Deep Learning Based Secure MISO Transmission

This paper investigates the secrecy parameter of the Multiple Input Single Output(MISO) system based on Unsupervised Deep Learning using an autoencoder fitted with a single antenna Eavesdropper. We worked on co-occurring physical layer illustration optimization via encryption and decryption procedures, having an end-to-end scenario of a transmitter having two antennas and a receiver having one antenna. During the work, the road of thinking was extended to the training schemes that performed more accurately than current-day schemes. In the communication scenario, a confusion matrix is given as an input to the Eavesdropper to keep it ignorant concerning the communication. The secrecy is achieved by optimizing the loss function using cross-entropy applied on the Rayleigh fading channel; thus, balance is maintained between a reliable association and data secrecy using spatial diversity and a confusion matrix. Therefore the Eavesdropper with higher noise is not able to decode the symbols correctly.