Fusion of Feature Ranking Methods for an Effective Intrusion Detection System

Expanding internet-connected services has increased cyberattacks, many of which have grave and disastrous repercussions. An Intrusion Detection System (IDS) plays an essential role in network security since it helps to protect the from vulnerabilities attacks. Although extensive research was reported IDS, detecting novel intrusions with optimal features reducing false alarm rates are still challenging. Therefore, we developed a fusion-based feature importance method reduce high dimensional space, identify attacks accurately less rate. Initially, improve training data quality, various preprocessing techniques utilized. The Adaptive Synthetic oversampling technique generates synthetic samples for minority classes. In proposed importance, use different approaches filter, wrapper, embedded methods like mutual information, random forest permutation Shapley Additive exPlanations (SHAP)-based statistical difference mean median standard deviation rank each according its rank. Then by simple plurality voting, most retrieved. fed models Extra Tree (ET), Logistic Regression (LR), Support vector Machine (SVM), Decision (DT), Extreme Gradient Boosting (XGBM). hyperparameters classification tuned Halving Random Search cross-validation enhance performance. experiments were carried out on original imbalanced balanced data. outcomes demonstrate that scenario knocked Finally, experimental analysis proved our performed well XGBM giving accuracy 99.86%, 99.68%, 92.4%, 9, 7 8 time 1.5, 4.5 5.5 s Network Security Laboratory-Knowledge Discovery Databases (NSL-KDD), Canadian Institute Cybersecurity (CIC-IDS 2017), UNSW-NB15, datasets respectively. addition, suggested been examined contrasted state art three datasets.