Genetic Algorithm Based Anomaly Detection for Intrusion Detection

This article describes a method for detecting anomalies in network intrusion data. The method utilizes a clustering-based strategy to efficiently cluster the data. Here, Genetic Algorithm (GA)-based clustering was employed, with chromosomes represented by strings of real integers encoding the cluster centres. We have adopted the Davies-Bouldin (DB) index as a fitness function in this study. After clusters have been constructed, points that are closer to the cluster centroids are deemed inliers and are eliminated. Pruning strategies considerably minimize the computing effort and complexity required to identify outlier scores. To generate outlier scores, we've used a technique based on local distance and the outlier factor. The n points with the highest value were deemed to be outliers. Our results were compared to those obtained utilizing the explicit local distance-based outlier factor algorithm.