The impact of identifiable features in ML Classification algorithms with the HIKARI-2021 Dataset

Network Intrusion Detection Systems keep being the critical part of cybersecurity that researchers are exploring the most. Both datasets and algorithms are being improved, with more computational power, available data, advanced analysis of the data, and better enhancements in the selected algorithms.This paper comes as a need for a new analysis of the HIKARI-2021 dataset where we revisit the features used in previous articles and we discuss the feature importance of the dataset and we notice a significant drop of 20% in the accuracy when removing identifiable features that should be used when training the models.