We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
基于 CNN 与 LightGBM 的入侵检测研究.
- Authors
魏明军; 闫旭文; 纪占林; 陈 钊
- Abstract
Aiming at the problem that the minority class samples of network intrusion detection not only have low accuracy rate and recall rate, but also affect the overall accuracy rate and recall rate, a new combined algorithm based on generative adversarial network ( GAN), convolutional neural network and LightGBM was proposed. Firstly, the GAN was improved using a denoising variational autoencoder to handle the unbalance datasets. Secondly, the residual convolutional neural network was improved using the convolutional block attention mechanism to extract the key features of the data better. Finally, the processed dataset was classified using the LightGBM ensemble learning algorithm. The experimental results showed that after improvement,the accuracy rate,the recall rate,the precision rate,and the F1 score were all increased on the NSL-KDD test set. The model alleviated the negative impact of minority class samples on model classification and had good classification performance.
- Publication
Journal of Zhengzhou University (Natural Science Edition), 2023, Vol 55, Issue 6, p35
- ISSN
1671-6841
- Publication type
Article
- DOI
10.13705/j.issn.1671-6841.2022190