Convolutional One-Dimensional Variational Autoencoder Based Intrusion detection

Abstract

Intrusion detection technology is an important part of network security. It detects various intrusions by collecting and analyzing various information on the network, which is the focus of maintaining network security. With the popularization of the network and the increase of the network rate, the network attack behavior is increasing and the attack methods are constantly updated. The traditional detection technology can not meet the demand of the new network environment for network security. In view of the defects of the existing intrusion detection algorithm model based on deep learning, such as long training time, more hyperparameters and high data demand, it is feasible to detect malicious traffic by learning the characteristics of normal traffic. In this paper, a variational self-encoder (COD-VAE) intrusion detection model based on one-dimensional convolutional neural network is proposed, which can identify malicious traffic more accurately. The intrusion detection algorithm is used to detect the intrusion of the intrusion detection system. Experimental results show that all the indexes of the algorithm are higher than the comparison algorithm, which verifies the effectiveness of the algorithm.