CSK-CNN: Network Intrusion Detection Model Based on Two-Layer Convolution Neural Network for Handling Imbalanced Dataset

In computer networks, Network Intrusion Detection System (NIDS) plays a very important role in identifying intrusion behaviors. NIDS can identify abnormal behaviors by analyzing network traffic. However, the performance of classifier is not good traffic for minority classes. order to improve detection rate on class imbalanced dataset, we propose model based two-layer CNN and Cluster-SMOTE + K-means algorithm (CSK-CNN) process dataset. CSK combines cluster Synthetic Minority Over Sampling Technique (Cluster-SMOTE) under sampling algorithm. Through network, only be identified, but also classified into specific attack types. This paper has been verified UNSW-NB15 dataset CICIDS2017 proposed evaluated using such indicators as accuracy, recall, precision, F1-score, ROC curve, AUC value, training time testing time. The experiment shows that CSK-CNN this obviously superior other comparison algorithms terms performance, suitable deployment real environment.