Int J Performability Eng ›› 2020, Vol. 16 ›› Issue (12): 1875-1887.doi: 10.23940/ijpe.20.12.p4.18751887

• Orginal Article • Previous Articles     Next Articles

Network Security Situation Prediction based on Combining 3D-CNNs and Bi-GRUs

Jie Lin*, Minghua Wei   

  1. Department of Information and Technology, Fuzhou Polytechnic, Fuzhou, 350108, China
  • Submitted on ; Revised on ; Accepted on
  • Contact: * Corresponding author. E-mail address: linjielw2020@163.com
  • Supported by:
    This work was supported by the Fuzhou Polytechnic Research Foundation (No. FZYKJJJJC202001).

Abstract: Conventional neural networks-based network security situation (NSS) prediction methods have low prediction accuracy and low efficiency. To solve such shortages of NSS prediction, this paper proposes a novel method based on combining 3D convolutional neural networks (3DCNNs) and bidirectional recurrent neural networks (Bi-RNNs). Because the prediction data of NSS includes multi-dimensional time series, the NSS can be better predicted by combining the spatial features and sequential features, and the prediction accuracy and efficiency will be improved by using the combined features. Therefore, the 3DCNNs model is adopted to extract spatial features from different network nodes, and the Bi-RNNs model with gated recurrent units (Bi-GRUs) is adopted to extract the sequential features based on the extracted spatial features. Finally, the NSS prediction results are obtained by using the fused spatial-sequential features. In order to validate the feasibility and effectiveness of the proposed method, comparable experiments are performed on three different datasets. Experimental results have shown that the proposed 3DCNNs-Bi-GRUs model achieves the optimal NSS prediction results among all datasets under different situations. The efficiency of the proposed model meets the requirements for real-world application scenarios. By combining the spatial features and sequential features, the proposed model confirms higher prediction accuracies of NSS, and such a model has good application value for the rapid development of computer networks and intelligent technologies.

Key words: network security situation prediction, 3D convolutional neural networks, recurrent neural networks, bidirectional gated recurrent units, combination features