A Comparative Study of Current Dataset Used to Evaluate Intrusion Detection System

Abstract

Today, since the world increasingly depends on computers and automation, securing applications and systems has become critical. The possibility of threat and violence is expanding rapidly due to the growing complexity of modern networks and services. This study focuses on current datasets, the gaps and limitations of these datasets, and their effects on creating Intrusion Detection Systems (IDS). The study outlines the existing datasets that show an apparent absence of real-world threats, attack representation, and outdated threats, limiting the efficiency of machine learning IDS techniques. This work presents the most Machine Learning techniques for implementing IDS and a comprehensive review of the available benchmark datasets. We believe this review will enable future researchers in this field to understand better different kinds of IDS datasets, advantages, and limitations.