摘要
针对网络入侵检测准确率低的问题,笔者提出一个基于卷积神经网络和长短期记忆网络的深度分层网络模型,并在UNSW-NB15数据集和CICIDS2017数据集上进行二分类实验。首先,基于卷积神经网络和长短期记忆网络提取局部特征和长距离依赖特征;其次,引入自注意力机制计算各属性特征的重要性;最后,使用sigmoid分类器对特征数据进行二分类。实验结果表明,与其他检测模型相比,该模型的网络入侵检测效果较好。
Aiming at the problem of low accuracy of network intrusion detection, the author proposes a deep layered network model based on convolutional neural network and short-term memory network, and carries out binary classification experiments on UNSW-NB15 data set and CICIDS2017 data set. Firstly, local features and long-distance dependent features are extracted based on convolutional neural network and short-term memory network;Secondly, the self attention mechanism is introduced to calculate the importance of each attribute feature;Finally, sigmoid classifier is used to classify the feature data. The experimental results show that compared with other detection models, the network intrusion detection effect of this model is better.
作者
贺飞飞
王恒
HE Feifei;WANG Heng(School of Information Engineering,Ningxia University,Yinchuan Ningxia 750000,China)
出处
《信息与电脑》
2021年第23期55-57,共3页
Information & Computer
基金
结合注意力机制基于深度学习的网络入侵检测研究(项目编号:2021AAC03114)。
关键词
入侵检测
卷积神经网络
长短记忆网络
注意力机制
intrusion detection
convolutional neural network
short and long memory network
attention mechanism
