Abstract:

The traditional communication jamming methods usually use fixed jamming wave forms and make jamming decisions based on target signals features obtained during the communication reconnaissance, Which is difficult to deal with the situation that the target signal characteristics are unknown or the parameters change dynamically. An communication jamming waveform generation technology based on generative adversarial networks (GAN) is proposed to solve these problems. GAN are used to extract the potential features of target signal and generate the jamming waveform with the same features as the target signal. Firstly, Based on the introduction of GAN principle, the network design and learning rate optimization experiment are carried out to make GAN more suitable for the generation of time series communication jamming waveform. Then, the generalization experiment is carried out on communication signals with different types and parameters, and results showed that this technology works well on other communication signals. Finally, the comparision experiment of jamming effect of generated jamming waveform, Gaussian noise and optimal jamming waveform is carried out, and results showed that the jamming effect of generated jamming waveform can approximate the jamming effect of optimal jamming waveform when the jamming to signal ratio increases gradually.

Key words: communication jamming waveform generation, machine learning, generative adversarial networks (GAN)