摘要
针对目前网络上未知二进制协议种类繁多、不便于管理的问题,提出一种基于一维CNN的二进制协议分类方法,利用聚类得到协议数据的标签进行训练,直接将经过一类分类的二进制协议报文作为一维卷积神经网络的输入,训练分类模型,构建了一个二进制协议分类器,能够自动学习原始输入与预期输出之间的非线性关系,实现协议的自动分类功能。这是首次将一类分类与CNN网络应用于二进制协议分类领域。并且针对最大频度池化和一维卷积网络作了对比试验,验证了改进的有效性。经过实验验证对协议的识别率达到了98%以上,分类时间优于聚类方法。
In order to solve the problem that there are many kinds of unknown binary protocols on the network,which are not easy to manage,a binary protocol classification method based on one-dimensional CNN is proposed,which is trained by the tags of the protocol data obtained by clustering.The binary protocol message is directly used as the input of one-dimensional convolution neural network,and the classification model is trained to construct binary protocol classifier,which can automatically learn the nonlinear relationship between the original input and the expected output to realize the automatic classification function of the protocol.As far as we know,this is the first time that a class of classification and CNN networks have been applied to the field of binary protocol classification.The contrast experimental results between maximum frequency pooting and one-dimensional convolution show that the recognition rate of the protocol is up to 98%,and the classification time is better than that of the clustering method.The results show that the method is effective.
作者
尹世庄
王韬
陈庆超
刘丽君
YIN Shi-zhuang;WANG Tao;CHEN Qing-chao;LIU Li-jun(Shijiazhuang Campus of Army Engineering University,Shijiazhuang 050003,China)
出处
《火力与指挥控制》
CSCD
北大核心
2020年第11期163-167,172,共6页
Fire Control & Command Control
基金
国家重点研发计划基金(2018YFC0806900)
江苏省自然科学基金资助项目(BK20161469)。
关键词
深度学习
分类
二进制协议
一维卷积神经网络
deep learning
classification
binary protocol
one-dimensional convolution neural network
