基于改进的CNN-BiLSTM和三支决策的网络入侵检测方法
Network intrusion detection based on improved CNN-BiLSTM and three-way decision
王宇1
作者信息
- 1. 安徽理工大学计算机科学与工程学院,淮南 232001
- 折叠
摘要
针对网络入侵检测模型特征提取不足以及相关数据集不平衡问题,提出一种基于卷积神经网络与双向长短期记忆网络的网络入侵检测方法.首先结合SMOTE过采样算法对UNSW-NB15数据集进行预处理;其次建立基于CNN-BiLSTM的入侵检测模型,提取数据集的局部特征和长距离依赖特征,通过注意力机制进一步加强特征的重要性;最后通过基于三支决策的分类器获得分类结果.实验结果表明,所提方法在各个评价指标上均有所提升,能够有效提高检测性能.
Abstract
Aiming at the problems of insufficient feature extraction of network intrusion detection model and imbalance of re-lated dataset,a network intrusion detection method based on convolutional neural network and bidirectional long short-term memory network is proposed.Firstly,the UNSW-NB15 dataset is preprocessed by combining the SMOTE oversampling algorithm.Secondly,the intrusion detection model based on CNN-BiLSTM is established to extract the local features and long-distance-dependent features of the dataset,and the importance of the features is further strengthened by the attention mechanism.Finally,the classifica-tion results are obtained by the classifier based on the three-branch decision.The experimental results show that the proposed method is improved in each evaluation index and can effectively improve the detection performance.
关键词
入侵检测/卷积神经网络/双向长短期记忆网络/三支决策Key words
intrusion detection/convolutional neural networks/bidirectional long short-term memory networks/three-way decision引用本文复制引用
出版年
2024
NETL
NSTL
万方数据