Date of Award
5-2023
Culminating Project Type
Thesis
Styleguide
apa
Degree Name
Information Assurance: M.S.
Department
Information Assurance and Information Systems
College
Herberger School of Business
First Advisor
Dr. Jim Q. Chen
Second Advisor
Dr. Shakour A. Abuzneid
Third Advisor
Dr. Akalanka B. Mailewa
Fourth Advisor
Dr. Abdullah Abu Hussein
Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.
Keywords and Subject Headings
Intrusion Detection Systems (IDS), Machine Learning, Deep Learning, Principal Component Analysis (PCA), Support Vector Machine (SVM), Autoencoder Neural Network (ANN), NSL-KDD Dataset
Abstract
Advancements in computing technology have created additional network attack surface, allowed the development of new attack types, and increased the impact caused by an attack. Researchers agree, current intrusion detection systems (IDSs) are not able to adapt to detect these new attack forms, so alternative IDS methods have been proposed. Among these methods are machine learning-based intrusion detection systems. This research explores the current relevant studies related to intrusion detection systems and machine learning models and proposes a new hybrid machine learning IDS model consisting of the Principal Component Analysis (PCA) and Support Vector Machine (SVM) learning algorithms. The NSL-KDD Dataset, benchmark dataset for IDSs, is used for comparing the models’ performance. The performance accuracy and false-positive rate of the hybrid model are compared to the results of the model’s individual algorithmic components to determine which components most impact attack prediction performance. The performance metrics of the hybrid model are also compared to two deep learning Autoencoder Neuro Network models and the results found that the complexity of the model does not add to the performance accuracy. The research showed that pre-processing and feature selection impact the predictive accuracy across models. Future research recommendations were to implement the proposed hybrid IDS model into a live network for testing and analysis, and to focus research into the pre-processing algorithms that improve performance accuracy, and lower false-positive rate. This research indicated that pre-processing and feature selection/feature extraction can increase model performance accuracy and decrease false-positive rate helping businesses to improve network security.
Recommended Citation
Collen, Samantha, "Comparing a Hybrid Multi-layered Machine Learning Intrusion Detection System to Single-layered and Deep Learning Models" (2023). Culminating Projects in Information Assurance. 136.
https://repository.stcloudstate.edu/msia_etds/136
Comments/Acknowledgements
Thank you to my Thesis Chair, Dr. Jim Chen for sharing your knowledge and guiding me through this process. Your encouragement was always appreciated. Thank you to Dr. Abu Hussein for helping me at the initial stage of finding my research interests and selecting a topic. Thank you to Dr. Akalanka Mailewa for providing me resources and helping me to narrow down my research topic to intrusion detection and machine learning. Thank you to Dr. Abuzneid for your encouragement and guidance with machine learning and python code. I am blessed to have had such an encouraging and knowledgeable committee to lead me through this journey. Thank you to everyone I did not mention that assisted me in this exploration. I could not have completed this process without your support, and I am truly grateful. A special thank you to my Academic Advisor, Dr. Lynn Collen for your guidance and direction, and to all my family for always being there for me.