Validation of Machine Learning Algorithm on the Intrusion Detection System (IDS) of Navy Smart Grid
Abstract
The U.S. Navy currently deploys a centralized smart grid that consists of several control systems to analyze energy consumption and utilize data to drive more efficient operations. Being interconnected as a system-of-systems through the internet interface, the smart grid faces threats from the cyber realm aimed at disruption, intelligence gathering and destruction. In this thesis, an intrusion detection system (IDS) is developed for the smart grid as the first line of defense to guard against cyber attacks. We study the architecture of that Navy Smart Grid that utilizes Supervisory Control and Data Acquisition (SCADA) for control and monitor operations. We build the IDS using a random forest machine learning algorithm that can classify and identify malicious traffic. We then compare our approach to two machine learning benchmarks, namely the k-nearest neighbor and Bayesian learning models. We train our algorithm on an open-source SCADA data set that closely aligns with the type of traffic the smart grid would transmit. Simulations run on MATLAB show the efficacy of each of the algorithms and its ability to accurately classify different network threats.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 2021
- Accession Number
- AD1164397
Entities
People
- Wee S. Ng
Organizations
- Naval Postgraduate School