The Metering Automation System based Intrusion Detection Using Random Forest Classifier with SMOTE+ENN

The intrusion detection system (IDS) plays an important role in network security. The KDDCup99 dataset is the benchmark dataset. There is a huge imbalance between the classes in the KDDCup99 dataset. In this paper, we propose a model framework that combines Synthetic Minority Oversampling Technique (SMOTE) and the Edited Nearest Neighbor Rule (ENN) using Random Forest Classifier. In terms of dealing with irrelevant and redundant features in the dataset which tends to increase time complexity and resource consumption as well as decrease detection rates, we combine ranks obtained from both information gain and correlation. Besides, 25 features were selected. The experimental results obtained showed that the SMOTE+ENN method can achieve higher precision, recall and F1-value.