A Novel Approach to Reduce False-Negative Alarm Rate in Network-Based Intrusion Detection System Using Linear Discriminant Analysis

In recent years, Security is remaining as a major concern in various state-of-the-art computers and network systems. Expanding technologies and their relevant framework can also be the focal point, to exploit it. To prevent those frameworks, network systems, servers, files, and systems, enterprises use the intrusion detection system [IDS] and it tries to avoid certain traffic otherwise, it can lead to an intrusion in the system. Nowadays, Machine Learning and Deep Learning techniques used to keep an efficient IDS in almost every area. In this paper, We used a supervised classification algorithm, onto a Network-based IDS dataset—“UNSW-NB15 dataset”, and compared it with different algorithms to increase the efficiency of the IDS dataset which can further reduce the false negative alarm rate. We had applied a feature selection algorithm to sort out favorable features that can reduce the false alarm and then use Linear Discriminant Analysis as a classification algorithm that will refine our result analysis. The result will be based on Accuracy, False-Negative alarm rate, and ROC-AUC Score.