A DEPTH PRE-PROCESSING DATA ANALYSIS FOR INTRUSION DETECTION SYSTEM USING OUTLIER DETECTION AND BOX-COX TRANSFORMATION TECHNIQUE

Abstract

An Intrusion Detection System (IDS) seeks to identify unauthorized access to computer systems’ resources and datausing a statistical approach. The scale on which a dataset variable is measured may not the most appropriate forstatistical analysis or describing variation, and may even hide the basic characteristics of the data. This paper proposed apre-processing analysis for detecting unusual observations that do not seem to belong to the pattern of variabilityproduced by the other observations. The pre-processing analysis consists of outliers detection and Transformation.Outliers are best detected visually whenever this is possible. Usually, the original data sets are not normally distributed.If normality is not a viable assumption, one alternative is to make non-normal data look normal. This paper explains thesteps for detecting outliers’ data and describes the Box-Cox power transformation method that transforms them tonormality. The transformation obtained by maximizing lamda functions usually improves the approximation tonormality.

Keywords : IDS, dataset, outliers, transformation, pre-processing