towards effectiveense mbl e classification for a nomaly based intrusion detection

Abstract

As information technology rolls out, the applications of the newlineInternet continue to impact our daily routines including communication, ecommerce, newlineentertainment, e-learning, etc. The advent of computing and newlinecommunicating devices as well as the infiltration of intrusive actions and newlinehacking tools into the networks make data communication increasingly newlinevulnerable. Generally, an intrusion would cause a loss of confidentiality, newlineintegrity, and availability (CIA triad) of information. An Intrusion Detection newlineSystem (IDS) is widely employed to detect cyberattacks preferably in realtime newlineand to protect the valuable information of the users. Albeit, numerous newlineMachine Learning (ML) algorithms have been proposed to improve the newlineperformance of IDS, it is a challenge to process massive unrelated and newlineredundant information in current big data environments. newlineThis research proposes an Intelligent Classifier using Ensemble newlineTechnique (ICET) to classify the intrusive activities significantly. The newlineproposed ICET includes two elements: (i) feature selection module; and newline(ii) ensemble classifier. To cope with high dimensional traffic in large newlinenetworks, the feature selection module exploits a Correlation-based Feature newlineSelection (CFS) algorithm to select the appropriate features. Besides, it newlineexploits the optimized RelieF algorithm to calculate the quality of attributes. newlineThe attributes with a low-quality index are eliminated to reduce the newlinedimensionality of the feature space. The performance of the proposed feature newlineselection approach is further enhanced by integrating CFS with Bat-inspired newlineOptimization (BIO) algorithm. This integration (hereafter called BIOCFS) is newlineembedded in an ensemble classifier to increase the performance of the IDS. newlineThis study proposes an ensemble classifier that includes three newlinedifferent classifiers including Balanced Forest (BF), Random Forest (RF), and newlineC4.5 decision tree. The BF exploits the Forest by Penalizing Attributes (FPA) newlinealgorithm to construct a set of highly balanced and accurate decision trees. newlineThe RF classifier integ