Deep Learning Algorithm Based Computer Network Security Analysis Modeling

Abstract

There is a non face-to-face intellectual game between the defender and the attacker of Cyberspace Security. The attacker in the game has various motives and uses various techniques and procedures, while the defender in the game is often relatively passive. One way to change the rules of the game is to use human-computer intelligence, rather than hard coded control mechanism to resist all kinds of attacks. Especially against complex attacks, the important embodiment of this intelligence is machine learning. Based on the comprehensive analysis of intrusion detection and deep learning, this paper proposes a hybrid intrusion detection model based on deep structure. Firstly, the model adopts multi-layer depth structure for unsupervised feature learning. The high-dimensional and nonlinear intrusion data are mapped to the low dimensional space, and the relationship mapping between high-dimensional and low-dimensional is established. Then, fine-tuning algorithm is used to transform the model to achieve the best expression of features. Finally, classification method is used to identify the intrusion data. The experimental results show that the deep structure network model has better feature learning effect than the shallow model.