Managing and Evaluating Data Integrity in machine Learning-Based Cyber Intrusion Detection

Abstract

This study investigated the impact of data quality on the performance of intrusion detection systems. Experiments were conducted using intrusion datasets and machine learning models. The pre-trained models were less affected by data duplications and overlaps compared to classic ML models. Removing overlaps and duplicates from training data improved the pre-trained models' performance in most cases, but had adverse effects in datasets with highly similar sequences. The study also proposed a framework for model selection and data quality assurance for building high- quality intrusion detection systems. Additionally, we focus on optimizing nine hyperparameters within a 1D-CNN model, using two well-established evolutionary computation methods—genetic algorithm(GA) and particle swarm optimization(PSO).The performances of these methods are assessed using three majordatasets—UNSW-NB15,CIC- IDS2017, and NSL-KDD. The key performance metrics considered in this study include the accuracy, loss, precision, recall, and F1-score. The results demonstrate considerable improvements in all metrics across all datasets, for both GA- and PSO-optimized models, when compared to those of the original non optimized 1D-CNN model.