Network intrusion detection method based on deep learning

Abstract

The invention discloses a network intrusion detection method based on deep learning, and the training process of the method comprises the steps: inputting an obtained data set into a to-be-trained convolutional neural network model, and extracting network traffic features through the to-be-trained convolutional neural network model; repeatedly extracting features with richer network traffic through a spatial pyramid model including a convolutional layer, an upper sampling layer and a lower sampling layer, and obtaining a multi-scale effective feature layer; and finally, predicting network intrusion classification confidence through logistic regression, predicting categories by using a logic classification model, and calculating loss errors of a real frame and a prediction frame through an error model. Then, repeated iterative optimization is carried out through reverse gradient, and the to-be-trained network intrusion detection model with the minimum loss error is used as a trained network intrusion detection model; according to the method, the detection precision and speed of network intrusion are further improved, the detection capability of unknown attacks is improved, and the false alarm rate is reduced.

Description

technical field [0001] The invention relates to the field of network technology, in particular to a network intrusion detection method based on deep learning. Background technique [0002] With the increasing development of science and technology, the network has become one of the tools widely used by people. When people enjoy convenient Internet services, Internet security issues have gradually attracted widespread attention. Especially in the era of widespread use of social networking, e-commerce, and big data, problems such as network intrusion and data theft are also emerging, so network security has become an issue that cannot be ignored. [0003] Although the existing network intrusion detection technology has improved the accuracy, when faced with high-dimensional data sets such as KDD99, due to the influence of data dimensions, the detection rate of attack categories is often low when training models, which affects overall detection rate. When traditional machine l...

AI Technical Summary

Problems solved by technology

[0003] Although the existing network intrusion detection technology has improved the accuracy, when faced with high-dimensional data sets such as KDD99, due to the influence of data dimensions, the detection rate of attack categories is often low when training models, which affects overall detection rate

When traditional machine learning intrusion detection technology detects high-dimensional data, it often needs to artificially select some features based on a large amount of experience and professional knowledge. This process of feature selection often directly affects the final detection result, and the efficiency is low.

Traditional machine learning intrusion detection often uses a single classifier when predicting network attack categories, which leads to low accuracy and high false alarm rates when the attack type has multiple labels

Images