Review of Network Intrusion Systems Evaluated on NSL-KDD and CIC-IDS2017 Datasets

The increase in the role of information technology in day-to-day tasks inevitably calls for heightened security measures so as to protect sensitive data from falling into the wrong hands. The evolution of computer systems has incentivized the spawning of even more advanced and dangerous types of cyber-attacks, making it more of a challenge for security systems to identify them in an efficient and accurate manner. Network Intrusion Detection Systems (NIDSs), mostly operating on anomaly-based detection schemes, must comprise machine learning frameworks which are robust enough to effectively detect most network attack groups. Recent studies which focus on building efficient NIDSs present an amalgamation of techniques, from the stage of data pre-processing, feature selection to classification, each having its own strengths and limitations. This paper reviews several existing NIDS models which have been evaluated on benchmark datasets NSL-KDD and CIC-IDS2017. For both datasets, genetic algorithm-implemented models outperformed all other models across most performance metrics.