的)签名都被映射到所使用的 QCF 中,以便稍后用作监控传入流量的识别工具。
系统通过监控传入流量并根据已映射到 QCF 的签名检查每个到达的数据包来工作

Evaluation
Insertion throughput
QCF 比其他过滤器具有更高的插入吞吐量,这是减少计算开销的结果。
Query throughput
QCF 消耗的计算量更少,因此它比 CF、BF 和 QF 具有更高的查询吞吐量。
Deletion throughput
与 QF 相反,CF 和 QCF 提供稳定的性能,因为它们的删除是独立于过滤器占用的。
DPI time improvement
与使用 CF 相比,平均时间改进高达 77%,与使用 BF 和 QF 相比,平均时间改进高达 98%。
Conclusion

边栏推荐
- leetcode汇总
- 9、学习MySQL DELETE 语句
- Matlab solution of [analysis of variance]
- MySQL builds master-slave replication
- 晶品特装递交注册:第一季营收降80% 陈波控制68.5%股权
- [1.2. return and risk of investment]
- 31. Opinion based relational pivoting forcross domain aspect term extraction reading notes
- 写综述,想用一个靠谱的整理文献的软件,有推荐的吗?
- Crystal special decoration submitted for registration: the first quarter revenue fell by 80%, and Chen Bo controlled 68.5% of the equity
- How to get 5L water in a full 10L container, 7L or 4L empty container
猜你喜欢

Summary and analysis of image level weakly supervised image semantic segmentation

10 schemes to ensure interface data security

外文文献查找技巧方法有哪些

Classic line style login interface

Advanced MySQL v. InnoDB data storage structure

基于CAS的SSO单点客户端配置

目标跟踪相关知识总结
![[integer programming]](/img/e5/aebc5673903f932030120822e4331b.png)
[integer programming]

SiamRPN:建议区域网络与孪生网络

RPN: region proposal networks
随机推荐
写综述,想用一个靠谱的整理文献的软件,有推荐的吗?
Network pictures are transferred locally, causing the kernel to exit
Leetcode659. split the array into continuous subsequences (hash table)
自编码器 AE(AutoEncoder)程序
4 kinds of round head arrangement styles overlay styles
llcc68广播唤醒简单说明
Crystal special decoration submitted for registration: the first quarter revenue fell by 80%, and Chen Bo controlled 68.5% of the equity
Qt开发高级进阶:如何在显示时适合视窗宽度和高度(fitWidth+fitHeight)
14. Bridge-Based Active Domain Adaptation for Aspect Term Extraction 阅读笔记
postman 环境变量设置代码存放
大学生如何申请实用新型专利?
CAS单点登录
Minecraft 1.16.5模组开发(五十二) 修改原版生物战利品 (Loot Table)
VBA 上传图片
JS creative range select drag and drop plug-ins
Stacked noise reducing auto encoder (sdae)
外文文献查找技巧方法有哪些
基于CAS的SSO单点客户端配置
BSN IPFS(星际文件系统)专网简介、功能、架构及特性、接入说明
The difference between torch.nn and torch.nn.functional in pytorch