Predicting Jamming Systems Frequency Hopping Sequences Using Artificial Neural Networks

This paper proposes a neural network architecture that was designed to predict and reverse engineer frequency hopping jamming systems. The neural network was trained for frequency hopping sequences that use maximum-length sequences that utilize minimal polynomials as the primitive polynomial used in the linear-shift feedback register. This information is then used to generate a hopping sequence that reduces the jamming interference to 0 with as few as 4 jammer hopping samples. The model is also capable of determining if the jammer is utilizing a sequence that the model is trained for in as few as 25 jammer hopping samples.