AN APPLICATION OF LINEAR DISCRIMINANT ANALYSIS IN THE DESIGN OF INTRUSION DETECTION SYSTEM (LINEAR DISCRIMINANT ANALYSIS & INTRUSION DETECTION SYSTEM)

Background and objective: Intrusion Detection Systems essentially involve processing of high dimensional network data using various multivariate statistical methods. In this work the Intrusion detection systems designed for network features and non-network features are compared for their efficiency. Materials and Methods: This current work has used Linear Discriminant Method to analyze the network data in the design of Intrusion Detection System by applying it to the network attacks from UNSW-NB15 intrusion detection data along with MITRE ATT&CK framework. The Linear Discriminant Method is applied the trained data set to build the model and also to test data set to validate the model. Separate models were built for connection features and non-connection features. Results: The performance measures are deduced using Confusion Matrix for the models. The results indicate that network features are best suited to detect cyber intrusions as compared to the other network features. Conclusion: The network features of the computer system coupled with balanced training dataset and test data set shows improvement in the efficiency of the intrusion detection system.