Application of Convolutional Neural Network in Intrusion Detection

Abstract

Advanced persistent threat (APT) attack is an advanced network intrusion technology. Because of its strong concealment, pertinence, persistence and long-term intrusion mechanism regardless of cost, the security of all kinds of core networks is greatly threatened. This paper proposes an intrusion detection model based on hybrid convolutional neural network. Compared with the traditional machine learning model, the hybrid deep learning network structure can mine more complex structure features of the whole network traffic matrix, and extract and encapsulate the unknown malicious behavior features. Firstly, the correlation between different feature spaces in network traffic matrix is extracted by Convolutional Neural Network (CNN). Then, Recurrent Neural Network (RNN) is used to find out the time dependence of intrusion traffic data, fully mine the temporal and spatial characteristics of the whole network traffic matrix, and improve the accuracy of intrusion detection model.