Date of Award
Fall 10-2023
Document Type
Dissertation
Degree Name
Doctor of Philosophy in Cyber Defense (PhDCD)
First Advisor
Austin O’Brien
Second Advisor
Ashley Podhradsky
Third Advisor
Varghese Vaidyan
Abstract
In the rapidly advancing field of space technology and defense, the imperative for Security by Design in autonomous spacecraft systems has never been more pressing. This dissertation presents a pioneering approach to adapting this principle, traditionally rooted in software engineering and cybersecurity, to the specialized domain of security-oriented autonomous spacecraft. It addresses a crucial gap between cyber and astronautical sciences, advocating for a proactive stance in security—anticipating and mitigating potential threats in the physical space environment rather than responding to incidents post-factum.
Central to this research is the development of an innovative computer vision anomaly detection system that integrates a convolutional autoencoder with a glowworm swarm optimization algorithm. This novel union is designed to refine anomaly detection methods and establish robust measures for true and false positive detections. The resulting artifact is a cornerstone of a broader, general-purpose computer vision system intricately woven into the spacecraft’s operational fabric, encompassing both flight software and remote agency.
The primary objective of this endeavor is to devise a specialized visual anomaly detection system tailored for autonomous spacecraft, enhancing their capability to conduct security and reconnaissance missions autonomously. By harnessing advanced AI-driven methodologies, the system is engineered to perform nuanced inspections within the unpredictable and complex environment of space, a vital component for preserving situational awareness and mission safety.
A focal point of the study is the system’s ability to monitor and detect anomalies vigilantly—be they natural phenomena or adversarial actions—that could imperil space operations. The proactive detection framework established by this system is poised to revolutionize the operational security of space assets, ensuring a preemptive defense mechanism against emergent and evolving threats. The implications of this research extend beyond immediate applications, as it lays the groundwork for future innovations in autonomous defense systems, potentially transforming the landscape of space exploration and security.
Recommended Citation
Lancelot, Jonathan F., "Convolutional Autoencoder Neural Network Design Evaluation for an Anomaly Detection Subsystem in Autonomous Spacecraft Computer Vision Systems" (2023). Masters Theses & Doctoral Dissertations. 477.
https://scholar.dsu.edu/theses/477