Wireless technology identification using deep Convolutional Neural Networks
Naim Bitar, Siraj Muhammad, Hazem H. Refai · 2017
With the proliferation of wireless technologies and the ever-increasing growth in Internet of Things (IoT) devices operating the license-free Industrial, Scientific, and Medical (ISM) band, intelligent access systems capable of coexisting in crowded spectrum regions are of vital importance. In this work we study the adaptation of Convolutional Neural Networks (CNNs) to the problem of identifying coexisting wireless devices. We develop a machine learning conduit to facilitate the detection and identification of frequency domain signatures for 802.x standard compliant technologies. Spectrum scans across the entire ISM region (80-MHz) are recorded and a data-driven training process for a wide range of Signal-to-Noise Ratios (SNRs) is completed. Model accuracy is compared to that attained using standard feature based classification methods. Results indicate CNN models outperform their counterpart methods in terms of classification accuracy, connoting them to be highly effective tools for detecting and identifying coexisting devices despite acute overlap and interference presence. The proposed approach aims to advance cognitive wireless awareness by enhancing automatic detection and identification accuracy.