Shelby Hall Graduate Research Forum Posters

Download

Download Full Text (385 KB)

Description

Non-linear phase-space analysis models data represented as a graph transitioning between states in the time domain. By studying data transitions, we can predict the time a particular behavior occurs and classify the events (states) in a system. For example, we could classify neurological sensor data to determine if a person is asleep (state), or predict the direction in which a stock will move (transitions) based on micro trade patterns.

Previous research has demonstrated success in phase-space graphs in classifying malware, detecting network intrusions, and predicting seizures. However, the solutions either require calculating global graph features as inputs to a classifier, which results in information loss, or converting the graph into an image as inputs to convolutional neural networks (CNNs). The CNN solutions, however, require fixed-sized images, which limits the size of a graph.

This study proposed graph neural networks (GNNs) to analyze phase-space graphs without the limitations above. GNNs do not limit the graph complexity or size and do not require the upfront calculation of either global or local features, which is time prohibitive. Preliminary results of this research to power measurements from computer operating systems for rootkit detection, indicate that GNNs can obtain a high accuracy (99.6%) with substantially less training time than other methods. Future work includes examining the use of different GNNarchitectures and the effectiveness of the approach for similar problems such as epilepsy prediction and network intrusion detection.

Publication Date

3-2025

Department

Computer Science

City

Mobile

Disciplines

Databases and Information Systems | Graphics and Human Computer Interfaces | Numerical Analysis and Scientific Computing | Other Computer Sciences

Application of Graph Neural Networks with Phase Space Graphs

Share

COinS