This repository contains main codes for simulation of the paper "Deep Learning Driven 3D Robust Beamforming for Secure Communication of UAV Systems", which has been accepted by IEEE Wireless Communications Letters.

Python == 3.7.0, Tensorflow-gpu == 1.14.0

csi_gen.m: to generate the training data sets
Ray_channel.m: complement for csi_gen.m
Folder named data: to store the training data sets
Folder named c_e: to store the errored location of Eve
Folder named output: to store the trained network parameters and R_s
opt_tf.py: main .py file
fun_1_tf.py: complement to opt_tf.py
para_cor.txt：parameters and data sets corresponding illustration
f_G.py: to generate the corresponding beamforming vector
fig3.m: to plot Fig. 3 in the paper (2D scenario is similar, which is omitted)
fig5.m: to plot Fig. 5 in the paper opt_multiEve.py and multiEve_comp.m: for multiple Eves scenario

You need to create directories to store files before running.
Step1: run csi_gen.m, and data should be generated and placed into .\data\1~3
Step2: run opt_tf.py, you can switch data sets by changing X in path, e.g., './data/X/'，'./output/network_DSN/X/'
Step3: run f_G.py
Step4: run fig3.m
Step5: run fig5.m
Step6: run opt_multiEve.py and multiEve_comp.m (for multiple Eves scenario)

Fig3:

Fig5:

multiEve_comp:

• If you have any questions, please don't hesitate to contact me via [email protected].