Title: Network Intrusion Detection Using Deep Learning and Machine Learning for Multinomial Classification

Issue Number: Vol. 9, No. 4
Year of Publication: Dec - 2020
Page Numbers: 155-181
Authors: Thomas A. Woolman, Sanghyun “Philip” Lee
Journal Name: International Journal of Cyber-Security and Digital Forensics (IJCSDF)
- Hong Kong

Abstract:


The paper utilizes several cutting-edge machine learning and artificial intelligence technologies for data mining IP network traffic data in order to classify network intrusion anomalies. A novel approach using a deep learning algorithm (multi-layer perceptron neural network) is used in comparison with a number of machine learning classifiers, including the use of oversampling techniques for balancing training data classifiers. The results of these experiments are investigated to compare the natural dispersion between variances of multiple AI/ML algorithms using Levene’s test for equality of variances for null hypothesis testing. The paper concludes with a rejection of the null hypothesis and experimentally determines an optimal deep learning classification methodology for accurately predicting the multiclass dependent variable factor associated with IP network intrusion anomalies in the dataset.