Continuous Radio Frequency Oddity Watchdog

To address the U.S. Navy’s need for an automated system that characterizes the radio frequency (RF) emitter behaviors and patterns of life for a geographic area with no to minimal operator intervention, Intellisense Systems, Inc. proposes to develop a new Continuous Radio Frequency Oddity Watchdog (CROW), based on a multidimensional autoencoder matrix. Specifically, the innovation in the development and training of the autoencoder matrix will enable automatic modeling of RF spectral patterns of life, with excellent frequency and temporal resolution. By comparing a learned RF model with real-time data, CROW will be capable of identifying and logging any observed signals with atypical signal constructions, unexpected modulation types, and novel frequency content. The software structure also will detect abnormalities in signal scheduling; CROW automatically alerts operators if an observed signal is occurring at unexpected times. The system is designed for integration with one or more RF receivers. Using more than one receiver in a detection network enables CROW to further develop a spatial model of the spectrum; and, thereby, identify signals with atypical origins. CROW will provide anomaly detection in congested RF environments, which directly addresses the Navy requirements for a spectrum management and monitoring solution. In Phase I, Intellisense will demonstrate the feasibility of CROW by developing the core autoencoder matrix algorithms; designing inference and continuous learning protocols, assuming a single channel receiver with <=500 MHz instantaneous bandwidth and 18+ GHz tuning range; and demonstrating the system’s capability to both learn spectral patterns and automatically bring an operator’s attention to unusual activities in a simulated RF environment. Intellisense will also develop a Phase II plan. In the Phase I Option, Intellisense will develop additional system components and assemble test environments to accelerate developments in Phase II. In Phase II, Intellisense plans to refine CROW for use with three to four receivers, geographically distributed over 100 square miles, and demonstrate the new model’s capability to detect anomalies over space, time, frequency, and signal type. Finally, Intellisense plans to build a prototype system, train the learned model, and then operate in real-time to identify spectral anomalies in a real congested RF environment.