Network Intrusion Detection Method Combining CNN and BiLSTM in Cloud Computing Environment

Abstract

A network intrusion detection method combining CNN and BiLSTM network is proposed. First, the KDD CUP 99 data set is preprocessed by using data extraction algorithm. The data set is transformed into image data set by data cleaning, data extraction, and data mapping; Second, CNN is used to extract the parallel local features of attribute information, and BiLSTM is used to extract the features of long-distance-dependent information, so as to fully consider the influence between the front and back attribute information, and attention mechanism is introduced to improve the classification accuracy. Finally, C5.0 decision tree and CNN BiLSTM deep learning model are combined to skip the design feature selection and directly use deep learning model to learn the representational features of high-dimensional data. Experimental results show that, compared with the methods based on AE-AlexNet and SGM-CNN, the network intrusion detection effect of this method is better, the average accuracy can be improved to 95.50%, the false-positive rate can be reduced to 4.24%, and the false positive rate can be reduced to 6.66%. The proposed method can significantly improve the performance of network intrusion detection system.