Anomaly Detection in Cybersecurity Datasets via Cooperative Co-evolution-based Feature Selection

Anomaly detection from Big Cybersecurity Datasets is very important; however, this a challenging and computationally expensive task. Feature selection (FS) an approach to remove irrelevant redundant features select subset of features, which can improve the machine learning algorithms’ performance. In fact, FS effective preprocessing step anomaly techniques. This article’s main objective quantify accuracy scalability both supervised unsupervised effort, novel using FS, called Detection Using Selection (ADUFS), has been introduced. Experimental analysis was performed on five different benchmark cybersecurity datasets with without feature performance techniques were investigated. The experimental results indicate that instead original dataset, dataset reduced number yields better in terms true positive rate (TPR) false (FPR) than existing for detection. For example, technique, multilayer perception increased TPR by over 200% decreased FPR about 97% KDD99 dataset. Similarly, local outlier factor more 40% 15% 36% Windows 7 NSL-KDD datasets, respectively. addition, all require less computational time when suitable rather entire datasets. Furthermore, have compared six other state-of-the-art based decision tree (J48).