Date of Award
Doctor of Philosophy
Dr. Ala Al-Fuqaha
Dr. Robert Trenary
Dr. Ajay Gupta
Dr. Ikhlas Abdel-Qader
The focus of this research is the application of the Artificial Immune System (AIS) paradigm to a new research area along with the modifications necessary to adapt it to a new problem. In the past 10 years, there has been much research into the use of various Machine Learning (ML) algorithms in Network Flow Traffic Classification. AIS algorithms have thus far not been applied to this problem. Because AIS algorithms have been used extensively for Network Intrusion Detection applications, which is a similar area of research, the motivation to extend them to the network flow classification problem is clear.
This research also shows a technique for faster execution of the training and classification portions of an AIS algorithm, which are meant to speed-up the execution of the AIS algorithms and adapt them to resource-constrained environments. Additionally, the research performed for this study seeks to expand the knowledge available about the behavior of Artificial Immune System algorithms. Specifically, the effect of several different distance functions as well as different kernel functions on the accuracy of the AIS classifier. The optimization is also applied to the class of algorithms known as Negative Selection Algorithms (NSA).
This study includes a survey of the network traffic classification literature. It also contains a presentation of the history of Artificial Immune System algorithms, their inner workings, and their previous applications. Furthermore, the reasoning for applying this type of algorithm to the network traffic classification problem is explained. Finally, the performance of the algorithm described in this study is analyzed by giving its big O complexity as well as a bound for its generalization error.
Schmidt, Brian Haroldo, "Artificial Immune Systems: Applications, Multi-Class Classification, Optimizations, and Analysis" (2017). Dissertations. 3096.