Classification of Intrusion Detection System (IDS) Based on Computer Network

Intrusion Detection System (IDS) is made as one of the solutions to handle security issues on the network in order to remain assured free of attack. IDS's work is developed by 2 models that using signature-based detection, how it works is limited to the pattern of attack behavior that has been defined in the database. The next is the Anomaly-based IDS model. It works by detects unusual activity of network in the normal conditions, but this model gives a lot of false positiv messages. Several previous studies have shown that the IDS approach with machine learning techniques can provide high accuracy results. The first step that must be done in the application of mechine learning technique is preprocessing the selection of features / attributes to optimize the performance of learning algorithms. In this study, intrusion detection system with mechine learning classification technique is proposed by using naivebayes algorithm with NSL-KDD dataset. The processes in this reseach start from normalization of data, discretization features continuous variables with k-means method and the selection of features using Information Gain algorithm. The result of this reseach shows that the application of k-means clustering method for continuous variabe discretization and feature selection can optimize the performance of naivebayes algorithm in classifying intrusion types.