Performance of the Convolutional Neural Network Model in Identification of the Wireless Frequency Bands

The appearance of many security vulnerabilities is also increasing today with the widespread use of the wireless frequency band. This situation brings with it the research and development of various security measures as a necessity. Identification of wireless signals and transmitters properly in the environment becomes crucial situation. The efficient achievement of this process directly on the end devices also becomes necessary for the process to be fast. The purpose of this paper is to show the performance of deep learning methods for classification of wireless signals by making use of the software-defined radio devices. In particular, it is aimed to implement real-world applications and evaluate the results by utilizing software-based radio platforms. We investigated the performance of the classification made by deep learning algorithms on the I/Q data samples as a wireless signal representation that obtained on the receiver device by collected from transmitter devices. As an experimental result of the study, it has been seen that successful results are seen up to %80 in the detection of radio signals of the CNN model, which is especially efficient in image processing and identification applications. It has also been seen as a practical result that the increase in the number of layers in the CNN network model is not directly related to the performance of the network.