Predicting network anomalies with deep sequence analysis

UNCG Author/Contributor (non-UNCG co-authors, if there are any, appear on document)
Saed SayedAhmed (Creator)
Institution
The University of North Carolina at Greensboro (UNCG )
Web Site: http://library.uncg.edu/
Advisor
Somya Mohanty

Abstract: Network attacks can be very costly to victims and due to the complexities in they disparate types of attacks they are also hard to detect/predict. Understanding the underlying network traffic is critical in the developing automated solutions which can prevent such attacks in future. Within our study, we develop data-driven machine learning approaches to detect and predict such attacks based on the traffic behavior. Our study compares the differences in detection versus prediction of attacks/network anomalies where we compare traditional machine learning models for detection to the developed approach of leveraging network traffic as sequences of states in order to predict future network behavior. We also provide a comprehensive comparison of the different approaches taken with a wide range of feature-sets, hyperparameters, and variables evaluated for detection and prediction accuracy.

Additional Information

Publication
Thesis
Language: English
Date: 2019
Keywords
Cyber-Security, Intrusion Detection Systems, Machine Learning, Sequence Modeling
Subjects
Computer security
Machine learning
Intrusion detection systems (Computer security)

Email this document to