Ensemble-Based Filter Feature Selection Technique for Building Flow-Based IDS

Intrusion Detection systems play a crucial role in maintaining network security. It keeps track of network traffic for anomalous activities and detects any vulnerabilities in the network. It is not a trivial task to build one due to the high number of features in the dataset, which increases the computational overhead on the system. It is necessary that we select only the relevant features from the dataset to ensure that the model thus built provides high accuracy in low computational time. This paper works on different filter-based feature selection techniques to lower the complexity of intrusion detection systems while preserving the performance of the system. The use of feature selection techniques followed by ensemble learning provides an optimal subset of features. The proposed method attempts to handle the imbalance of classes in CIC-IDS2017 and NSL-KDD datasets by separately classifying the minority and majority classes. The model's performance is explored in terms of precision, accuracy, and F1 score, that has been observed to be superior to existing works in the field of intrusion detection.