A deep learning based covert communication method in internet of things

Communication security has played a more and more important role in modern wireless communication systems, especially in Internet of Things (IoT) for its privacy data security requirement. However, the IoT faces severe security threats due to the broadcast nature of radio propagation in wireless communication network. Therefore, physical layer security has been a hot topic nowadays which can effectively protect the private data transmission from jamming and eavesdropping at the physical layer of a wireless network. Inspired by the widely discussed deep learning based wireless communications, this paper adopts the widely used autoencoder framework for covert communications in IoT, which aims at waveform hiding for physical layer security. This proposed method employs deep complex neural networks (DCNNs) between legitimate users in IoT to jointly perform modulation, synchronization and demodulation. The generated covert signal produced by the DCNNs presents Gaussian statistics on both time and frequency domains. Therefore, the communication security is strongly guaranteed due to the difficulty of unauthorized detection and decoding for eavesdroppers. Moreover, computer simulations under single-user and multi-user cases demonstrate the effectiveness of this proposed deep learning based covert communication method, and the symbol error rate performance shows the superiority of our proposed method.