ANOMALY DETECTION OF EMS HARDWIRED INFRASTRUCTURE USING SUPERVISED AND UNSUPERVISED ARTIFICIAL INTELLIGENCE MACHINE LEARNING
Loading...
Authors
Ries, Jeffrey S.
Subjects
microgrid
Marine Corps Air Station
MCAS
fifth generation technology
5G
non-standalone
NSA
distributed energy resources
DER
energy management system
EMS
Naval Postgraduate School
NPS
Sea Land Air Military Research
SLAMR
Marine Corps Air Station
MCAS
fifth generation technology
5G
non-standalone
NSA
distributed energy resources
DER
energy management system
EMS
Naval Postgraduate School
NPS
Sea Land Air Military Research
SLAMR
Advisors
Thulasiraman, Preetha
Date of Issue
2024-06
Date
Publisher
Monterey, CA; Naval Postgraduate School
Language
Abstract
The microgrid currently deployed at Marine Corps Air Station (MCAS) Miramar, California began operations in 2021. It is unique in its efforts to leverage a Verizon, fifth generation technology (5G), non-standalone communications network to provide connectivity between distributed energy resources and the MCAS energy management system (EMS). With this new technology comes additional risks in the form of cyber attacks. Therefore, novel approaches to combat this threat are necessary to protect vital energy assets. In this thesis, we discuss the development of anomalous traffic detection models that use an unsupervised machine learning autoencoder trained on benign data sets captured from an AT&T 5G cellular tower at the Naval Postgraduate School Sea Land Air Military Research facility and Raytheon hardwired Modbus network at the EMS. We created synthetic anomalies for each data set to test our autoencoder and assess its effectiveness at classifying these packets as anomalous or benign. F-score, accuracy, precision, and recall were used as performance metrics. Through experiments conducted with Python and TensorFlow, we demonstrate the autoencoder models can successfully be trained and tested using benign network data using carefully crafted synthetic anomalies. This research establishes a baseline of research for an autoencoder to be used as an effective intrusion detection system to demonstrate the utility and operability of unsupervised machine learning for use in a microgrid.
Type
Thesis
Description
Series/Report No
Organization
Identifiers
NPS Report Number
Sponsors
ONR, Arlington, VA 22217
Funder
Format
Citation
Distribution Statement
Distribution Statement A. Approved for public release: Distribution is unlimited.
Rights
This publication is a work of the U.S. Government as defined in Title 17, United States Code, Section 101. Copyright protection is not available for this work in the United States.