Research on Intrusion Detection Technology Based on CNN-SaLSTM

Abstract

As Internet-connected application devices become more and more popular, more and more services need to be done through the network, which also leads to users paying more attention to network security performance. Due to the continuous iterative development of cyber attack means and attack scale, it is difficult to conduct passive security detection systems such as traditional intrusion detection mechanisms to conduct endless attacks. Later, intrusion detection was studied as an active defense technique to compensate for the shortcomings of traditional safety detection techniques. Active defense and response technology has also attracted the attention of researchers at home and abroad. The complex, engineering and large-scale scenarios presented by network attacks prevent the original passive intrusion detection system to meet the users’ needs for network security performance. With the continuous expansion of network scale, the continuous increase of network traffic scenarios and the rapid iteration of attack means, the performance of network intrusion detection system has put higher requirements. Therefore, we introduced the CNN, LSTM and self attention mechanisms in deep learning into invasion detection and performed experiments in the tensorflow framework, increasing the accuracy to 97.4%.