Intrusion Detection System using Ensemble Learning Analytics

An Intrusion Detection System (IDS) monitors and analyses data to find any intrusions into a system or network. The network generates data at a tremendous volume, variety, and speed, making it difficult to detect attacks using conventional techniques like a virus detection system, misuse detection software i.e. the database of attack signatures that it uses to compare packets. Despite the researchers' significant efforts, IDS still struggles to identify new intrusions, to improve detection accuracy, and to reduce false alarm rates. To overcome the problems mentioned above this paper proposes an unique model named Intrusion Detection System using Machine Learning Analytics (IDSMLA), which uses SMOTE oversampling technique to deal with class imbalance problem, it also uses Minimum Redundancy Maximum Relevance (mRMR) to perform feature selection as feature selection reduces time complexity by eliminating irrelevant features and hence increasing the accuracy of the model and finally to perform classification task, the proposed model IDSMLA uses Extra Trees(ET) bagging ensemble technique. The performance of the proposed model IDSMLA is measured using accuracy and F1-score using 10-folds cross validation. Experimental results have demonstrated that the proposed model IDSMLA greatly outperforms different single-classifier based models, different ensemble models as well as different models present in literature.