SAJE: SATCOM Anti-Jam Exploration

This paper presents SATCOM Anti-Jam Exploration (SAJE), an effort to develop a machine-learned filter that protects a SATCOM signal of interest from multiple simultaneous jammers. Although several methods exist for mitigating specific jammers, few can protect against multiple simultaneous jammers of different types, and there is currently no single-antenna solution that protects a waveform from a wide variety of jammers. SAJE attempts to find that one-size-fits-all solution by leveraging machine learning as a non-traditional method of removing jammers from a signal of interest. Our results have shown that SAJE offers significant protection against jammers both in simulation and in a lab scenario. We can currently mitigate combinations of up to three simultaneous CW, FM, and BPSK jammers, as well as weaker matched-spectral jammers. To achieve this, we have developed a flexible and modular training process that trains a neural network to produce an un-jammed SATCOM signal when given a jammed signal. Within our training pipeline, we can easily change the signal of interest, jammers, and channel conditions we wish to emulate. We have taken extra care to faithfully represent the SATCOM signal, jammers, and channel effects in our simulation so that anti-jam performance translates well between simulation and the lab. The success of SAJE thus far demonstrates the promise of applying machine learning to the problem of SATCOM anti-jamming and the viability of using machine learning to develop an all-encompassing solution to protect a waveform from multiple kinds of jammers.