Uncertainties classification in cyberspace using ensemble learning model

  • M.E. Irhebhude
  • Z.O. Musa
  • A.O. Kolawole
Keywords: UNSW-NB15, Network Intrusion Detection Systems, Classification, Ensembled Classifier


Studies have shown that different techniques can help classify uncertainties in the cyber space, however, a lot of these studies did not report the false predictions. Ensembled classifier was applied in this paper to curb the uncertainties in cyberspace. Classification learner in MATLAB was used as a tool to train the machine learning model on the publicly available University of New South Wales Network-Based (UNSW-NB15) and locally gathered datasets. A multiclass classification was done on the two datasets which consist of various attack categories. An experiment was performed with the proposed model on the datasets with the use of an ensemble classifier in MATLAB classification learner with 30% held for validation. Performances were measured using accuracy, confusion matrix, and receiver operating characteristics (ROC) curve. The experiments resulted in excellent classification accuracy of 99.1% and 99.4% on the merged Comma Separated Value (CSV) UNSW-NB15 dataset and self-acquired dataset respectively. Experimental results from the two datasets have shown that ensembled gave a more robust classification accuracy compared to artificial neural network classifier. With the results, the ensemble model help solved the problem of classification of attacks in network environment and uncertainties in cyberspace. Infrastructures in cyberspace and user interaction will be well secured with the adopted solution.


Journal Identifiers

eISSN: 1597-6343