Efficient UAV Physical Layer Security based on Deep Learning and Artificial Noise
This addresses eavesdropping risks in UAV communications, but it is incremental as it combines existing deep learning and noise techniques for a specific security application.
The paper tackles secure image transmission in UAV networks by using autoencoder CNNs for compression and artificial noise to reduce eavesdropper channel capacity, achieving an MSE below 0.05 for authorized nodes with SNR under 5.
Network-connected unmanned aerial vehicle (UAV) communications is a common solution to achieve high-rate image transmission. The broadcast nature of these wireless networks makes this communication vulnerable to eavesdropping. This paper considers the problem of compressed secret image transmission between two nodes, in the presence of a passive eavesdropper. In this paper, we use auto encoder/decoder convolutional neural networks, which by using deep learning algorithms, allow us to compress/decompress images. Also we use network physical layer features to generate high rate artificial noise to secure the data. Using features of the channel with applying artificial noises, reduce the channel capacity of the unauthorized users and prevent eavesdropper from detecting received data. Our simulation experiments show that for received data with SNR fewer than 5 in the authorized node, the MSE is less than 0.05.