Proposed Hybrid CorrelationFeatureSelectionForestPanalizedAttribute Approach to advance IDSs

摘要

NetworkIntrusionDetectionSystem(NIDS), widely used network infrastructure. Although many datamining has been used to increase the effectiveness of IDSs, current ID still struggle to perform well. therfore; proposed a new NIDS focused on feature_selection. The proposed CorrelationFeatureSelection_ForestPanalizedAttributes(CFS_FPA) used for dimensionality_reduction and selects the optimal_subset. based on two steps: first check each feature with a target(class) and choose only features that most effective by applying CFS filter using a statistical_method, then applied FPA to select only features will enhance ID and reduce_dimensionality. proposal tested with the NSLKDD experimental results of accuracy 0.997% and 0.004 FAR, wherein UNSWNB15_dataset accuracy and FAR are 0.995%, 0.008 consequently.