An encrypted end-to-end communication system based on DCGAN
An encrypted end-to-end communication system based on deep convolutional generative adversarial networks(DCGAN)is proposed to address the secure transmission problem in end-to-end learning-based wireless communication systems.The system consists of DCGAN and autoencoder(AE)based on convolutional neural networks(CNN).The encrypted end-to-end communication system can adapt to different channel types through adjusting the network structure and designing the parameters.Simulation results show that DCGAN can encrypt messages in various forms.The proposed system can realize encoding and decoding of input bits of arbitrary lengths with a good generalization capability.It obtains a bit error rate(BER)performance similar to those of the conventional digital modulation systems.Compared to the basic AE end-to-end communication system,the proposed system are more difficult to be eavesdropped and its signals are less likely to be deciphered,thanks to the encryption module at its transmitter side.
end-to-end communication systemgenerative adversarial networksautoencoderphysical layer security
万方数据
维普
