Term of Award
Master of Science, Information Technology
Document Type and Release Option
Thesis (restricted to Georgia Southern)
Copyright Statement / License for Reuse
This work is licensed under a Creative Commons Attribution 4.0 License.
Department of Information Technology
Committee Member 1
Committee Member 2
The intrusion detection system (IDS) has been evolving as demand for safeguarding sensitive information on the network has grown since the term network was first introduced. As anomaly-based IDS being one of the security measures that could be used to detect intrusion, much research in the IDS field has developed an abundance of anomaly-based IDS with various deep learning algorithms such as LSTM and GRU. However, most of those research papers neglected the importance of using the newest dataset for their IDS. Instead, their methods used old intrusion datasets such as KDD-99 making their IDS unadaptable to modern attacks. Also, these research papers didn’t consider the feature of anomaly-based IDS creating excessive false alarms, such as false positive or true negative. To mitigate such problems, in our study, we implemented an IDS with two different deep learning algorithms, LSTM and GRU, with the newest dataset UNSW-NB15 while reducing the false alarm rate by using the algorithms we developed in a finite state machine.
Kim, Seongsoo, "Towards Effective and Efficient Intrusion Detection with Deep Learning" (2022). Electronic Theses and Dissertations. 2445.
Research Data and Supplementary Material