Utilization of Machine Learning to Optimize Radio Frequency Interference Identification for U.S. Naval Communications
Abstract
The proliferation of electronic devices emitting radio waves has led to Radio Frequency (RF) spectrum congestion. This poses a significant threat to Department of Defense (DOD) environments, especially naval communications heavily reliant on satellite systems, which are susceptible to electromagnetic interference.The lack of sufficient interference identification and characterization capabilities further compounds the operational risks faced by naval units. This thesis investigates the utilization of machine learning (ML)techniques for interference detection in RF transmissions. With their advanced data analysis and pattern recognition capabilities, ML algorithms can enhance interference detection and mitigation. Two architectures,a basic autoencoder and Long Short-Term Memory (LSTM) autoencoder, were evaluated for their ability to identify anomalous RF data within a dataset. The research methodology involved generating RF data with varying Additive White Gaussian Noise (AWGN) levels in a basic transmission pathway. The ML models were trained using normal RF data and evaluated on their ability to detect and classify signals with and without interference. The results demonstrate that both the basic autoencoder and LSTM autoencoder models could effectively identify interference. The LSTM autoencoders achieved a success rate of about 99%, indicating their potential use as a solution to the capabilities gap for interference identification.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 2023
- Accession Number
- AD1213273
Entities
People
- Rorey E Garnett
Organizations
- Naval Postgraduate School