Neural Network-Based RF Device Identification with Constellation Pattern

We propose the identification of neural network (NN)-based radio-frequency (RF) devices to enhance the security of normal communication. An access point (AP) receives an RF signal from a device and down-converts it into baseband I/Q symbols, which are grouped into their corresponding constellation points to form a constellation pattern. This pattern has a unique feature of the device's inherent RF impairments and identified with the NN. Simulations verified the effectiveness of the device identification.