Anomaly detection in Operational Technology (OT) environment is a unique challenge, especially in shipboard environments and naval installations where space constraints and system complexity can limit the integration of additional security or fault detection mechansims.  Naval power systems are vulnerable to unauthorized intrusions and system manipulations, affecting missions. The premise of this work stems from the limited study or insight into understanding how physical intrusions into different OT components manifest in power data and how to assess the transient security of Naval power systems. 

This project uses Non-Intrusive Load Monitorm (NILM) and machine learning (ML) to study cyber assuredness for Navy power systems, focusing first on microgrid installations at Naval facilities and eventually transitioning to shipboard power systems. The objective of this research is to develop cost-effective, cyber intrusion/anomaly detection techniques and methods to secure a Navy power system from single point anomalies to down stream cascading effects.

This project is an interdisciplinary effort; we collaborate with various faculty across NPS and with Naval organizations, including EXWC, Port Hueneme. 

The following tasks are expected to be executed:

Phase 1

• Develop multiple unsupervised ML tools to classify different types of single point anomalies in multi-signal microgrids in the presence of noise using an unbalanced data set (NILM data set). Work with NPS MSEE and NPS MSCS students.
• In collaboration with NAVFAC EXWC, analyze these tools using data from the microgrid testbed at EXWC; develop test procedures and collect experimental data from the testbed to validate the models and tools. 
• Develop training modules for facility engineers to learn about varied cyber anomalies that can be induced in a microgrid and how early detection of these anomalies will improve system resilience.
• Publications in the form of conferecences and journals will be expected.

Phase 2--this will classified at the SECRET level

• Analyze CUI shipboard NILM data and apply ML strategies for anomaly classification. Develop methods on how to use ML to study power signature analysis on waveforms to classify anomalies as enemy-induced, non-enemy induced (component failure) or a combination of the two.

Postdoc Requirements

• Must have either a PhD in Electrical Engineering or Computer Science.
• Must have experience in power systems and/or machine learning
• Must have some peer-reviewed publications.
• Must have effective oral and written communications skills.
• Must have the ability to interact and collaborate effectively with a diverse set of faculty and graduate students.
• US Citizen with at least a SECRET clearance (or has the ability to obtain a clearance) preferred.
• Foreign nationals will be relegated to Phase 1 work only.