Abstract

This research aims to enhance the accuracy and speed of the intrusion detection process by using the feature selection method to reduce the feature space dimensions that eliminate irrelevant features. Further, we employed ensemble learning in the UNSW-NB15 dataset, by using a classifier of the Stacking method, to prevent the intrusion detection system (IDS) from becoming archaic, to adjust it with a modern attack resistance feature, and to make it less costly. We used logistic regression as a meta-classifier and combined random forests, sequential minimal optimization (SMO), and naïve Bayes methods. Our approach allowed us to achieve 97.88% accuracy in intrusion detection.

摘要

本研究旨在通过使用特征选择方法减少消除不相关特征的特征空间维度，从而提高入侵检测过程的准确性和速度。此外，我们在 UNSW-NB15 数据集中采用了集成学习，通过使用 Stacking 方法的分类器，以防止入侵检测系统（IDS）变得陈旧，并使用现代抗攻击特性对其进行调整，并使其更少昂贵。我们使用逻辑回归作为元分类器，并结合了随机森林、顺序最小优化 (SMO) 和朴素贝叶斯方法。我们的方法使我们能够在入侵检测中达到 97.88% 的准确率。