Intrusion Detection Model of CNN-BiLSTM Algorithm Based on Mean Control

Abstract

The data preprocessing methods of traditional intrusion detection algorithms have some problems: difficulty in training, low classification accuracy, and poor generalization ability for some datasets with the unbalanced numerical distribution. To improve these problems, this paper proposes a data preprocessing method based on the mean control, Convolutional Neural Network (CNN), and Bidirectional Long Short-Term Memory Network (BiLSTM) algorithm, through experiments with different means to choose the optimal control of mean. Using the optimal mean to standardize the data of mean control can reduce the unbalance of data distribution and improve the detection accuracy. In this paper, NSL-KDD datasets are used for model training and testing. The experimental results show that the data preprocessing method's overall accuracy based on mean control can reach 99.10% on the CNN-BiLSTM algorithm, which has a good effect compared with the results without mean control.