摘要
传统的网络安全入侵检测系统在面对复杂多变的网络攻击时,往往存在误报率高及准确率低的问题。为了改善这种情况,结合循环神经网络(RNN)和卷积神经网络(CNN)的优点,并将长短期记忆(LSTM)网络模型进行改进,设计一种面向网络安全入侵检测的双向长短期记忆(Bi-LSTM)算法。经过实验验证,该算法在网络安全入侵检测中取得了显著的效果。在5种不同类型的攻击方式下,CNN-BiLSTM模型的准确率分别为94.8%、90.2%、96%、90.5%和93.7%,误报率分别为5.98%、7.2%、5.23%、6.84%和6.49%。这些数据表明,设计的Bi-LSTM网络在安全入侵检测中具有较高的准确率和较低的误报率,具有一定的应用价值和研究价值。
Traditional network security intrusion detection systems often have high false alarm rate and low accuracy when facing complex and ever-changing network attacks.In order to improve this situation,this study combines the advantages of recurrent neural network(RNN)and convolutional neural network(CNN),and improves the long short-term memory(LSTM)model to design a bidirectional long short-term memory(Bi-LSTM)algorithm for network security intrusion detection.After experimental verification,this algorithm achieves significant results in network security intrusion detection.Under five different types of attack methods,the accuracy of the CNN-Bi-LSTM model reaches 94.8%,90.2%,96%,90.5%and 93.7%,respectively,and the false alarm rates are 5.98%,7.2%,5.23%,6.84%and 6.49%,respectively.These data indicate that the designed Bi-LSTM has high accuracy and low false alarm rate in network security intrusion detection,so it has certain application value and research value.
作者
于继江
YU Jijiang(National Institutes for Food and Drug Control,Beijing 102629,China)
出处
《微型电脑应用》
2024年第11期222-225,共4页
Microcomputer Applications
关键词
深度学习
卷积神经网络
入侵检测
循环神经网络
长短期记忆网络
deep learning
convolutional neural network
intrusion detection
recurrent neural network
long short-term memory network